FORTH Thé   PI’Q.Ci)iQG
ACORNSQFT


Copyright C) Acornsoft Limited l982
All Rights Reserved
No part of this book may be reproduced by any means without the prior
permission of the copyright holder. The only exceptions are as
provided for by the Copyright (photocopying) Act or for the purposes
of review or in order for the software herein to be entered into a
computer for the sole use of the owner of this book.
SECOND EDITION
ISBN 0 907876 05 6
Published by:
Acornsoft Limited S
4a Market Hill
Cambridge
CB2 3NJ
England
ATOM FORTH on cassette and System FORTH on disc are available from
Acornsoft.


FORTH THEORY AND PRACTICE
CONTENTS
1 About this Manual 1
2 About FORTH 3
3 Starting FORTH 7
4 Stacks of Arithmetic 9
5 FORTH Definitions 19
6 Conditionals and Loops 35
7 The Ins and Outs of FORTH 45
8 Tape Interface and Editor 55
9 Graphics 55
10 What  Does 69
11 Further Examples 77
12 Error Messages 89
Glossary of FORTH Words D 91
Appendix A Two's-Complement Arithmetic 117
Appendix B System FORTH 119
Appendix C Dictionary Entry Structure 121
Appendix D Memory Allocation 125
Appendix E Further Reading 129
Index 131


To Annette
Acknowledgments
This version of FORTH is based on the public domain publications
provided through the courtesy of the
FORTH Interest Group,
P.O. Box llO5,
San Carlos,
CA. 94070 g
The greatest common divisor routine of Section 6.4 is by R.L. Smith.
The random number routine of Section 8.2.2 is by J.E. Rickenbacker.
These routines appeared in FORTH DIMENSIONS Vol.2, pages 167 and 34
respectively. The factorial calculations of Section ll.2.l were
provided by David Johnson—Davies of Acornsoft.
I am grateful for much helpful advice, particularly on the Acorn
operating systems, given by David Johnson—Davies, and Joe Dunn of
Acorn Computers.
I also owe a special debt of gratitude to Chris Jordan of
Acornsoft who has had the unenviable task of ensuring that this manual
makes sense. .He has made many improvements to the text and has, I
think, done an excellent job.


1 Ab my M t 1
Although this manual is written to explain the operation of FORTH for
the Acorn ATOM, most of the contents are applicable to any version of
the FORTH language. The implementation, like most others for home
computer use, is based on the model produced by the FORTH Interest
Group.
Individual addresses of registers, subroutines etc. and the
details of the memory map will, of course, be different on different
machines, and the machine code sections are necessarily concerned with
the 6502 microprocessor. In the descriptions of FORTH code, only the
tape interface (Chapter 8) and the graphics (Chapter 9) are likely to
be significantly different in other versions.
In this manual all FORTH words are written in upper—case letters,
exactly as they are typed in and appear on the display. Since FORTH
may use any character that can be typed.cx1 the keyboard, there may
occasionally be confusion between a FORTH word and punctuation marks.
In any cases where such confusion may arise, FORTH words are placed in
angle brackets, e.g. <.> , <,> and <."> .
In examples which contain text, both typed at the keyboard and
produced by the computer, the underlined sections represent the
computer's output, for example:
2 3 + . 5 OK
All keyboard input must be terminated by pressing the RETURN key, and
this will not normally be shown explicitly.
All numbers appearing in the text will, unless otherwise stated,
be given in decimal base.
I


2


FORTH was invented around 1969 by Charles H. Moore. It was originally
created as a convenient means of controlling equipment by computer.
Most high level languages that can be used on mini and micro computers
(e.g. BASIC) are too slow for such control and the only other
alternative is to use machine code routines. These, however, are very
tedious to write and enter.
FORTH solves many of these problems by allowing fast—executing
programs to be written in a high—level language. It also has the very
great advantage on small systems of using very little memory for
program storage. One further advantage, which will become more
apparent as you use the language, is that FORTH encourages the writing
of well—structured programs.
The speed of FORTH is largely due to the fact that it is a
compiled language, so that the stored program is in a form very close
to machine code. Unlike most other compiled languages, however, FORTH
is interactive, which means that each new word can be tested as soon
as it has been entered. If it does not do what you want it can·be
changed immediately until you are satisfied.
Perhaps the most powerful feature of FORTH is that it is an
extensible language. When you define a new word in FORTH, it becomes
an integral part of the language and can be used to produce further
definitions, in exactly the same way as the words resident in the
basic system. This allows the production of short, neat solutions to
complex problems. I
You may be beginning to realise, from what has been said so far,
that writing programs in FORTH is very different from writing in
languages like BASIC. A FORTH program consists of a series of
definitions of actions, each represented by a 'word'. These words are
then combined in further definitions until the required action of the
whole program is represented by a single word. The program can then be
executed by typing this single word at the keyboard.
The procedure for writing a program in FORTH begins with a
specification of the overall action of the program. This is then
broken down into a sequence of small tasks and these, if necessary,
are further divided into simpler tasks. Eventually the tasks are
reduced to the point where each is very easy to write in FORTH code.
The program is then written, starting with these simple routines and
building back up to the full program. Testing can be carried out at
each stage, greatly reducing the chance of errors in the final
program.
As an example we can consider the task of controlling a domestic
washing machine. The whole program might be represented by the word
WASHING which could be defined as:
: WASHING
WASH RINSE DRY ;
3


The words WASH, RINSE and DRY could themselves be defined as:
: WASH
FILL HEAT SOAP AGITATE SPIN ;
: RINSE
FILL AGITATE SPIN ;
: DRY
SPIN SPIN ;
At the next lower level the words FILL and HEAT, for example, could be
written:
: FILL
TAP ON
BEGIN ?FULL UNTIL
TAP OFF ;
and
: HEAT
HEATER ON
BEGIN ?TEMPERATURE UNTIL
HEATER OFF ;
Coding could then begin with the definitions of the actions of the
words TAP, HEATER, ON, OFF, ?FULL etc. The action of each word would
be checked, with a simulation of the machinery and sensors of the
washing machine, until the program is completed by the definition of
WASHING.
This example illustrates that, in FORTH, the problem to be solved
at any stage is simple and well—defined. Note also that many of the
words appear several times; once a word is defined it may be used in a
number of different situations, greatly easing the programming load.
FORTH is an example of threaded code. The words in a FORTH program
can be imagined to be strung together like beads on a thread. From one
word the thread loops to pass through all the words in its definition
and, if necessary, further loops include the words of lower level
definitions. Ultimately the thread returns to the highest level word
of the sequence.
FORTH is actually implemented as indirect threaded code, where
each 'bead' on the thread is not the routine itself but the address of
the routine. In the dictionary, therefore, each word consists of a
list of the addresses of the words out of which it is built.
So far there have been many references to 'words' in FORTH, so it
is about time to define what can be used as a FORTH word.
A word is defined as any combination of characters, separated by
one or more spaces from another word. Any character that can be typed
on the keyboard, including non-printing characters and control codes,
is allowed. The only characters that can not be used in a FORTH word
are a space, which is reserved as a delimiter to separate successive
words, and a null (ASCII zero), which is used to mark the end of input
text. In ATOM FORTH a word may be of any length up to a maximum of 3l
characters.
4


The following are examples of valid words:
§ORTH
II
+!
EMPTY—BUFFERS
The following are not valid:
EMPTY BUFFERS (includes a space)
THIS-WORD-CAN'T-BE-USED—IN—FORTH (32 characters)
  .... 5


6


ATOM FORTH is supplied in pre—compiled form on cassette. All that is
needed to load the system is to place the tape in the cassette
recorder and use the monitor command:
*RUN "FORTH"
Loading will take about five minutes, after which the system will
respond with the sign-on message:
ATOM FORTH
OK
FORTH has its own operating system for saving to and loading from
cassette, so the monitor commands will not be used again, unless you
decide to use them. The only exception is if your program crashes, or
if you press the BREAK key, when control will return to the cassette
operating system. The following example illustrates the procedure to
return to FORTH.
Once FORTH is loaded and the sign-on message has appeared type in
the following:
: STARS BEGIN 42 EMIT 2 SPACES AGAIN ;»
After you press the RETURN key, FORTH will respond with OK. This is
the standard response to all correct operations. If you do not get the
OK response, check if you have typed the example in correctly - it is
important to leave at least one space between each word.
You have caused a new definition named STARS to be entered in the
FORTH dictionary. This can be checked by typing:
VLIST
(and then RETURN) which will give a list of all the words present in
the dictionary. The listing may be stopped at any time by pressing the
ESC key. It may then be restarted by pressing the space bar — any
other key will abort the listing and return control to the keyboard.
The first word in the list will be STARS, showing that it is now
present in the dictionary.
When the message OK appears (either after the listing is complete
or when you have aborted the listing by, for example, pressing the ESC
` key twice) execute the word by typing STARS (don't forget the carriage
return) and the system will type an endless display of stars. You will
obtain no response from any key except BREAK, since an endless loop is
being executed.
Press the BREAK key to get the BASIC prompt '>' and then type:
LINK #2804
when the sign-on prompt will again appear. Typing in:
STARS
will give the same response as before, showing that STARS is still
present in the FORTH dictionary. This is an example of a 'warm' start,
in which all current dictionary entries are retained. The warm start
7


entry point to FORTH is at hexadecimal address 2804.
Repeat the sequence of executing STARS and pressing the BREAK key,
but this time re-enter FORTH by typing:
LINK #2800
The sign-on prompt will appear again but this time you will find that
a VLIST no longer includes the word STARS, showing that it is not in
the dictionary. An attempt to execute STARS will give:
STARS ? STARS MSG # 0
Error message 0 is given whenever a word is not recognised by FORTH.
Restarting FORTH at hexadecimal 2800 does a 'cold' start which forgets
everything except the nucleus dictionary. A cold start can be
performed from within FORTH by typing the word COLD, and a warm start
by typing the word WARM.
When either a cold or warm start is executed the sign-on prompt is
printed and FORTH is entered in the following state:
Numeric conversion base: DECIMAL
CURRENT vocabulary: FORTH
CONTEXT vocabulary: FORTH
Computation stack: cleared
Return stack: initialised
  (


4 sta 1< ofA in r'
4.1 Stacks
Most high level languages use one or more stacks for their internal
operations, e.g. for storing intermediate values during the
calculation cnf the result of ami arithmetical expression. Languages
such as FORTRAN and BASIC are designed so that the user needs no
knowledge cnf the internal structure cnf the computer, and they
therefore keep the stacks well out of sight.
A FORTH programmer has direct access to the stack with full
control of the values stored and their manipulation. Most words in
FORTH will place values on the stack or expect to find values there.
It is essential, therefore, to understand the structure and operation
of stacks.
The type cd? stack used by FORTH is one known as a last-in
first-out (LIFO) stack where the value most recently placed on the
stack is the one that is most accessible. The action is similar to the
pop-up pile of plates that is sometimes seen in restaurants. If a
plate is placed on the top of the pile it moves down until the new
plate is at counter level. When a plate is removed the pile rises so
that the plate which was underneath becomes the new top of the pile.
Because of this similarity the structure is also known as a push-down
stack.
Here's an illustration of the action of a LIFO stack:
TOP—> 27 TOP—> I5 TOP-> -3
-3 27 I9
19 -3 4
4 l9
4
a) b) c)
(a) is the initial state of the stack, and (b) is the state after the
value 15 has been 'pushed' onto the top of the stack. (c) is the final
state after the values I5 and then 27 have been ‘popped' from the top
of the stack.
This is the conventional view of the LIFO stack, in which the top
of the stack is always found at the same memory location. The contents
of the stack are moved to make room for a new top item, or to replace
an item that has been removed.
· This would be very slow in operation because of the need to move
the entire contents of the stack for each addition or removal. A more
efficient method is tx> leave the contents in the same positions in
memory and then change the pointer to the top of the stack when items
are added or removed. In FORTH it is convenient to make the stack grow
downwards in memory so that the 'top' of the stack is at the lowest
memory location used by the stack contents.
9


The scheme appears like this:
4 4 4
19 19 19
-3 -3 TOP—> -3
TOP-> 27 27
TOP-> 15
(a) (b) (c)
In this manual all descriptions will use the conventional wording, so
that the 'top' stack item is always the one that is most accessible.
In FORTH the top item of the stack is found at an address given by
the variable called SP (stack pointer). Each single-precision item in
the stack is stored as a 16-bit number, using two bytes of memory.
Thus the top item on the stack is found at address SP, the first item
from the top is at SP+2, the second from the top at SP+4, etc. The
address of the Nth item from the top of the stack is simply SP+2*N.
FORTH uses two stacks known as the 'computation stack' (sometimes
called the 'parameter stack') and the 'return stack'. The programmer
will generally use only the computation stack. This stack is used for
all arithmetic operations and to transfer information from one FORTH
word to another in the execution of a program. In this manual the
computation stack will be referred to as 'the stack' unless confusion
may arise. _
The return stack is mainly used by the system:
a) to store the address of the routine to which control is
returned after execution of the current word,
b) to store the current loop index in a DO ... LOOP .
In addition the return stack may be used, with caution, by the
programmer as a temporary store for values from the computation stack.
This is one possible method of gaining access to a stack value which
is not at the top of the cmmputation stack. In ATOM FORTH the
computation stack can hold up to 36 single—precision numbers, and the
return stack up to 44 addresses or single—precision numbers.
4.2 Arithmetic I
Arithmetic in FORTH is performed on integers rather than
floating—point numbers. There is ru> reason why floating—point
arithmetic should not be used but this would reduce the operating
speed. The integer operations in FORTH are designed to allow fast and
accurate arithmetic, without the need to use a floating—point format.
It has been said that if you need to use floating-point arithmetic in
FORTH, you do not fully understand your application! This is rather an
extreme viewpoint but makes the point that there are very few problems
that cannot be solved by the use of FORTH's integer operations.
All the arithmetic operators in FORTH expect to find their values
on the stack and replace them by their result. A consequence of this
is that the numeric values must be placed on the stack before the
operator is used.
10


Thus to add the numbers 2 and 3, the following sequence should be
typed at the keyboard:
2 3 +
where 2 places the number 2 on the stack
3 places the number 3 on the stack
+ removes the top two items from the stack, adds them and places
the result on the stack.
The FORTH word <.> removes the top item from the stack and prints it
on the display so the following result should be found
2 3 + . 5 OK
(Don't forget to type 'RETURN' after the <.>)
Placing the operator after the numbers on which they act is known as
postfix, or reverse-Polish, notation and will be familiar to anyone
who has used a Hewlett—Packard calculator. The normal method of
writing arithmetic operations is luunni as infix notation. One
advantage of using postfix notation is that there is no need to use
brackets to indicate the order of evaluation as the order is
completely unambiguous.
4.3 Single—Precision Operations
4.3.1 Single—Precision numbers
In FORTH, single-precision numbers are of 16 bits (2 bytes) with the
most significant byte at time lower address. Unsigned numbers are in
the range () to 65535 inclusive. Signed numbers are stored in two's
complement form and are in the range -32768 to +32767 inclusive (see
Appendix A).
A number may be placed on the stack simply by typing it at the
keyboard and following it by RETURN. The top stack item may be removed
from the stack and printed on the VDU by the use of the word <.>
(dot). This word interprets the number as a signed integer. To show
the action of <.> , try the following examples:
l7 . l7 OK
-2l . -2l OK
32767 . 32767 OK
32768 . -32768 OK
In single precision, numbers greater than 32767 are interpreted by <.>
as being negative. 3
Numbers greater than 32767 can be printed as unsigned integers
using the word  .
32768 U. 32768 OK
4.3.2 Single—Precision Arithmetic
FORTH does not provide an exhaustive set of arithmetic operators,
since the needs of different applications vary widely. There is,
however, a sufficiently large range of general purpose operators so
that any required operation can be defined by the user.
The following list contains all the single-precision arithmetic
operators provided in FORTH. In the stack action the notation is
ll


(stack before ... stack after) with the top of the stack to the right,
and the items separated by '\':
WORD Stack action Description
+ (nl\n2 ... sum: nl+n2) Add
- (nl\n2 ... difference: nl—n2) Subtract
* (nl\n2 ... product: nl*n2) Multiply
/ (nl\n2 ... quotient: nl/n2) Divide (integer)
MOD (nl\n2 ... remainder) Remainder of nl/n2
/MOD (nl\n2 ... rem\quotient) Leave quotient with remainder
beneath
*/ (nl\n2\n3 ... nl*n2/n3) Intermediate product nl*n2 is
stored in double precision
*/MOD (nl\n2\n3 ... rem\nl*n2/n3) As */ but also leave remainder
beneath
MINUS (nl ... —nl) Change sign
ABS (nl ... |nl|) Absolute value
+— (nl\n2 ... n3) Leave, as n3, the value of nl
with the sign of n2
l+ (nl ... n2) Add l to the top stack item
2+ (nl ... n2) Add 2 to the top stack item
2* (nl ... n2) Fast multiply by two
The following list gives examples of the nun; of the first four of
these in postfix notation, compared with the corresponding infix form:
Infix Postfix
2 * 3 2 3 *
9 / 4 9 4 /
2 — (3 * 5) 2 3 5 * —
(2 + 3) * 5 2 3 + 5 * (or 5 2 3 + *)
If you are not familiar with postfix notation you may find it useful
to try these examples at the keyboard, using the <.> word to print the
result. Try a few examples of your own, using <.> to check if the
result is what you expected. If the operators run out of numbers to
work on, FORTH will give error message number one (empty stack) but
there will be no indication if too many numbers are left on the stack
at the end. Once you have completed a calculation keep using <.> until
error message one is given, to make sure that there are no numbers
unexpectedly remaining.
FORTH can be used to calculate a formula, such as the value of the
quadratic expression
3x2 — 5x + 4
for various values of x. If, for example, x has the value 2 the result
could be found as follows:
2 2 * 3 * 2 5 * - 4 + . 6 OK
12


This involved typing in the value of x (2) in three places. It can be
improved upon by using the stack operators described in the following
section.
Try using the other words in the list. Use each one with a range
of numerical values, both large and small, positive and negative, to
become familiar with their actions.
4.3.3 Single—Precision Stack Operators.
There are several words in FORTH which act directly on the numbers on
the stack. These words are given in the following list:
WORD Stack Action Description
DROP (n ...) Remove the top stack item.
DUP (n ... n\n) Duplicate the top item.
-DUP (n ... n\n) or (n ... n) Duplicate the top item if it is
non—zero, otherwise do nothing.
OVER (nl\n2 ... nl\n2\nl) Copy the second item over the top
item.
SWAP (nl\n2 ... n2\nl) Exchange the top two items.
ROT (nl\n2\n3 ... n2\n3\nl) Rotate the top 3 items, so that the
third item moves to the top.
There are also two words which act on numbers further down the stack.
These are:
a) PICK — used as n PICK to make a copy. on the top of the stack, of
the nth number in the stack. For example,
l PICK is equivalent to DUP
and 2 PICK is equivalent to OVER.
b) ROLL - used as n ROLL to rotate the top n items on the stack,
bringing the nth item to the top. For example,
2 ROLL is equivalent to SWAP
and 3 ROLL is equivalent to ROT.
These two words are useful to extract a needed number that is some
depth below the top of the stack, but are relatively slow in their
operat*on and should be used sparingly.
A further three words act to transfer numbers between the
computation stack and the return stack (see Section 4.l). These are:
R Copy the top item of the return stack to the computation stack.
The return stack is unchanged.
>R Transfer the top item of the computation stack to the return
stack.
R> Transfer the top item of the return stack to the computation
stack.
Since these last two words modify the contents of the return stack,
which is used for system control, they should be used with caution.
They should never be executed directly from the keyboard and, within a
definition, they should normally be used only as a pair. This will
l3


ensure that the state of the return stack is unchanged between entry
and exit when the new definition is later executed. The main use of >R
and R> is as a temporary store for the top value on the computation
stack when a calculation needs to use the number(s) below it.
The stack contents may be manipulated by the use of several of the
stack operators in succession. The following list includes a number
of useful stack manipulations which require two words:
Stack before Stack After Words
l 2 l l 2 A OVER SWAP
l 2 2 l 2
l 2 2 l l
l 2 l 2 l l
l 2 3 2 l 3
l 2 3 3 2 l
It is useful practice to work out the solutions for this list. If you
are not sure your solution is correct, try it out at the keyboard.
Remember that <.> prints the top of the stack first so that the
correct response for the first of these is:
l 2 OVER SWAP . . . 2 l l OK
If we now return to the earlier problem of calculating the value of
the quadratic function:
3x2 — 5x + 4
we can see that it is possible to perform the calculation in such a
way that the value of x needs to be typed once only. The following
example shows how this could be done, and is broken into several
sections so that the stack contents can be shown at each stage.
Remember that the top stack item is on the right.
Stack contents
2 2
DUP DUP 3 2 2 2 3
* * 2 l2
SWAP 5 l2 2 5
* l2 l0
— 4 2 4
+ 6
The advantage of this is that everything except the value of x can be
made into a definition (see Chapter 5):
: QUADRATIC
DUP DUP 3 * *
SWAP 5 * - 4 + ;
This can then be used with many values of x. It expects to find the
value of x as the top stack item and replaces it by the value of:
3x2 - 5x + 4 4
l4


4.3.4 Relational and Logical Operators
Most of the relational operators in FORTH apply a test to the top one
or two stack items, returning a true or false value, depending on the
result of the test. A false result is indicated by zero and a true
result by a non—zero value being left on the stack. As usual the words
replace the arguments with the result, in this case a true or false
flag.
The relational operators provided in FORTH are listed below;
unless otherwise stated they all act on signed numbers:
0= Leave true if the top stack item is zero; otherwise false.
0< Leave true if the top stack item is negative.
= Leave true if the top two stack numbers are equal.
< Leave true if the second stack item is less than the top item,
for example; 2 3 < leaves true. .
> Leave true if the second stack item is greater than the top
item, for example; 3 2 > leaves true.
U< As < , but the two numbers are treated as unsigned integers.
MAX Leave the larger of the top two numbers on the stack.
MIN Leave the smaller of the top two numbers on the stack.
Note the difference between < , which compares two signed integers in
the range -32768 to +32767, and U<. They act identically on numbers in
the range 0 tx> 32767 but will give different results outside this
range.
The action of U< is to subtract the two numbers and examine the
sign of the result. This should be used carefully, since if the
difference between the two numbers is greater than 32767 the result
will be interpreted as being negative and gives an apparently wrong
result. The action of U< is:
Second stack value Top stack value Result of U<
0 32767 l
32767 32769 4 l
32769 62769 l
0 32769 0 r
The logical operations in FORTH usually act on the top two numbers on
the stack and are:
AND Leaves a bit—by—bit logical AND of the top two stack
I numbers.
OR Leaves a bit—by—bit logical OR of the top two stack
numbers.
XOR Leaves a bit—by—bit logical EXCLUSIVE-OR of the top two
stack numbers.
TOGGLE Performs a bit—by—bit EXCLUSIVE—OR of the low order byte of
the top stack number with the byte whose address is second
on the stack. The result is replaced at this address.
One application of XOR is to determine the sign of the product of two
numbers and is used in this way for many of the multiplication words
in FORTH. A negative number has the most significant bit set to l. The
exclusive—OR of two numbers of the same sign (i.e. whose most
i l5


significant bits are both 0 or both 1) will be a number with a zero
most significant bit, indicating a positive result. With two numbers
of opposite sign the exclusive—OR will leave a number with most
significant bit 1, showing the result to be negative. Note that the
value of the result has no meaning in this context. An example of this
use of XOR is the definition of MD* in Section 4.4.
The use of TOGGLE is illustrated in the definition of SMUDGE
(defined in hexadecimal base):
: SMUDGE LATEST 20 TOGGLE ;
LATEST returns the address of the name header of the most recently
defined word in the dictionary, and 20 TOGGLE changes the 'smudge' bit
in the header to allow or prevent the word from being found in a
dictionary search. This is discussed more fully in the description of
CREATE in Chapter 5.
4.4 Higher Precision Arithmetic
In addition to single precision, FORTH also supports double-precision
arithmetic. Double—precision numbers are stored in 32 bits, using four
successive bytes of nmmory, with the least significant byte at the
lowest address. Again, two's—complement form is used, giving a range
of values from -2l47483648 to +2l47483647 inclusive.
Note that, because of the way the LIFO stack is implemented in
FORTH, a double precision number on the stack has its high order part
'above' the low-order part.
A double-precision number may be entered from the keyboard by
including a decimal point anywhere in the number.
Thus typing in:
12.
583.2478
will place the numbers 12 or 5832478 on the stack, in double precision
form. Note that the position of the decimal point has no effect on the
way in which the number is stored. Thus
123456.
1234.56
123.456
.123456
will all be stored on the stack as the double-precision integer
123456. The number of digits to the right of the decimal point is,
however, stored in the user variable DPL and may be used, for example,
to control the format of numeric output. When a single-precision
number is input from the keyboard the value in DPL is always set to
-1.
There is one purely double-precision arithmetic word and this is:
D+ Double—precision add
In addition there are five mixed—precision operators:
M* Multiply two signed single—precision numbers to give a
signed double-precision product.
U* As M* , but all numbers are unsigned. This is the
multiplication primitive (machine code).
16


M/ Divide the double-precision number second on the stack by
the single-precision number on the top. A single-precision
quotient is left and all numbers are signed.
U/ As M/ , but the remainder is left beneath the quotient and
all numbers are unsigned. This is the division primitive.
M/MOD As U/ , but leaving a double-precision quotient. Again all
quantities are unsigned.
There are also three sign-changing words for double-precision numbers:
DMINUS Change the sign of a double-precision number.
DABS Leave the absolute value.
D+- Apply the sign of the single-precision number on the top of
the stack to the double-precision number beneath.
Finally, there are two stack operators in double precision:
2DROP Remove the double-precision top stack item.
2DUP Duplicate the double-precision top stack item.
Note that these two words can also be usd to act on the top two
single-precision numbers, i.e ZDROP is equivalent to DROP DROP and
2DUP is equivalent to OVER OVER.
There are ru> relational or logical operations provided for
double-precision numbers.
As an example of the use of some of the double-precision words
consider the following definition of the word MD*. It also illustrates
the use of XOR to determine the sign of a product, as discussed in
Section 4.3.4.
The word MD* performs a mixed-precision multiplication which
leaves the signed double-precision product nd2 of the signed
double-precision number ndl and the signed single-precision number n:
: MD* ( ndl\n ... nd2 )
2DUP XOR >R ( keep sign of product )
ABS >R DABS R> ( modulus of multiplicand & multiplier )
DUP ROT * >R ( high order product )
U* R> + ( low product, plus high product )
R> D+- ( apply sign to product )
This word is used in the factorial routine in Chapter ll to allow the
calculation of the factorial of numbers up to l2.
l7


l8


5 FORTH Def` Y
5.1 Introduction
Programs written in FORTH are usually, and more accurately, known as
x applications. The idea of a program implies the generation of a
sequence of actions, distinct from the set of instructions which form
the language in which the program is written. In FORTH the distinction
between the language and the 'program' is far less clear. The sequence
of actions are created as additional words in the FORTH vocabulary and
can be used in exactly the same way as the original words, to produce
a more complex process. In effect the language is simply being
extended. Many people argue that FORTH should not be described as a
language since it contains no rules or structures that cannot be
changed by the user. Whether this is a valid argument or not, the fact
remains that the great power of FORTH lies in its ability to be
extended to cope with any situation that may arise.
There are several ways in which new words may be placed in the
dictionary. These use defining words, of which the most common are:
: (colon)
CONSTANT
VARIABLE
USER
CREATE
VOCABULARY
The exact formats of words created by each of the above are given in
Appendix C. It will, however, be useful to give here a general
description of the construction of a typical dictionary entry.
All dictionary entries consist of two parts, the head and the
body. The head contains:
a) The name of the entry (variable length),
b) a link pointer to the name of the previous entry,
c) a code pointer to the machine code used in the execution of
the entry.
The starting addresses of these fields are known as the name field
address, the link field address, and the code field (or execution)
address respectively.
The body of the entry, also known as the parameter field, contains
the information which defines the action of that particular entry. The
nature of this information differs according to which defining word
was used jJI its creation. For E1 colon-definition, for example, the
parameter field contains a list of the execution addresses of the
words in the definition, terminated by the execution address of <;S>
which causes an exit from the word.
l9


The diagram illustrates these points for a dictionary entry
created by a colon-definition.
Name Field Address NAME
Link Field Address LINK POINTER -—> to previous
C name field
Code Field Address CODE POINTER --> to machine code for
a colon-definition
Parameter Field EXECUTION ADDRESS l
Address {
. EXECUTION ADDRESS 2
S EXECUTION ADDRESS OF ;S
In this and all such diagrams in this manual, memory addresses
increase from top to bottom.
There are za number of words supplied which allow the address of
one of these fields to be converted into the address of another. The
initial address is expected on the stack, and it is replaced by the
new address. The words are as follows:
Word Action
PFA Convert the name field address to the parameter field address. i
CFA Convert the parameter field address to the code field address.
LFA Convert the parameter field address to the link field address.
NFA Convert the parameter field address to the name field address.
There are no words provided for the conversion of the code field or
link field addresses. This is not normally a problem since any search
for a word will return either the parameter field address or the name
field address. If such a conversion is required it is, however, very
simple since
code field address = link field address + 2
parameter field address = code field address + 2 ‘
5.2 Colon-Definitions F
5.2.1 Form /
The colon-definition is the most—frequently used way of defining a new
action in FORTH and has been used in several of the examples in
earlier chapters. The form of a colon-definition is:
: NAME ..... ;
20


The colon indicates the start of the definition of a new dictionary
entry for the word NAME. The NAME is followed by a sequence of actions
in terms of FORTH words which have been previously defined. The
definition is terminated by a semicolon <;> . Once defined the word
can be executed by typing its NAME at the keyboard. Since
colon-definitions are used extensively throughout this manual, no
specific examples are given here.
5.2.2 Separating Applications
When starting to write a new application it is useful first to make a
null definition such as:
: TASK ; V
This is a word which has no function except to mark the start of the
application - executing TASK will do nothing. When the application is
no longer required, however, typing
FORGET TASK
will erase TASK and all subequently defined words, clearing the
dictionary for a new application.
5.3 CONSTANT, VARIABLE and USER
5.3.1 CONSTANT
Numerical values may be compiled into a colon definition as literal
values. An alternative is to define them as constants. The sequence:
l0 CONSTANT LENGTH
will create a dictionary entry for a constant with the name LENGTH and
value l0. The entry has the single precision value of the constant in
its parameter field and the code field contains a pointer to machine
code which will copy the value from the parameter field to the stack.
Thus, when LENGTH is later executed it will place the value l0 on the
stack, just as if the number l0 had itself been typed, i.e. typing:
LENGTH . will give l0 OK
There are two advantages in using constants rather than literal
values:
a) When used in a colon definition LENGTH will compile its two—byte
execution address, whereas the literal value requires four bytes -
two bytes for the address of the literal handling routine and two
bytes for the value. If the value is used many times there is a
net saving in memory space, despite the space needed for the
definition of LENGTH .
b) If it is necessary to change the value at some later time it is
simpler to change the definition of LENGTH , rather than every
occurrence of the literal value.
To change the value of a CONSTANT, the operators <'> and  are used.
<'> followed by the name of the CONSTANT leaves its parameter field
address on the stack.  uses two values from the stack; a numeric
2l


value with an address above it. It acts to store the numeric value in
the two bytes starting at the address. Thus
100 ' LENGTH 1
changes the value of LENGTH to 100, leaving the stack unchanged.
Typing:
LENGTH . will now give 100 OK
5.3.2 VARIABLE
A dictionary entry for a variable may be created by typing, for
example:
30 VARIABLE XLENGTH
This will create a variable with name XLENGTH and initial value 30.
The dictionary entry will contain the single precision value of the
variable in its parameter field.
The difference between a CONSTANT and a VARIABLE is that, on
execution, the CONSTANT places its value on the stack but the VARIABLE
places on the stack the address of the memory containing the value.
The value of the variable is returned by the <@> operator. This takes
the address from the stack, replacing it with the two—byte value
fetched from the corresponding location. Hence
XLENGTH @ `
puts the value of XLENGTH on the stack. This method is chosen so that
the storing of a new value in the variable is made simple by, for
example,
40 XLENGTH 1
which replaces the old value of XLENGTH by the new value, 40 .
The value of a variable can be incremented by the use of +1 e.g.
1 XLENGTH +1
will increment the value of XLENGTH by one. The increment may be of
any magnitude (subject to the valid range for single precision
numbers) and may be positive or negative.
The value of a variable can be printed by using . Its action
is as one would expect after:
: ? @ . ;
5.3.3 USER
This word is provided to allow system modifications, and will not be
used in most applications. A user variable may be created by, for
example, the sequence:
50 USER TERMINAL
In this case the value of the variable is not initialised. The above
sequence creates a new user variable whose name, TERMINAL , is stored
22


in the dictionary, but the value of the variable will be stored in a
separate user variable area. The number (50) is used as an offset in
the user area from the value of the user variable pointer, UP . The
above sequence will therefore reserve two bytes of memory, 50 bytes
above the start of the user area (see memory map).
The user area is reserved for system variables, many of which are
initialised on a COLD start of FORTH, and should not be used for
variables in an ordinary application. The user variable area provided
has its base at address 97C4 hex with a maximum offset of 3A hex (58
decimal). The first unused offset is 32 hex (50 decimal), allowing up
to five additional user variables to be defined.
The user variables provided in the system are:
TIB The start address of the terminal input buffer.
WIDTH The maximum width of a dictionary entry name, normally 3l.
WARNING Error message control, normally 0.
FENCE Lower limit for FORGET.
DP Dictionary pointer.
VOC-LINK Points to the most—recently defined vocabulary.
BLK If 0, input is from terminal, otherwise from tape buffer.
IN Current offset into the input buffer.
OUT No. of characters output (not used by system words).
SCR Current tape/disc screen number.
CONTEXT Vocabulary pointer for dictionary searches (see Section 5.5).
CURRENT Vocabulary pointer for new definitions (see Section 5.5).
STATE Indicates the compilation state, non-zero when compiling.
BASE Contains current numeric conversion base.
DPL Position of decimal point in last number from keyboard.
CSP Current stack pointer value - used in compiler security.
R# Pointer to the editing cursor.
HLD The address of the last converted character during numeric
output conversion.
There are in addition two user variables that are used by the system
but do not have name headers. They are:
RO The address of the start of the return stack
S0 The address of the start of the computation stack
They may be given headers by:
8 USER RO 6 USER S0
23


5.4 CREATE
CREATE is a word that will produce a dictionary entry consisting of a
name header, a link pointer, and a code field pointing to the start of
a blank parameter field. It is used by <:> , CONSTANT , VARIABLE , and
USER to generate their name headers.
The main use of CREATE in ATOM FORTH is to allow the definition of
new machine—code routines, without the use of a FORTH assembler which
would require loading from tape and take up valuable dictionary space.
In such a use we are performing a hand compilation of a dictionary
entry and make explicit use of the compiling words <,> and . The
word <,> takes a two—byte value from the top of the stack and places
it in the first two unused bytes of the dictionary (pointed to by the
user variable DP). Usually, as in the following example, this will be
in the parameter area of the dictionary entry which is being compiled.
The value of DP is incremented by two. The word  has a similar
action except that it places only the low byte of the top stack item
in the dictionary and increments DP by one.
As an example of the use of CREATE, consider the definition of
2DROP (which is provided in the nucleus dictionary):
HEX CREATE 2DROP 4C C, 29lE , SMUDGE
where p
HEX Sets the base to hexadecimal for the machine code.
CREATE 2DROP Generates the header for the word, with a code field
pointing to the following code. I
4C C, Compiles the single byte 4C into the parameter area.
29lE , Compiles the two byte address 29lE into the parameter
area.
SMUDGE Toggles the 'smudge' bit in the name field to allow the
new definition to be found in a dictionary search.
This word will execute the machine code
Machine Code: Assembler Mnemonic:
4C lE 29 JMP POPTWO
(POPTWO is the label of machine code to erase the top two stack items
and then execute NEXT, which moves on to the next word).
The dictionary entry produced by CREATE has the 'smudge' bit set
to prevent the possibility of a partially completed definition from
being found in a dictionary search, and it must be reset by the use of
SMUDGE before the new definition can be either executed or used in a
further definition. It will, however, appear in a VLIST in its smudged
or unsmudged state. SMUDGE always acts on the last definition in the
dictionary.
Note that an error in the use of a colon definition will leave a
partially completed entry, in a smudged state, which can be found by
VLIST but not used or forgotten. In this case using SMUDGE will then
allow you to FORGET the partial definition and start again.
All machine code routines must terminate with a jump to existing
machine code, which will, directly or indirectly, execute the code of
NEXT. Valid terminating jumps are:
24 U


Routine Hex Address Description
NEXT 2842 Transfer execution to the next word in the
sequence, or, if it is the last word, return
to the keyboard.
PUSH 283B Push the accumulator (as high byte) and one
_ ' byte from the return stack as a new number
on the computation stack and execute NEXT.
PUT 283D Replace the current top stack item from the
A accumulator and return stack and execute NEXT.
PUSHOA 2B21 Push zero (high byte) and accumulator (low
byte) to the computation stack and execute NEXT.
POP 2920 Drop the top stack item and execute NEXT
POPTWO 29lE Drop the top two stack items and execute NEXT.
The X-register of the 6502 is used by ATOM FORTH as the computation
stack pointer. If a machine code routine uses the X-register its
contents must be saved before it is used and restored before exit from
the machine code by one of the above jumps. One byte of memory at
address #8E in zero page is reserved as a temporary store for the
X—register, and is known as XSAVE. This and other reserved zero page
locations are given in the following table:
Name Hex address Format Comments
W 88 xXx 1+2 bytes, codefield pointer
IP 8A Xx 2 bytes, interpretive pointer
UP 8C Xx 2 bytes, user area pointer
XSAVE 8E X 1 byte, temporary for X-register
N .90 xXxxxxxxx 1+8 bytes, scratch pad
Note that W and N both use one extra byte before their stated
addresses.
25


The significance of UP (see Section 5.3.3) and XSAVE have been
discussed. To explain the functions of IP and W, consider that the
word NAME in the definition of CCCC is being executed, as shown:
NAME
LINK POINTER
+-9 CODE POINTER
CCCC
LINK POINTER
CODE POINTER
W -—+ EXECUTION ADDRESS
OF NAME
IP —-—— > EXECUTION ADDRESS
OF NEXTNAME
W is the code field pointer and at this time contains the address
of ('points to') the code pointer of NAME . IU? is the interpretive
pointer and holds the address of the next instruction in CCCC . When
the execution of NAME is complete the contents of the location pointed
to by IP are transferred to W. This means that W will now contain the
address of the code pointer of NEXTNAME . The value in IP is then
incremented by two so that it again points to the instruction
following the current one. An indirect jump to W will then start the
execution of the code for NEXTNAME . This sequence of actions is
performed by the machine code of NEXT .
The scratchpad area N is a nine—byte area into which up to four
stack values can be transferred by use of the subroutine SETUP at
address #2863. This expects to find the number of l6—bit values to be
transferred in the accumulator. It uses the Y register and will return
26


with zero in Y and the value in A doubled.
As an example of its use the sequence:
A9 O3 LDA @3
20 63 28 JSR SETUP
will transfer the top three stack items into N. Each stack item
occupies two bytes of N with its low-order byte at the lower address.
The top stack item is first in the list. On return from the
subroutine, Y will contain 0 and A.wdll contain 6. The three items
will have been dropped from the stack. The byte immediately preceeding
N contains the number of bytes transferred (i.e. 6). This subroutine
is useful to place stack values at a known, fixed location for use by
machine code routines. Since this scratchpad area is used by many of
the system words, its contents will change frequently, so the contents
should only be used within the definition that placed them there.
The following examples are for words which exist in the nucleus
dictionary. If you enter them at the keyboard you will find error
message number 4 given, which is a warning that you are redefining an
existing word. This will not, however, affect their operation. The
examples are given in two forms, the first being how the routines are
entered and time second is aa conventional assembly language listing
with explanations.
HEX (all the following are in HEX base)
CREATE AND (logical AND (section 4.4.4))
B5 C, O0 C, 35 C, 02 C,
48 C, B5 C, Ol C, 35 C,
O3 C, E8 C, E8 C, 4C C,
283D , SMUDGE
B5 00 LDA 0,X load low byte, top of stack
35 O2 AND 2,X AND with low byte 2nd on stack
48 PHA save result on return stack
B5 Ol LDA l,X load high byte, top of stack
35 03 AND 3,X AND with high byte 2nd on stack
E8 INX drop the top
E8 INX stack value
4C 3D 28 JMP PUT replace (old 2nd) by result (as new top),
then execute NEXT
Note that <,> stores the two bytes of 283D in reverse order. Provided
this is remembered the amount of typing can be reduced, as in the next
example, particularly as leading zeroes need not be typed.
CREATE + ( single precision add )
l8 C, B5 , ( 'High' byte is zero )
275 , 295 , lB5 ,
375 , 395 , E8E8 ,
4C C, 2842 , SMUDGE
l8 CLC clear carry for addition
B5 00 LDA 0,X load low byte top of stack
75 02 ADC 2,X add low byte 2nd on stack
95 O2 STA 2,X store at low byte 2nd on stack
B5 Ol LDA l,X load high byte top of stack
75 O3 ADC 3,X add high byte 2nd on stack
95 O3 STA 3,X store at high byte 2nd on stack
E8 INX drop top of
27


E8 INX stack
4C 42 28 JMP NEXT execute NEXT
The next example uses the X—register which must therefore be saved and
restored as described earlier.
CREATE RP@ ( leaves the contents of the return stack pointer )
8E86 , 8ABA , 8EA6 ,
  [   [
4C C, 283B , SMUDGE
86 8E STX XSAVE save X—register
BA TSX stack pointer (low byte) to X
8A TXA and then to accumulator
A6 8E LDX XSAVE restore X—register
48 PHA push accumulator to (machine) stack
A9 Ol LDA @1 high byte of stack pointer is Ol
4C 3B 28 JMP PUSH push stack pointer to computation stack
and then execute NEXT “
5.5 VOCABULARY
A VOCABULARY is a subset of the dictionary and the VOCABULARY
structure of FORTH is the means by which the order of a dictionary
search is controlled. Normally if an existing word is redefined a
dictionary search will find only the latest definition. The old word
will still be used in earlier definitions but only the most recent
version will be available for new definitions. The following example
illustrates this:
: QUOTE ." THIS IS A LITERAL STRING" ;
: PRINT QUOTE CR ;
Executing PRINT will type out the message of QUOTE. The word CR
performs a carriage return and line feed on the display. If QUOTE is
redefined as:
: QUOTE ." A DIFFERENT MESSAGE"
MSG #4 will be given, warning that QUOTE is already in the dictionary.
When the new definition of QUOTE is made and used in:
: NEWPRINT QUOTE CR ;
NEWPRINT will type out the new nmssage. PRINT will, however, still
respond with the original message. Executing a VLIST will show that
there are now two entries for QUOTE .
Typing FORGET QUOTE and then executing a VLIST will show that the
second QUOTE ( and NEWPRINT ) will have disappeared from the
dictionary but the earlier definition of QUOTE will still remain. It
is necessary to type FORGET QUOTE a second time to remove both
versions from the dictionary. If the two versions of QUOTE are defined
in different vocabularies it is possible, by changing the dictionary
search order, to select which version will be used in further
definitions.
The order of search and the determination of which vocabulary a
new definition will be entered in is controlled by the two user
variables CONTEXT and CURRENT , each of which points to the most
recently defined word in a vocabulary. CONTEXT points to the
VOCABULARY that is first searched by a dictionary search, in either a
28


VLIST or a search for a word to compile into a colon—definition.
CURRENT points to the VOCABULARY into which new definitions are
placed. These two are usually, but not necessarily, the same.
A new vocabulary is created by, for example:
VOCABULARY TEST—VOC IMMEDIATE
This creates a new vocabulary with name TEST—VOC (by convention all
VOCABULARY words are IMMEDIATE ). This vocabulary can be made the
CONTEXT vocabulary by executing TEST—VOC . URM: CURRENT ·vocabulary
remains as that from which the new vocabulary was created (this would
normally be the FORTH vocabulary). Thus, after creating and then
executing TEST—VOC , a dictionary search will start in TEST—VOC , but
new definitions will still be entered into the old vocabulary. The
process of making a new definition automatically sets CONTEXT to be
equal to CURRENT so that after the sequence (assuming the initial
vocabulary to be FORTH ):
VOCABULARY TEST—VOC IMMEDIATE ( create TEST—VOC )
TEST—VOC ( set CONTEXT to TEST—VOC )
: NAME ; ( define NAME )
the word NAME will be in the FORTH vocabulary, which will now also be
the CONTEXT vocabulary.
The word DEFINITIONS sets the CURRENT vocabulary to be the same as
the CONTEXT vocabulary, so the sequence:
VOCABULARY TEST-VOC IMMEDIATE
TEST-VOC DEFINITIONS
: NAME ;
will place the definition of NAME in the TEST-VOC vocabulary which
will then be the CURRENT ( and CONTEXT ) vocabulary. Note that
TEST-VOC itself is created in the FORTH vocabulary. To forget TEST-VOC
it is necessary to type:
FORTH FORGET TEST—VOC
Each ·vocabulary· eventually links back. into ·the 'parent' vocabulary
(the CURRENT vocabulary at the time of its creation). It is normal to
ensure that each new VOCABULARY links directly into the FORTH
vocabulary, because though it is possible to chain vocabularies this
can result in a complicated and confusing search sequence and is
therefore not recommended.
One of the main uses of the VOCABULARY structure, in addition to
‘,~” separating the words of one application from those of another, is to
allow the use of the same word to represent several different actions
and still be able to find earlier definitions.
If we use the earlier example and type the following:
FORTH DEFINITIONS (make sure that FORTH is the current vocabulary)
: QUOTE ." THIS IS A LITERAL STRING " ;
VOCABULARY TEST-VOC IMMEDIATE
TEST-VOC DEFINITIONS
: QUOTE ." A DIFFERENT MESSAGE " ;
the warning of a duplicate entry will still be given but now the first
version is in the FORTH vocabulary and the second version is in the
TEST-VOC vocabulary. Typing:
on


FORTH QUOTE will give THIS IS A LITERAL STRING OK
whereas
TEST-VOC QUOTE will give A DIFFERENT MESSAGE OK
At this point VLIST will start in the vocabulary TEST-VOC .
The ability to use the same word for two different actions is used
in the EDITOR vocabulary where, for example, the word R is used to
replace a line of edited text. In the FORTH vocabulary the word R is
used to copy the top item of the return stack to the computation
stack.
5.6 The Compilation of a Colon-Definition
5.6.1 Normal Action
When a FORTH word is typed at the keyboard it is usually executed as
soon as the RETURN key is pressed. In the compilation mode, during the
creation of a colon-definition, the response is quite different. The
word is not executed but its execution address is added to the list of
addresses iJ1 the dictionary entry being constructed. This continues
until the terminating semi-colon is found, whereupon normal execution
is resumed. ,
5.6.2 IMMEDIATE Words
It is often necessary to define a word that will execute even in the
compilation mode. Examples include the conditional words IF, ELSE,
THEN which must execute in order to calculate the offsets for their
branches, and the vocabulary words e.g FORTH , EDITOR which allow the
changing of the CONTEXT vocabulary to include words from other
vocabularies in the current defintion.
The response to these words is identical in both execution and
compilation modes — they are always executed. They are classed as
IMMEDIATE words and are made so by including the word IMMEDIATE at the
end of their definitions, for example:
: DO—IT—NOW CR ." I AM EXECUTING" CR ; IMMEDIATE
If this is now used in another definition, for example: I
: TEST
CR ." I HAVE BEEN COMPILED" CR
DO-IT-NOW ; H
the message I AM EXECUTING will appear as soon as the RETURN key is
pressed after typing iJ1 DO—IT—NOW .. Executing TEST will produce the
message I HAVE BEEN COMPILED , but not the message of DO-IT-NOW ,
since this was executed and not compiled. Note that any type of word,
not just colon—definitions, may be made IMMEDIATE .
5.6.3 Making a Normal Word IMMEDIATE
It may be necessary, during the formation of a colon—definition, to
execute cxua or more words which would normally be compiled. This is
useful, for example, for the calculation of a numerical value, or to
change the numeric base, during the compilation of a colon-definition,
and is accomplished by the use of the words [ , which is itself
30


IMMEDIATE , and ]. The action of [ sh; to terminate compilation and
enter the execution mode, while ] has the opposite effect. They are
usually, but not necessarily, used as an pair (see Section 10.6.2).
Their use in a colon-definition is:
: NAME ... these words are compiled as usual ...
[ ... these words are executed ... ]
... compilation continues ... ;
[ and ] may also be used to include in a colon—definition a word which
has no name header in the dictionary, provided that its execution
address is known. For example, the word (ENTER) in ATOM FORTH is a
headerless dictionary entry with execution address #3AF5. Its function
is to interpret the contents of the tape buffer. It is given a header
in the full tape interface by:
HEX
: ENTER [ 3AF5 , ] ; V
5.6.4 Compiler Security
It is important to ensure that the sequence of words to be executed
between [ and ], or any IMMEDIATE word, does not change the number of
items on the computation stack. In the above example the number #3AF5
is added to the stack and then removed by <,>. Part of the compiler
security system is to check that the number of items on the stack is
unchanged across a colon—defintion. If the actions between [ and ]
have the net effect of adding or removing stack items an error message
will be given and the definition will be left in an incomplete form.
It is also important to realise that the words IF , ELSE , DO ,
BEGIN leave a number on the stack, to be checked and removed by the
corresponding THEN , LOOP , UNTIL etc. If these numbers are changed or
removed by any immediate action the compiler‘ security' systemn will
again give an error message and the definition will be incorrectly
terminated.
5.6.5 Forcing the Compilation of IMMEDIATE words
It may occasionally be necessary to compile a word that is marked as
IMMEDIATE. You may wish, for example, to delay a change of CONTEXT
vocabulary until a word is executed rather than the change taking
place during the definition cnf the word, as normal. Each IMMEDIATE
word can be forced to compile by preceeding it with the word
[COMPILE].
As an example, we can compile the IMMEDIATE definition DO-IT-NOW
of Section 5.6.2:
: DO-IT-LATER
[COMPILE] DO-IT-NOW ;
The message of DO-IT-NOW is no longer displayed during the definition
and will only appear when DO—IT—LATER is executed.
The word [COMPILE] is, as indicated km; the square brackets, an
IMMEDIATE word and not itself compiled. Its only action is to force
compilation of the following word. You may however, if necessary,
force its compilation by:
... [coMP1LE] [COMPILE] ...
3l


5.6.6 C mpiling Into Another Word
The word COMPILE (without square brackets) is not an IMMEDIATE word
and will therefore be compiled as normal into a colon—definition. It
is used in the form:
: NAME ... COMPILE WORD ... ; I
In this sequence COMPILE and WORD are both compiled into the
dictionary entry for NAME . When NAME is executed, COMPILE will act to
place the execution address of WORD into the next free dictionary
space ie. to compile it. WORD is not executed during the execution of
NAME.
5.6.7 An Example
A non—trivial example of the use of IMMEDIATE words, [COMPILE] , and
COMPILE occurs in the literal numeric handler of FORTH.
The word LITERAL is used by the keyboard interpreter to compile a
literal numeric value into a colon—definition.
: LITERAL
STATE @ IF COMPILE LIT , THEN
; IMMEDIATE
In the execution of LITERAL
STATE @ returns a true value if compiling and a false value if
the keyboard input is being executed.
IF tests this value and skips to THEN on a false result
(see Chapter 6), so LITERAL has no action in execution
mode.
COMPILE LIT if compiling a new colon-definition, compiles the
literal handler LIT into the new definition, and
, adds the numeric value from the top of the stack into
the definition.
The new definition will now include the sequence
... LIT (value) ...
and when it is later executed, LIT will act to put the following value
onto the stack, as required.
Note that LITERAL has to be an IMMEDIATE word so that it will
execute whenever a numeric value is to be included within a
definition. The word DLITERAL is used by the keyboard interpreter to
compile a double—precision numeric value into a definition, and uses
LITERAL twice. LITERAL must, however, be compiled into the definition
of DLITERAL by the use of [COMPILE] .
: DLITERAL
STATE @
IF SWAP
[COMPILE] LITERAL (low part)
[COMPILE] LITERAL (high part)
THEN
; IMMEDIATE
32


Like LITERAL , DLITERAL has no action in execution mode. A  
double—precision value is stored on the stack with its high-order
part above the low—order part so that on execution of DLITERAL in
compilation mode,
SWAP places the low-order part above the high—order part
LITERAL compiles the low-order part into the definition, and
LITERAL then compiles the high—order part.
The new definition will now contain the sequence:
... LIT (low part) LIT (high part) ...
and when it is later executed the double—precision number will be
pushed onto the stack with its two parts in the correct order.
33


34


6 c aw 1 a L
6.1 introduction
FORTH is a highly—structured language, in which all transfers of
control are accomplished. without the use of (KIN) statements. This
requires the writing of applications in a modular style, where each
module has only one entry and one exit point. Although very different
from writing a BASIC program, it is not too difficult since it is
almost impossible to write a FORTH application in any other way. Once
you have grasped the underlying ideas, writing structured programs
becomes natural and soon you begin to wonder how you ever managed to
do anything in an unstructured language.
6.2 Conditional Branches
The simplest conditional branch in FORTH uses
... IF ... THEN ...
These words may look familiar, but in FORTH their actions are somewhat
unusual. In BASIC the action of a statement such as
IF X=2 THEN GOTO 2137
is interpreted as
IF this test is true THEN do this statement, otherwise go on to
the next line (and just what does line number 2137 do, anyway?).
Just like operators, the IF in FORTH is post—fix, so the value to be
tested comes before the II' . The IT' ... THEN structure in FORTH is
interpreted as:
IF the result of the test was true, do this sequence, otherwise
skip it
THEN continue with the following sequence, in either case.
Some FORTH systems attempt to make this clearer by using IFTRUE to
replace IF , and ENDIF to replace THEN .
One restriction in FORTH is that the branch words, and the loop
words of the following sections, can.only be used inside a
colon-definition and may not be directly executed from the keyboard.
The way in which they are used is:
: EXAMPLE
?TEST IF DO-THIS THEN CONTINUE ;
A FORTH definition with the same function as the BASIC statement in
the earlier example could appear as:
: =2? 2 = IF ." VALUE WAS TWO " THEN ;
Where has X gone? The sequence 2 = will test the top number on the
stack and leave a true (non—zero) result if it were equal to 2 and a
false (zero) result otherwise. In FORTH it is often not necessary to
35


use an explicitly-named variable; as long as the appropriate value is
placed on the stack at the right time it doesn’t matter how it got
there.
In the above example the value to be tested can be entered
directly from the keyboard, for example:
2 =2? VALUE WAS TWO OK »
3 =2? Q5
Note the use of the word WAS. The general rule in FORTH is that words
remove from the stack the numbers they use. The word <=> will remove
the 2, placed on the stack by =2? , and the number being tested,
leaving only a true/false flag. IF will then remove the flag, so =2?
leaves the stack unchanged. If the value being tested is needed again
DUP must be used before =2? .
In many cases you may wish to execute one sequence if the test is
true and a different sequence if the test is false. The sequence
IF ... ELSE ... THEN
will, if the result of the test was true, execute the words between IF
and ELSE and skip to THEN . If the result of the test was false it
will skip to ELSE, and execute the words between ELSE and THEN . The
words (if any) after THEN will_be executed in either case. To
illustrate this we can
FORGET =2? Q5
and redefine it as follows:
: =2?
2 = IF ." VALUE WAS TWO "
ELSE ." VALUE WAS NOT TWO "
THEN ;
(The layout is irrelevant — type it any way you like — as long as you
do not press the RETURN key in the middle of a word, or of ." ...“ ,
all will be well. The above layout looks good and makes the structure
clearer.)
Trying the new version gives, for example,
2 =2? VALUE WAS TWO Og
3 =2? VALUE WAS NOT TWO Q5
Incidentally, the definition can be re—written to use less memory.
FORGET the old definition and replace it by:
: =2?
." VALUE WAS"
2 - IF ." NOT" THEN
•II TWOII ;
This has exactly the same effect as the earlier, longer, version
(remember that a true result may be represented by any non—zero
value).
Often, the number being tested for truth by IF may be needed for a
calculation in the IF ... THEN sequence, but not needed otherwise. One
way of doing this is:
... DUP IF ... ELSE DROP THEN ...
36


DUP duplicates the number to be tested, and the copy is discarded by
DROP if it is false. ·
A neater solution is to use —DUP , which will only duplicate a
number if it is non—zero. Thus the sequence is equivalent to
... —DUP IF ... THEN ...
If the number is zero it is not duplicated and there is obviously no
need then to DROP it, since the only copy is removed by IF .
The IF ... THEN and IF ... ELSE ... THEN forms may be nested to
any required depth, provided that the nested structure lies completely
within the outer structure. The following are examples of valid
nestings. The nested structure is underlined for clarity.
... IF ... IF ... ELSE ... THEN ... THEN ...
... IF ... IF ... THEN ... ELSE ... THEN ...
Too many levels of nesting, however, make the definition hard to
understand and should be avoided. It is much clearer if a long
definition with many levels of nesting is split up into a number of
short definitions. The nested structure of the second case given above
is much clearer if it is written as:
: NESTED-IF
IF ... THEN ;
: OUTER-IF
... IF ... NESTED-IF ...
ELSE ... ‘
THEN ... ;
A general rule in FORTH is
long definitions = bad definitions
— keep them short!
6.3 Definite Loops 6
A loop whose number of repetitions is known before entry will use the
DO ... LOOP structure. DO takes two values from the stack, the start
index and the loop limit. If we take as an example the definition:
: TENCOUNT l0 0 DO I . LOOP ;
then executing TENCOUNT will give:
TENCOUNT 0 l 2 3 4 5 6 7 8 9 OK
The loop index is post-incremented, i.e. the increment occurs in LOOP,
after the body of the loop is executed. The loop will terminate when
the (incremented) loop index equals or exceeds the loop limit.
This has two important consequences:
a) Regardless of the value of the loop limit and the starting index,
the body of the loop will be executed at least once.
b) The last execution of the loop body will be with an index which is
one less than the loop limit. Thus, in the example, the loop was
37


executed ten times but the last execution was with a loop index of
9.
The word I , which should only be used within a loop, places the
current loop index on the stack. Note that in this example, I is
immediately followed by <.> which types (and removes from the stack)
the value left by I .
Programming errors that cause a net change in the number of items
on the stack inside the body of the loop can lead to a stack overflow
resulting in a system crash. One of the easiest ways of crashing the
system is to execute the following definition (not recommended):
: CRASH 100 0 DO I LOOP ;
Small stack overflows will result in an error message, either message
7 (stack full) or message l (stack empty). Even large stack overflows,
such as that in tina above example, will not cause the loss of the
system. Pressing the BREAK key and restarting at the warm entry point,
#2804, as described in Chapter 3, should cause a successful recovery.
As long as there are two values on the stack for DO ... LOOP , it
does not matter how they got there. In the examples so far the values
have been placed on the stack within the definition. They may,
however, be entered directly from the keyboard:
: DELAYS 0 DO LOOP ;
In this example, only the starting index is put on the stack
within the definition. The loop limit may be entered from the keyboard
so that
10000 DELAYS (pause) OK
30000 DELAYS (longer pause) OK
can be used to give a variable length delay.
8000 DELAYS
will give approximately a one—second delay.
Of course both values may be entered from the keyboard. The
definition
: COUNTER DO I . LOOP ;
will allow the following:
8 0 COUNTER 0 l 2 3 4 5 6 7 OK
52 47 COUNTER 47 48 49 50 5l OK
and so on.
It is awkward to have to remember to type the values in reverse
order and to have to add one to the last required value. The routine
can be made more 'user friendly' by defining
: COUNTS l+ SWAP DO I . LOOP ;
38 is


This can then be used in a more sensible way:
l 7 COUNTS 1 2 3 4 5 6 7 OK
10 14 COUNTS 10 ll 12 13 14 OK
or even to count in hexadecimal:
10 16 HEX COUNTS A B C D E F 10 OK
Remember to change the base back to DECIMAL .
For increment values other than 1 the DO ... +LOOP structure is
used. +LOOP expects to find its incremental value on the stack. This
value can be given in the definition as in the following example.
: 3-COUNT 1+ SWAP DO I . 3 +LOOP ;
0 15 3-COUNT 0 3 6 9 12 15 OK
The increment could, if you feel confident, be calculated within the
loop to give a variable increment, for example:
: SEQUENCE 1+ SWAP DO I DUP . 2* +LOOP ;
1 27 SEQUENCE 1 3 9 27 OK
By the use of a negative increment, the loop can count backwards:
: BACKWARDS DO I . -1 +LOOP ;
0 6 BACKWARDS 6 5 4 3 2 1 OK
Note that the loop still terminates when the incremented index equals
or passes the loop limit.
Loops may be nested, provided that the inner loop is completely
enclosed by the outer one. This is illustrated by the following
example, which is laid out in such a way that the nested structure is
made clear.
: 100-coUNT
10 0 DO I 10 *
10 0 DO DUP I + . LOOP
DROP CR
LOOP ;
The word I is used twice, once in each loop. The first I will leave
the outer loop index and the second I leaves that of the inner loop.
The sequence I li) * leaves cum the stack the tens value for the
count. In the inner loop this is first duplicated, so that it remains
available for the next time round the loop. The inner loop index (the
units value) is then added to it and the resulting value is printed.
On leaving the inner loop the old tens value is dropped, and a
carriage return ensures that the final display is 'tidy‘. Executing
100-COUNT will then display 100 integers, from 0 to 99 inclusive.
There is no reason why loops and conditional branches should not
be nested, again provided that the inner structure is completely
enclosed by the outer. Definitions in which the structures overlap,
such as: _
DO ... IF LOOP ... THEN
39


are not allowed.
In the following example it is assumed that the word LIST has been
previously defined to leave on the stack the starting address under
the number of single precision (l6-bit) items in a table of values
(the method of doing this is discussed in Section l0.4.3). The word
LOOK-UP will search the table for a particular value, initially on the
stack, and will leave either
a) the item offset within the table under a true flag, if the item is
found, or
b) only a false flag if the item is not found in the table.
It is used in the form:
n LIST LOOK—UP
where n is the value to be found.
The definition may be tested without the need to define LIST since
all it needs is three values on the stack. It can be used to search
any region of memory for a particular (l6-bit) word by giving it the
value to find, aa starting address and the length, in words (2—byte
units), of the region to be searched, for example
2345 l0240 2048 LOOK—UP
will search the first 4K of the dictionary for the value 2345 (and
fail to find it).
ll2l8 l0240 2048 LOOK—UP
should find the value with an offset of 1057.
: LOOK—UP ( val\addr\count ... offset\l ) ( found )
( val\addr\count ... 0 ) ( not found )
0 DO ( loop limit is count )
2DUP ( value under base addr )
I 2* + ( add byte offset to addr )
@ = ( table item = val? )
IF ( equal ) y
I 0 LEAVE ( item offset under 0, and exit loop )
THEN
LOOP ( top of stack is 0 if found, or )
( addr [assumed non—zero] if not )
IF ( not found )
DROP 0 ( val )
ELSE I ( found )
ROT ROT 2DROP l ( val and addr )
THEN ;
The word LEAVE, when executed, causes an exit from the loop. The exit
is not immediate but will occur when LOOP (or +LOOP) is next
encountered. The action of LEAVE is to change the loop limit to be
equal to the current value of the loop index, which is not changed.
The words, if any, between LEAVE and LOOP will be executed once before
exiting the loop.
An interesting variation on LOOK-UP is the following alternative
definition. It has exactly the same effect, but searches the region of
memory from high addresses to low. There is, however, one word fewer
to be executed in the loop for an unsuccessful match, so it is
slightly faster.
40


: LOOK-UP ( val\addr\count ... offset\l ) ( found )
( val\addr\count ... 0 ) ( not found )
OVER >R ( save addr for later )
2* OVER + ( calculate last address in table )
0 ROT ROT ( put 0 under addr and last )
DO
OVER I @ = ( table item = val? )
IF ( found )
DROP ( DROP the 0 )
I ( table address [assumed non-zero] )
I MINUS ( and its complement for +LOOP )
ELSE
- 2 ( increment for +LOOP )
THEN
+LOOP
SWAP DROP ( val )
R> ( recover addr )
OVER ( copy of either 0 or table address )
IF ( not the 0 )
— 2 / l ( calculate table offset under l )
ELSE
DROP ( DROP addr but leave the 0 )
THEN ;
An unsuccessful search of 4096 bytes of memory, i.e. 2048 comparisons,
takes about 3 seconds using the first method and about 2.4 seconds by
the second method.
The method of leaving the loop on a successful match does not use
LEAVE. For an unsuccessful match the loop index is decremented by -2,
but on a successful match the complement of the loop index is left for
+LOOP. This will guarantee that the increment will cause the loop
limit to be exceeded, thus terminating the loop.
It is a useful exercise to try to modify either (or both) of these
routines to search for a particular byte (8-bit) in memory. Not too
many alterations are needed. A further useful modification would be to
change the input stack requirements from val\addr\count to
val\addr\endaddr.
The word J can be used to leave, in an inner loop, the loop index
of the outer loop. Its action is demonstrated by the following
definition:
: JTEST
CR ." J ( OUTER ) I ( INNER )"
CR 0 DO
3 0 DO CR J .
l0 SPACES I .
LOOP CR
LOOP ;
Since the loop index and limit are kept on the return stack, which is
also used to keep track of the level of nesting of colon definitions,
the word J (and I) will only operate correctly if they are used at the
same level of definition. The following sequence, for example, will
not have the required effect.
: INNER 3 0 DO J . I . LOOP ;
: OUTER CR 3 0 DO INNER CR LOOP ;
4l


Executing OUTER will give the correct operation for I, but the value
of J will not be what was intended.
If DO ... LOOPs are nested to a depth of 3 then, from the
innermost loop,
I leaves the index of the inner loop
J leaves the index of the middle loop
K leaves.the index of the outer loop
The word K is not provided in the nucleus dictionary. Its definition
is:
: K RP@ 9 + @ ;
The idea can be extended to further levels by defining L , M , etc.
For each extra level the definition is similar to that of K , except
that the number to be added is increased by 4, for example:
: L RP@ 13 + @ ;
6.4 Indefinite Loops
There are three forms of indefinite loop:
BEGIN ... AGAIN
BEGIN ... UNTIL
BEGIN ... WHILE ... REPEAT
In each case BEGIN umrks the start of the sequence of words to be
repeated.
The word AGAIN causes a branch back to the corresponding BEGIN so
that the intervening words are repeated endlessly. This form of loop
was used in the definition of STARS in Chapter 3 to create an
application whose execution can only be terminated by pressing the
BREAK key. A BEGIN ... AGAIN loop is used only if it is to initiate a
repetitive sequence of actions which are to continue until the machine
is switched off. It is useful for turnkey applications where the user
is not expected to know, or wish to alter, the method of operation.
In the FORTH system it is, for example, used for the keyboard
interpreter which interprets all input to the computer. While FORTH is
in action all operations are at a more or less deep level of nesting
from within the keyboard interpreter, to which control must ultimately
return (when OK is displayed).
The remaining two forms may be terminated by the result of a test
made within the loop.
In the case of BEGIN ... UNTIL , the word UNTIL tests the top item
on the stack. If this value is false (zero) a branch will occur to the
corresponding BEGIN . If the value is true (non-zero) the loop will be
left and the words following UNTIL will be executed. The definition:
: PAUSE
CR BEGIN ?ESC UNTIL
." ESC KEY PRESSED" CR ; ,
will loop until the ESC key is pressed since ?ESC leaves a false value
on the stack unless the ESC key is pmessed, when it leaves a true
value.
42


If we define
0 CONSTANT HELL-FREEZES—OVER
the definition:
: WAIT
BEGIN HELL—FREEZES—OVER UNTIL ;
will, on execution, wait until the condition is satisfied!
On a slightly more useful level, the definitions:
: GCD ( nl\n2 ... gcd )
BEGIN
SWAP OVER MOD -DUP O=
UNTIL ;
: G—C—D ( nl\n2 ... )
GCD CR ." THE G—C—D IS " . ;
will calculate and display the greatest common divisor of the numbers
nl and n2. For example,
l5 25 G—C—D
will respond:
THE G—C—D IS 5 OK
Note how in these definitions, the calculation of the result and the
display routines are placed in separate words. This means that GCD can
be used as part of a longer calculation, where the value is not
required to be printed, without needing to re—write its definition.
When the value is to be displayed, however, the word G—C—D can be
used, as in the above example.
The BEGIN ... WHILE ... REPEAT structure will terminate as the
result of a test which should be made immediately before WHILE . This
word expects a true/false flag cxx the stack but in this case will
terminate the loop when the value is false. If the value is true,
execution will continue with the following words up to REPEAT which,
like AGAIN , causes an unconditional branch back to the corresponding
BEGIN . For a false value the words between WHILE and REPEAT are
skipped, the loop terminates, and then time words after REPEAT are
executed. An example of the use of BEGIN ... WHILE ... REPEAT is the
INPUT routine in Section 7.l.3 of the next chapter.
Indefinite loops may, of course, be nested with any of the other
structures to any reasonable depth, provided that the nested routine
is totally enclosed within the outer structure.
43


44


This chapter deals with the methods of controlling input and output in
FORTH. So far we have met two output operations, <.> and <."> , which
display a number (in the current numeric base) and a literal character
string respectively. All input, whether text or numeric, has used the
keyboard interpreter. The following sections give specific methods of
input and output.
7.1 Input
7.1.1 Character Input
A single character may be input by the use of KEY. This word waits for
a key to be pressed and leaves the corresponding ASCII code on the
stack. The definition:
: SHOWASCII KEY . ;
will accept a character from the keyboard and display its ASCII code
in the current numeric base.
The sequence:
KEY DROP
is a useful way of causing a wait until any key is pressed.
7.1.2 Text Input
The word QUERY will accept any sequence of characters typed on the
keyboard, up to a limit of 80 characters, or until RETURN is pressed.
The characters are stored in the terminal input buffer, followed by
one or more zeroes.
Text may be transferred from the input buffer to the word buffer,
immediately above the top of the dictionary, by use of WORD . This
expects to find a delimiter character on the stack and this is usually
. a space, ASCII code 32. Leading delimiter characters are ignored, and
text up to the next delimiter is transferred to the region of memory
starting at HERE . The first byte will contain the length of the text
string, with two or more space characters added to the end.
The text string at HERE may then be moved to another region of
memory or further manipulated, depending on what is required. As an
example, the definition:
: .STRING
QUERY 32 WORD HERE COUNT —TRAILING TYPE ;
will accept text from the keyboard and type it on the display.
It is important to realise that the keyboard interpreter itself
uses WORD so that all keyboard input is transferred, word by word, to
the word buffer, overwriting the previous contents. The implication of
this is that all uses should be from within a definition so that its
execution does not involve the keyboard interpreter. Simply executing
QUERY 32 WORD HERE COUNT —TRAILING TYPE
45


will not give the intended result.
The delimiter character need not always be a space. The word <.">,
for example, uses ASCII code 34 (#22) i.e. " as its delimiter. Since
<."> is a FORTH word it must be separated by a space from the text on
which it operates. The closing " is not a FORTH word but only a
delimiter and so does not need a space separating it from the text. If
a space is left, however, it will be included in the text string.
The word <."> is one of the most common ways of entering literal
text into a definition, for display when the definition is executed.
Outside a definition <."> will cause the immediate typing of the input
text.
7.1.3 Numeric Input
Most applications do not require special numeric input routines.
Since, shi general, words expect txn find their numeric data on the
stack, this can be placed there, by use of the keyboard interpreter,
before the word is executed. This is illustrated by the way that G—C—D
was used at the end of the last chapter.
Occasionally it may be necessary to wait for numeric input during
the execution of a word, and for this the following definition may be
used:
: NUMIN
CR ." ? " ( give prompt )
QUERY ( accept text to buffer )
32 WORD ( transfer characters to HERE )
( with space - ASCII 32 — as delimiter )
HERE NUMBER ( convert to double number )
DPL @ l+ O= ( was decimal point included? )
IF DROP THEN ( if not, make single number )
This routine leaves either a double- or single-precision number on the
stack, depending on whether a decimal point did or did not appear in
the input number. The number of digits to the right of the decimal
point is stored in DPL . If no decimal point was present the value of
DPL defaults to —l.
The disadvantage of this routine is that a standard error message
is given if a non-valid character is present in the input. This causes
execution to stop and the stack is cleared — a rather drastic action
for a mis—typed input!
Valid characters arezf
a) an optional minus sign as the first character
b) an optional decimal point at any position
c) all characters that may be interpreted as digits in the current
numeric base. In HEX , for example, valid characters are O to 9
and A to F inclusive.
It would be possible to define an alternative error—handling routine
and store its execution address in the parameter field of (ABORT) (at
address #347C). Setting the value of the user variable WARNING to —l
before the use of NUMIN would cause the error handler to use the new
routine. Don't forget to set WARNING back to its original value of O
afterwards, to restore the normal error messages. ,
An alternative solution is to use the word (NUMBER) which, in Atom
FORTH, is a headerless routine with execution address #33B7.
There are two main differences between NUMBER and (NUMBER).
46


Firstly, (NUMBER) does not generate an error message on detecting a
non-valid character, but simply leaves the address of the first
unconvertable character in the input text. Secondly, it does not test q
the first character of the input for the presence of a minus sign. In
addition, (NUMBER) requires a dummy double number on the stack, into
which the input value is built.
The following definition generates its own error message and will
not continue until a valid number is entered. Note that the base is
HEX so 20 WORD is the same as the 32 WORD in NUMIN , which was defined
in DECIMAL base.
HEX
: INPUT
BEGIN
CR ." ? " QUERY 20 WORD ( input text to HERE )
0 0 ( for building the number )
HERE ( text start address )
DUP l+ C@ 2D = ( test for minus sign )
DUP R> + ( save result and skip sign )
( if present )
-1 ( default DPL value )
BEGIN
DPL 1
[ 33B7 . ] ( (NUMBER) )
DUP C@ ( unconvertable character )
DUP BL = 0= ( leaves l if not a space )
SWAP 2E = ( and a l, if decimal point )
WHILE ( decimal point )
0= ( leave 0 to reset DPL )
REPEAT
WHILE ( invalid character )
R> 2DROP 2DROP ( clear the stacks )
." 1NvAL1D"
REPEAT ( and try again )
DROP ( address of space )
R> IF DMINUS THEN ( apply sign to number )
DPL @ l+ 0= ( no decimal point? [DPL= —l] )
IF DROP THEN ( if not, make single number )
The numeric conversion takes place in the inner loop, where DPL is
initially set to -1. During the conversion by (NUMBER) , DPL is
incremented for each converted digit, provided it is not equal to -l.
There are three cases which cause an exit from (NUMBER);
i) finding a decimal point (l\l)
ii) finding a space (0\0)
iii) finding any other non-numeric character (l\0).
The bracketed values in each case represent the top two items on the
stack after the two tests in the loop (note that the sequence <= 0=>
in the inner loop could be replaced by <-> since any non—zero value is
interpreted as true). (
In case (i), WHILE finds a l on the stack so the loop is repeated,
after 0= has changed the second l to 0, to reset DPL . Since DPL is no
longer -l it will now be incremented by (NUMBER) to give the number of
converted digits since the last decimal point.
In case (ii), the first 0 causes the termination of the loop,
leaving the second 0 to terminate the outer loop also.
47


In case (iii), the inner loop is again terminated, but the
remaining l causes the outer loop to be repeated, generating an error
message and returning for a new value to be typed.
The input prompt and the error message may, of course, be changed
to whatever you prefer.
The result of a successful conversion is identical to that of
NUMIN or to entering numbers via the keyboard interpreter.
7.1.4 Manipulating Blocks of Memory
The examples of the last two sections make frequent use of WORD, which
transfers a block of data from the input buffer to the region of
memory just above the dictionary. At this point it is worth examining
the words which allow such transfers to be made.
The usual way of transferring a block of bytes from one region of
memory to another is by the use of CMOVE. This word uses three values
from the stack; the starting address of the source block, the starting
address of the destination block and the number of bytes to be moved.
The stack action is therefore:
CMOVE ( from\to\count ... )
The byte with the lowest address is moved first and the transfer
proceeds in the order of increasing address. There is never a problem
if the destination address is less than the source address, but a
difficulty arises iii the destination address is higher than that of
the source and the two regions overlap. Consider, for example, that
the five bytes starting at FROM contain the characters F O R T H and
it is required to move them one byte forwards in the memory. It might
be thought that the following sequence would do the job.
FROM DUP l+ 5 CMOVE
This illustration shows what would happen:
FROM > F F F F F F
O > F F F F F
R R > F F F F
T T T > F F F
H H H H > F F
F
Moves O lst 2nd 3rd 4th 5th
In each column the arrow indicates the character that will be moved to
produce the next column. The final result is that the whole region of
memory is filled with the first character — probably not the required
effect! In order to avoid this problem the word 

4 will produce the following result:
FROM F F F F > F F
O O O > O O F
R R > R R O O
T > T T R R R
> H H T T T T
H H H H H
Moves 0 lst 2nd 3rd 4th 5th
If you don't want to have to worry about which of the two versions to
use, you can define an 'intelligent' version which will select the
correct one for you.
: CMOVE ( from\to\count ... )
>R 2DUP R> ROT ROT —
IF R OVER C!
DUP 1+ R> l — CMOVE r
The definitions of ERASE and BLANKS are very simple:
: ERASE 0 FILL ;
: BLANKS BL FILL ;
BL is a constant whose value is 32 i.e. the ASCII code for a space (or_
blank).
7.2 Number Bases
All numeric input and output is converted according to the current
value of the user variable BASE . Any value of BASE may be used,
subject to the restriction that it should lie in the range 2 to 255. A
practical upper limit is 36, to avoid the use of non-numeric or
alphabetic characters.
49


The internal numeric handling of FORTH is always in binary,
irrespective of the value of BASE, so there are no time overheads to
working in any base you choose.
The two numeric bases DECIMAL and HEX are provided with the
system. Any other base can be defined as follows:
: BINARY 2 BASE 1 ;
: OCTAL 8 BASE 1 ;
: BASE—36 36 BASE 1 ;
etc.
On first entry to the system, or after executing COLD or WARM, the
base will always be DECIMAL .
Many decimal—to-hex routines have been published in BASIC, with a
greater or lesser degree of complexity. In FORTH such a routine is
simply:
DECIMAL ( make sure you start in decimal )
: D->H
HEX . DECIMAL ;
Here is an example of its use:
DECIMAL 31 D->H lF OK
The routine can be modified to translate between any two bases.
Finally, you may like to try executing the following, having
previously defined BASE—36 as above. `
505030. 38210. 676 1375732.
BASE—36 CR D. . D. D. CR DECIMAL
7.3 Output I
7.3.1 Character Output
To output a single character, the word EMIT can be used. This will
display the character whose ASCII code is on the stack.
Examples:
65 EMIT A OK
49 EMIT 1 OK
It can also be used to execute control codes (see "Atomic Theory and
Practice", page 131) from within a definition, e.g.
: BELL 7 EMIT ;
7.3.2 Text Output
Text strings in FORTH are stored with a preceding length count byte,
as mentioned in Section 7.1.2. Access to a string is usually via the
address of this byte.
The display of a string is performed by TYPE which expects on the
stack the address of the first character, under a length count. The
conversion to this form from the address of the count byte is done by
50


the word COUNT. Thus
HERE COUNT TYPE
will display the string starting with its count byte at the address
given by HERE. The character count may include a number of blank
spaces at the end. These can be removed by the use of -TRAILING, which
deletes all trailing spaces from the string, for example:
HERE COUNT —TRAILING TYPE
Remember that the use of strings stored at HERE should only be from
within a colon-definition to avoid their being overwritten by the
keyboard interpreter.
The FORTH system does not provide string handling facilities but
they are fairly easy to include if required. For example the following
two definitions provide left and right string extraction. They assume
that the address of the count byte of the string is initially on the
stack under the number of characters to be extracted.
: LEFT$ ( addr\nl ... addr2\n2 )
SWAP COUNT ROT MIN ;
: RIGHTS
SWAP COUNT ROT 2DUP >
IF DUP >R — + R> ELSE DROP THEN ;
In both cases the stack is left in a state ready to TYPE the selected
character string. If the number of characters exceeds the length of
the string, the entire string will be displayed by TYPE .
As an example, type in the following:
: SIN ( STRING input to HERE with ' as delimiter )
CR ." $'" QUERY 39 WORD ;
: STRINGS
$IN HERE DUP
CR COUNT TYPE CR
DUP lO LEFT$ TYPE
6 RIGHTS TYPE CR ;
Then execute STRINGS as follows:
STRINGS 
SLTHIS IS A LONG STRING' 
THIS IS A LONG STRING
THIS IS A STRING
Q5
Further discussion of strings is deferred to Chapter l0.
7.2.3 Numeric Output
The numeric output operators provided in FORTH are:
. (n ...) Display the signed number n followed by one space
.R (nl\n2 ...) Display the signed number nl at the right of a field
n2 characters wide. No following space is printed.
D. (nd ...) Display the signed double number nd in the format
5l


of <.>.
D.R (nd\n ...) Display the signed double number nd to the right of
a field n characters wide. No following space is
printed.
U. (un ...) Display the unsigned number un in the format of <.>.
The words .R and D.R are useful for tabulating information. Their use
is illustrated in the following routine which will dump 64 bytes of
memory, given its starting address. When the listing stops, pressing
the space bar will display a further block. Pressing any other key
will terminate the routine. The display is in HEX , regardless of the
initial value of BASE , which is restored on exit from the routine.
z DUMP ( addr ... )
BASE @ SWAP ( SAVE CURRENT BASE >
HEX
BEGIN
DUP 64 + SWAP ( SET ADDRESS OF NEXT BLOCK )
8 0 DO CR
DUP I 8 * +
DUP 0 4 D.R SPACE ( SHOW ADDRESS )
8 0 DO
DUP I + C@ ( GET A BYTE )
3 .R ( DISPLAY IT )
LOOP
DROP
LOOP
DROP CR
KEY BL - ( WAIT FOR KEY PRESS AND TEST )
UNTIL ( IF SPACE, REPEAT LOOP )
DROP
BASE 1 ( RESTORE BASE )
CR
To illustrate the action of U. try the following:
30000 . 30000 OK
30000 U. 30000 OK
This shows that U. and . give the same result for positive signed
numbers.
-30000 . -30000 OK
-30000 UL 35536 OK
The number -30000 is interpreted by . as a signed integer, in the
range -32768 to +32767, but by U. as an unsigned integer in the range
0 to 65535. Whether a number is to be treated as a signed or an
unsigned number is a matter of context.
The word U. is simply defined as < 0 D. > in other words it
prints the value as a double-precision number with a high order part
of zero.
The sequence:
0 4 D.R
in DUMP also uses this idea to display a 2 byte (4 hexadecimal digits)
address as an unsigned number.
52


7.2.4 Numeric Output Formatting
The numeric output operations of the previous section enable the use
of two formats: the printing of a number at the current cursor
position (using <.>) and the placing of a number at the right of a
field of specified width (using .R and D.R ).
Other formats may be produced by use of the special numeric output
formatting words:
<# ' Set up for numeric conversion
#> Terminate numeric conversion
# Convert one digit
#5 Convert the remaining digits
SIGN Insert a minus sign in the converted string.
HOLD Insert the specified character in the converted string
The words # , #5 , SIGN , HOLD may only be used between <# and #>, and
all act on a double precision number on top of the stack. On
completion of the conversion, the word #> leaves the number ready to
be displayed by TYPE .
The first example will display a double-precision number as pounds
and pence. The original number is the value in pence and the routine
will handle amounts up to 2l474836.47, which should be enough for most
purposes! One problem is that the Acorn ATOM does not have a pound
sign in its character set. In this example the character # is used
since the ASCII code for this symbol is often shown as a pound sign on
a printer.
: .P0uNns ( nd ... >
DUP ROT ROT ( KEEP HIGH PART, INCLUDING SIGN )
DABS ( MAKE POSITIVE )
<# ( START CONVERSION )
# # ( c0NvERT 2 DIGITS - Pnucn )
46 HOLD ( INSERT DECIMAL POINT )
#5 ( CONVERT REMAINING DIGITS )
SIGN ( INSERT SIGN IF NEEDED )
35 HOLD ( INSERT # )
#> ( END CONVERSION )
TYPE SPACE ( DISPLAY CONVERTED NUMBER )
-12345. .POUNDS #-123.45 OK
The following example will display a double number with the decimal
point in the position indicated by the value stored in DPL. If DPL is
zero or negative the decimal point will be at the extreme right hand
side of the number.
53


: .REAL
DUP ROT ROT DABS
DPL @ 0 MAX ( MAKE SURE NOT LESS THAN 0 )
<#
-DUP IF ( IF NON-ZERO )
0 DO # LOOP ( CONVERT DPL DIGITS )
THEN
46 HOLD
#S SIGN #>
TYPE SPACE
1234.5 .REAL 1234.5 OK
.12345 .REAL 0.12345 OK
-12.345 .REAL -12.345 OK
123 0 5 DPL ! .REAL 0.00123 OK
Note that all rummric conversion starts with the least significant
digit and proceeds towards the more significant digits. The conversion
process produces the string of output characters in a scratchpad area
whose start address is placed on the stack by the word PAD. Numeric
strings are built up starting at PAD and working towards the lower
addresses. The characters in the string are therefore in the correct
order (most—significant digit first) for display by the numeric output
words.
54


8Ta ur dEd`t
8.1 Introduction
When the nucleus dictionary of FORTH is loaded it will allow the use
of the word LOAD to load an application from tape. The source code on
tape is divided into 'screens'. Each screen is 512 (#200) bytes long
and is divided into eight 'lines' of 64 characters. On a printer it
will be displayed in this format, but on a VDU screen it will appear
as sixteen lines of 32 characters.
when an application is loaded from tape each screen is first
loaded into E1 buffer area, starting at address #95C0, and then
interpreted in the same way as if it had been typed at the keyboard.
Almost anything that can be typed at the keyboard can be used in a
tape screen, but there are two major exceptions.
l. A loading screen will not prompt for input from the keyboard. This
is sensible since it is impossible to predict how long it will take
to complete the keyboard action and loss of synchronisation with
the tape will almost certainly result. A running tape, like time
and tide, waits for no man.
2. A screen cannot be used as a 'load screen' to load a number of
other screens. Again, this is run; a disadvantage in a pmimarily
tape—based, system since the order of loading` screens from. tape
cannot be changed. The loadimg of a series of consecutively
numbered screens is done by the use of the word ——> (see Section
8.2.2).
If either of these two is attempted the system will 'go away', and a
BREAK followed by a COLD or WARM start will be necessary.
In order to view the contents of a screen, or to write or modify
screens, it is first necessary to load the full tape interface and
screen editor. These are provided as an application on the ATOM FORTH
tape, in screens 6 to 14 inclusive. Before continuing with this
chapter, load these screens. Put the tape in the cassette recorder and
type:
6 LOAD
when the machine will respond with:
>6 PLAY TAPE
Start the tape and press the space bar. After a few seconds you will
notice some slight interference on the screen, indicating that screen
6 is being loaded. After a short pause the message
2.1
will appear, indicating that screen 6 has been loaded successfully and
a search is now being made for screen 7. There is no need to press any
keys; the loading will take place automatically until all the required
screens have been loaded when the response
Q5
55


will be typed.
During the loading of screen ll the messages
R MSG #4 I MSG #4
will appear, but don't worry. All that this means is that the words R
and I are redefined in this screen (as part of the EDITOR vocabulary).
8.1 The Tape Interface
When the OK prompt has appeared, at the end of the loading of screen
14, the full tape interface and screen editor are loaded and ready for
use. when this happens type:
14 LIST
This will list on the display the contents of the last screen to be
loaded (screen 14) which contains definitions of the words TILL and C.
Now type:
6 LIST
This time the response will be:
>6 PLAY TAPE
Rewind the tape, start it playing and press the space bar. After a
short pause the contents of screen 6 will be displayed, containing the
definitions of .LINE , LIST itself, and ENTER . As soon as the listing
has started the tape recorder can be switched off. Typing
6 LIST
again will give an immediate listing of screen 6. The word LIST
expects to find a screen number on the stack. If this is the number of
the screen present in the tape buffer area it will be listed
immediately. Otherwise the loading prompt will be given and the screen
loaded from tape before being listed. Now type:
14 LIST
and start the tape in the usual way. As the tape is searched for
screen 14, the number of each screen will be displayed as it is
encountered. When screen 14 is found it will be loaded into the tape
buffer and listed as before. The search may be stopped at any time
during the period before the specified screen is found by pressing the
CTRL key. This will cause a re—entry of FORTH via a WARM start. The
only point to remember is that the current screen number will have
been set to the value used with LIST (eg. 14 in the above case) so
that typing 14 LIST again will list the current contents of the tape
buffer as screen 14. This is not really a disadvantage since the main
reason for terminating a search is if it is realised that the wrong
screen number is being used.
This method of terminating a tape search can also be used with
LOAD . Zhu both cases the process cannot be terminated by use of the
CTRL key once the process of loading the screen into the tape buffer
has started.
56


If you tried stopping the search by the above method, change the
stored screen number by typing:
13 SCR 1
and then type l4 IJST again. When you have a listing of screen 14
type:
: TASK ; (
(to mark this point in the FORTH vocabulary) and then:
ENTER
This is the word which causes the contents of the tape buffer to be
interpreted, as though it had been typed at the keyboard. In this
case it will cause an immediate error and the message:
#LEAD ? MSG 0
will appear. This has occurred because the system is in the FORTH
vocabulary, but the word #LEAD is in the EDITOR vocabulary and
therefore has not been found. It does allow the demonstration of a
very useful featune of the system. Immediately after the error has
been notified, type:
WHERE
This will give the screen and line number, and a listing of the
offending line, where the FORTH system thinks the error has occurred.
We know about this error so now we can simply type:
FORTH FORGET TASK
to clear the dictionary and continue (WHERE enters the EDITOR
vocabulary, so FORTH must be used to return).
Remove the applications tape from the cassette and try using the
word:
SAVE
This is used to save the current contents of the tape buffer to tape.
No screen number is needed; the value of the user variable SCR is
assumed. The system will respond with:
RECORD TAPE
Do not switch the tape to record on this occasion (since there is no
tape in the recorder). Normally you would start the tape recording and
then press the space bar. Just press the space bar this time, and
wait. After a short pause the reminder
STOP TAPE
will appear and the system will wait until you press the space bar
again to indicate that you have done so. This reminder can save many
feet of blank tape! Now type 14 LIST again, to make sure that the
contents of the screen are still there, and then try typing:
CR 0 .LINE
57


which should type the contents of line 0 of the tape buffer. Now try:
EMPTY—BUFFERS l4 LIST
when you should find that an empty screen l4 is listed. The word
EMPTY-BUFFERS clears the contents of the tape buffer.
After this brief survey of the facilities of the tape interface we
are ready to try the screen editor.
8.2 The Screen Editor
8.2.1 Introduction
When a BASIC program has been written and is working satisfactorily it
may be saved as a named file. This is possible because a BASIC program
is stored in source form — very similar, if not identical, to the way
it is typed at the keyboard. Since FORTH is a compiled language it is
not stored in source form , but as a list of addresses of routines. It
is, therefore, to be expected that the writing of a FORTH application
which is to be saved on tape will be somewhat different from the
corresponding method for BASIC.
In FORTH the source text is edited into a screen, whose contents
can be compiled, tested and, if necessary, modified until its action
is correct. The screen can then be saved and any further screens
written in the same manner.
It is ai good idea to develop all but the very simplest of
applications by use of the editor since it allows the modification of
a definition without the need for excessive retyping. The editing
functions are placed in a separate EDITOR vocabulary. When the tape
interface and editor screens are loaded, note the difference between
the commands:
FORTH VLIST
and
EDITOR VLIST
8.2.2 A Sample Editing Session
The best way to learn the actions of the editing facilities is, as
with the rest of FORTH, by using them. The following is an example of
how the EDITOR vocabulary can be used to write, modify and save an
application.
Before the editor can be used a blank screen must be set up and
the EDITOR vocabulary declared. To do this the following word sequence
can be used:
DECIMAL
EMPTY-BUFFERS
300 SCR 1 ( we shall use screen 300 )
SCR @ LIST ( list this screen )
EDITOR
To simplify the process the word PROGRAM has been included in the
tape/editor application. This asks for the starting screen number and
automatically sets up the system for editing a screen. This example
saves a definition of RND , which generates a random number, and tests
58


its action. It is shown exactly as it would appear on the display.
PROGRAM
FIRST SCREEN NUMBER ? 300
scR # 300
9
l
Z
2
A
2
Q
1
Q5
0 P ( RANDOM NUMBER GENERATOR ) OK
l P DECIMAL OK
2 P 0 VARIABLE SEED OK
3 P : (RND) ( ...RAND ) OK
4 P SEED @ 259 * 3 + OK
5 P 32767 AND DUP SEED ! ; QQ
6 P : RND ( RANGE...RANDOM ) OK
7 P (RND) 32767 */ ; OK
L
scR # 300
0 (RANDOM NUMBER GENERATOR)
l DECIMAL
2 0 VARIABLE SEED
3 : (RND) ( ...RAND )
4 SEED @ 259 * 3 +
5 32767 AND DUP SEED 1 ;
6 : RND ( RANGE...RANDOM )
7 (RND) 32767 */ ;
95.
ENTER OK
SEED ? 0 OK
.0.?MSG#l
(RND) . 3 OK
.0.:>MsG#1
l0 RND . 0 OK
l0 RND . l OK
l0 RND . 8 OK
.0.?MSG#l
l 2 3 Q5
: TEST CR 0 DO DUP RND . LOOP DROP ; OK
20 l0 TEST
6 l7 l 17 ll 6 l8 l4 19 7 OK
. . . 3 2 l OK
SAVE
RECORD TAPE
( At this point, place a tape in the recorder, set it recording, and
press the space bar)
STOP TAPE
9.5
The random number generator is saved as screen 300 and may be entered
into the dictionary at a later time by:
300 LOAD
59


Screens may be numbered from O to 999 inclusive. Any attempt to save a
screen with ea screen number outside this range will cause an error
message (6) to be given.
The EDITOR word P will put the following text onto a line of the
screen. It expects to find the line number on the stack and is used
as:
n P text for line n
Since P is a FORTH word it must be separated from the text by a space.
It is good practice to leave a space after P, even if you then just
press RETURN (to clear a line of text) — otherwise an ASCII null will
be placed in the line and will stop interpretation of the screen at
this point.
It is conventional for line 0 of every screen to contain a comment
giving the contents of that screen. The FORTH word <(> is used to
start a comment and causes all text up to a right parenthesis <)> , or
to the end of the current line, to be ignored. Since <(> is a FORTH
word it must be separated by at least one space from any other text.
The right parenthesis is simply a delimiter and so need not be
separated by a space from the end of the comment. It is usual,
however, to leave a space to improve the appearance of the text. It
must, of course, be separated by one or more spaces from any following
words.
A FORTH application can be rather difficult to follow,
particularly if it makes great use of the stack so that variable names
do not appear. For this reason it is a good idea to use plenty of
comments when you unite aa screen. Since they are ignored when the
screen is interpreted they cost nothing in terms of dictionary space
or execution time. The only cost is in the use of cassette tape and
loading time.
The applications provided on-tape do not follow this rule and are
not commented. This is tx> reduce the loading time, since they are
likely to be used frequently. They are, in any case, fully documented
in this manual.
The word L is an EDITOR command to list the current screen and
saves typing
SCR @ LIST
every time.
When the screen is completed it should be fully tested before
being saved. In the example the words SEED (RND) and RND are tested in
turn. The word <.> is used frequently to check that nothing is left on
the stack.
As a final check the word TEST was defined and used to generate
ten random numbers in the range O to l9. The numbers l, 2 and 3 were
placed on the stack before TEST was used and printed again at the end.
This is a simple way to check that the words do not affect values
lower down on the stack.
When it is apparent that the words are operating correctly_ the
screen may be saved.
If an application extends to further screens the word -—> should
be put at the end of the screen. This is an instruction to continue
interpretation with the next screen on the tape (if the current screen
is screen n the next screen must be numbered n+l). The word ——> is
IMMEDIATE and so will execute even if the text is being compiled. This
means that it is possible for a single definition to extend over a
screen boundary. It is preferable not to do this as definitions
60


should, if possible, be kept short. It is, however, sometimes
necessary to write a definition (e.g. a machine code primitive) that
cannot be fitted into a single screen, particularly as the screens are
relatively short. A standard FORTH screen is sixteen lines of 64
characters, i.e. 1024 (lk) bytes. On the Acorn ATOM the screens are
half this size to enable the whole screen to be shown on the display
at once.
If -—> is not used at the end of a screen, interpretation will
stop at the end of the screen or at the word <;S> . This word
terminates the interpretation and may, therefore, be followed by
further comments or instructions, which do not need to be enclosed in
parentheses.
In the writing of an application extending to several screens the
word MORE can be used to save the current screen and set up the next
blank screen (it does not insert -—> at the end of the screen).
8.2.3 The Line Editor C mmands
So far the only word from the line editor we have used is P, which
puts text onto a given line. A complete list of line editor commands
appears below. Each command expects to find the relevant line number
on the stack. Most of the commands make use of the scratchpad area,
whose starting address is given by PAD, where a line of text can be
stored. Text is stored starting at PAD and working towards higher
addresses, as opposed to the use of PAD for numeric strings (see
Section 7.2.4).
Command Description
P Put text onto line.
D Delete the line, moving up the lower lines to close the gap,
but hold the deleted line at PAD (Does not work for line 7).
E Erase the line, leaving it blank. The contents of the line
are not saved.
H Hold the contents of the line at PAD. The line also
remains in the screen.
I Insert the text from PAD at the specified line. The lower
lines are moved down to make room for the insertion and
line 7 is lost.
R Replace the contents of the line with the text from PAD.
S Spread the text by inserting a blank line. Line 7 is lost.
T Type the contents of the line and also copy it to PAD. The
text remains in the screen.
In addition, the word TEXT will allow text to be put directly into
PAD. Like WORD it expects a delimiter character on the stack. It
accepts text from the keyboard up to the first appearance of the
delimiter, or until 64 characters are typed, or RETURN is pressed. It
is usual to use a delimiter that would not normally appear in the
input from the keyboard so that the end of text is marked by RETURN.
TEXT is used by all the EDITOR commands that put text at PAD with
ASCII code Ol as a delimiter character, i.e. it is used as:
l TEXT (... wait for text input)
It is worth setting up a dummy screen to practise using the line
6l


editor commands. Once you are familiar with each command, try to:
i) transfer a line to another position
ii) exchange two lines
Make sure that both of these work with line 7.
The following definition, as an example, will invert the order of
the lines in a screen.
EDITOR DEFINITIONS ( make sure it is in the EDITOR vocabulary )
: INVERT
8 O DO
7 H ( LINE 7 TO PAD )
FORTH I ( LOOP INDEX )
EDITOR I ( INSERT FROM PAD )
LOOP
L
It shows how a word can be defined to provide arbitrarily complicated
editing features. It also illustrates the way in which the vocabulary
structure can be manipulated to use both definitions of the word I .
Note that, since the vocabulary words are IMMEDIATE , they change the
CONTEXT vocabulapy for the following word(s) but do not themselves
appear in the compiled definition for which they are used.
8.2.4 The String Editor C mmands
The string editing facilities allow the location, insertion and
deletion of individual character strings withir1 a. screen. This is
accomplished with the aid of an editing cursor (displayed as #) whose
byte offset, from the start of the screen, is stored in the user
variable R# . The cursor is set to the beginning of the screen by the
word TOP .
The remaining string editing commands are given in the lists
below. They are divided into groups according to the type of input
they require.
The members of the first group need to be followed by typed text,
on which they act (like the line editing command P). There must, of
course, be one space between the command and the text, but (again like
P) any additional spaces are regarded as part of the text.
Command Description
C Insert the given text at the current cursor position.
F Find the given text and position the cursor immediately after
its first occurrence.
TILL Delete all text, from the current cursor position, to the end
of the text given. This command will only act on text within
a single line of the screen. ’
X Find and delete the first occurrence of the given text.
Each of the above commands will also leave the given text at PAD .
They give an error message if the text is not found in the screen.
The following example shows their actions. It is assumed that line
6 of the screen contains:
THIS IS A SILLY EXAMPLE
62


and the cursor has been reset by the use of TOP.
F IS A
THIS IS A# SILLY EXAMPLE 6 OK
C N
THIS IS AN# SILLY EXAMPLE 6 OK
TILL LY
THIS IS AN# EXAMPLE 6 OK
X PLE
THIS IS AN EXAM# 6 OK
The words of the following group require no additional input, but
expect to find their text at PAD.
Command Description
N Find the next occurrence of the text at PAD.
B Move the cursor back by the number of characters in the text
at PAD.
Each command in the next group requires a character count on the
stack.
Command Description
DELETE (count ...) Delete count characters, backwards from
the current cursor position.
M (count ...) Move the cursor by count characters, either
forwards or backwards depending on the sign
of count. The text itself is unchanged (0 M
is a simple way of displaying the current
cursor position).
Finally (we always save the best until last!) there is the word MATCH.
This is the string—matching routine used by all the words which search
for text. It is written, for speed, as a machine code primitive. Like
most machine code in FORTH, it is relocatable and so may be copied
from the tape/editor screen into your own application if you require a
routine to search an area of memory for a particular character string.
Its action on the stack is, however, a bit complicated.
The stack action is:
MATCH ( addrl\length\addr2\count ... f\offset )
using 4 and leaving 2 stack values.
The routine attempts to match the string, whose starting address is
addr2 and whose length is count bytes, to the contents of memory
starting at addrl and finishing at addrl + length. It leaves a flag,
which will be true (non—zero) if the match succeeds and false (zero)
if it fails, beneath the offset, from addrl, to the byte immediately
following the matched string (addrl + offset — count gives the address
of the start of the matched string).
8.3 A Note on Screen Numbering
The screen numbers, which may be in the range 0 to 999 inclusive, are
handled by the system in DECIMAL base, regardless of the current
numeric conversion base. If you are working, for example, in base l6
63


and you type in:
30 LOAD
the response will be:
>48 PLAY TAPE
since hexadecimal 30 is decimal 48. If you actually wanted screen 48
(decimal) then proceed as normal. Otherwise the load can be aborted by
pressing the space bar and then the CTRL key.
The sign—on message will appear and you can try again. Note that
the abort. will automatically· reset the numeric: conversion. base to
DECIMAL, so
30 LOAD
should now work as expected. You could have avoided this by typing
DECIMAL 30 LOAD
8.4 Using Discs
Although the FORTH system for the Acorn ATOM is designed for use with
the cassette operating system, it is fully compatible with floppy disc
drives using the Acorn DOS. The only change that you may like to make
is to remove the STOP TAPE prompt from the SAVE routine (tape
interface screen 7), since writing to the disc is controlled by the
system.
This can be done by loading the tape interface and editor, listing
screen 7 and using the EDITOR commands to delete the sequence
." STOP TAPE" KEY DROP CR
from line 7.
Screen 7 may then be re-saved, replacing the old version. A COLD
start, followed by a re-load of the tape interface and editor, will
allow the new version to be used.
8.5 Econet
ATOM FORTH is fully compatible with the Acorn Econet networking
system. This system is, however, designed mainly for use with discs
under the Acorn DOS. To load files or screens from tape it is
therefore necessary tx> activate the cassette operating system. This
can be done by using the *COS command before loading FORTH, or from
within FORTH by use of the MONITOR routine of Section ll.4. Remember
that in this case the '*' is not needed and the correct command is
MONITOR COS ‘.
64


9 cz 11
9.1 Introduction
You will probably have noticed that, when FORTH is executing, there is
a greater or lesser amount of interference on the display. This is an
unavoidable consequence of the use of the top 3K (#8COO to #97FF) of
the graphics memory for the user's dictionary and tape buffer. It
does, however, have one advantage in that it is possible to see when
the tape screens are loading and compiling! Because of this use, only
the lower half of the upper memory area is available for graphics and
the maximum resolution available is l28xl92.
If you do not need to use the higher resolution graphics, the
whole of the upper memory (with the exception of the VDU memory) can
be used as extra dictionary space. All that is necessary is to change
the value of the dictionary pointer before any applications are
loaded. When you have loaded the nucleus dictionary from tape, enter:
HEX 8200 DP z
and 8200 FENCE 1
New definitions will then start loading from #8200 upwards, giving an
extra 2.5K of dictionary space. The lowest graphics mode, with a
resolution of 64x48, is still available.
It is also possible to compromise, e.g. setting the initial value
of the dictionary pointer and FENCE to #8600 will give an extra l.5K
of dictionary space and allow graphics up to a resolution of l28x96.
Note that a COLD start will reset the upper dictionary space to start
at #8C00.
The graphics package, provided in screens 18 to 2l of the
applications, assumes that the upper dictionary area is set to start
at #8COO and therefore requires no alterations to the system. The
graphics package can be loaded, with or without the presence of the
EDITOR vocabulary, by typing:
18 LOAD
in the usual way.
9.2 Graphics Modes
Four modes of graphics are provided, similar to the ATOM BASIC
graphics modes 0, l, 2 and 3. The resolutions and the memory used are
given below.
Mode: Resolution Memory:
X: Y: #8000 tO:-
0 B 64 48 #8lFF (0.5K)
l 128 64 #83FF (1K)
2 128 96 #85FF (1.5K)
3 128 192 #8BFF (BK)
The graphics modes are set by typing:
n CLEAR
65


(after having declared the GRAPHICS vocabulary) where n may be 0, 1, 2
or 3. The word CLEAR expects to find this number on the stack and it
may be either typed in, as in the above example, or left on the stack
as the result of a calculation performed by some other word. A
9.3 Point—P1otting
The five words connected with point—plotting are:
Word Description
BLACK All subsequent plots are in black.
WHITE All subsequent plots are in white.
INVERT All subsequent plots invert the point.
PLOT (X\Y ...) Plot a point at the position with co-ordinates
(X,Y). The co-ordinates of the point are kept
in the GRAPHICS variables X and Y.
(PLOT) (X\Y ...) As PLOT but X and Y are not updated.
One of the words BLACK, WHITE or INVERT must be executed before any
use of PLOT or (PLOT). The 'co1our' will remain in effect for all
subsequent uses.
The following example illustrates the use of these procedures.
GRAPHICS DEFINITIONS (definitions to be in the GRAPHICS vocabulary)
: DIAGONAL
0 DO I I PLOT LOOP ;
Then execute
O CLEAR 48 WHITE DIAGONAL
3 CLEAR 192 DIAGONAL
and
12 EMIT (to return to text mode)
The next example shows the possibilities for animated graphics.
: FLASH
3 CLEAR INVERT  
BEGIN
192 DIAGONAL ?ESC
UNTIL 12 EMIT
Pressing the ESC key will end the display and clear the screen. The
use of animated graphics is restricted by the interference appearing
on the screen. This can only be eliminated by adding extra memory from
#3COO upwards (minimum of 2K) so that the dictionary, user variables
and tape buffer does not need to use the graphics memory (see Section
9.7 and Appendix D).
9.4 Line-Drawing
A line may be drawn from the last plotted position to the point with
66


coordinates (x»Y} by use of the LINE (x\y ...). The points on the line
may be plotted, erased or inverted by previous use of WHITE , BLACK or
INVERT as in the case of PLOT .
To draw a line from the point (0,0) to the point (15,25), for
example, the following sequence can be used.
0 X 1 0 Y 1 WHITE 15 25 LINE
A triangle with vertices at (X0,Y0), (Xl,Yl), (X2,Y2) may be drawn by
the definition:
GRAPHICS DEFINITIONS
: TRIANGLE (X0\Y0\Xl\Yl\X2\Y2 ...)
DUP >R Y 1 ( SET X & Y TO X2 & Y2 )
DUP >R X 1 ( AND ALSO SAVE THEM FOR LATER )
LINE LINE ( DRAW 2 SIDES )
R> R> LINE ( RECOVER COORDINATES & DRAW THIRD SIDE )
It leaves the coordinates of the final point i.e. X2 and Y2 in the
graphics variables X and Y, and may be used, for example, as
3 CLEAR 10 15 73 20 100 180 WHITE TRIANGLE
The variables XDIR , YDIR , DELTAX , DELTAY , ERR and the words SETXY
and (LINE) are used internally by LINE and are not intended to be
executed directly by the user.
The word MOVE (X\Y ...) will change the current plot position to
the point with coordinates (X,Y) without any plotting action. We
could, for example, redefine TRIANGLE as:
: TRIANGLE
2DUP >R >R
MOVE ( TO X2,Y2 )
LINE LINE
R> R> LINE ,
which has the same action as the earlier definition.
9.5 Relative Plotting
The words RPLOT , RLINE and RMOVE have similar actions to PLOT , LINE
and MOVE , except that the x and y values on the stack are interpreted
as being relative to the last point plotted. The word REL is used to
convert the relative coordinates to absolute values.
9.6 Clearing the Screen
The screen may be cleared and returned to text mode by:
: CLS l2 EMIT ;
9.7 Use of Mode 4 Graphics
Very few changes need to be made to the Graphics Package to enable the
use of mode 4 graphics with the applications dictionary moved to
extension memory below #8000. They will all fit into the existing
screens 18 — 2l, with a little editing.
67


Screen 18:
Insert the following as line 6:
: 4MODE F7AA F0 00 1800 ;
Modify the new line 7 to read:
CASE: NMODE OMODE IMODE 2MODE 3MODE 4MODE ; ——>
Screen l9:
On line l, replace the words:
FFFC AND
with:
ABS 4 >
No further changes are necessary.
68


10 What, to produce a dictionary entry for later (level 0) execution. All
(levels higher than 0 result in a new entry being made in the
dictionary. This chapter is concerned with the next higher level,
level 2, in which new defining words are formed. The sequence of
operations which is involved is:
a) generate a new defining word (level 2),
b) use the defining word to produce a new dictionary entry
(level l),
c) execute the new entry (level 0).
One higher level is possible: to produce alternative ways of
generating defining words. This level, which is often termed
"meta—FORTH", enables the writing of totally new FORTH-like languages,
and is beyond the scope of this manual.
Each defining word in FORTH can be considered as a mini—compiler,
dedicated to compiling a particular type of structure into the
dictionary. If a new structure is required, e.g. an array, a new word
is required to allow its compilation. Just as the generation of a new
word (level 1) extends the FORTH language, the generation of a new
defining word (level 2) extends the FORTH compiler.
The two words  are used for this purpose. The <
and > signs (which are not pronounced!!) are present to indicate that
the two words should be used together and to show their order of use.
10.2 The Actions of 
The two words are used in a definition of the following form (level
2):
: FAMILY  ..... ;
where an optional list of words may follow each of the two. This is,
in one sense, an ordinary colon—definition and all the words are
compiled in the normal way. The use of  , however,
make it a level 2 definition.
The words following  determine the action of the new word, and though they are
compiled into the definition of FAMILY , they are not executed until
the new word is used. It is important to remember that the  words at
execution time.
The execution of FAMILY takes the form:
FAMILY MEMBER
and may expect one or more values on the stack, depending on the
69


definition of FAMILY . This is a level l process and creates a
dictionary entry for the word MEMBER .
When  takes the address of the first word after
DOES> in the dictionary entry of FAMILY and places it in the two-byte
space reserved by  a routine that
will handle this execution.
When MEMBER is executed (level 0) the address of its parameter
area is placed on the stack and then the words following DOES> , in
FAMILY , are executed. Execution of all words defined by FAMILY begins
with this code, so FAMILY produces a group of words with related
actions.  use the address of the parameter
area to gain access to these items, allowing each of the words created
by the same defining word to have its own function.
10.3 The use of 
Some simple examples may make the use of these words more clear.
Let us first have a look at an alternative definition of VARIABLE.
This word appears in the nucleus dictionary and its action has been
described in Section 5.3.2. It creates a dictionary entry with space
for a single variable, and initialises it. The following definition of
VARIABLE is identical except that the values of the variables it
creates are not initialised.
: VARIABLE  ;
When this is executed by typing:
VARIABLE SIZE
 so when SIZE is executed it just leaves the address of the
parameter field on the stack. This gives access to the value, which is
initally indeterminate, through ! and @ as normal.
The definition of the VARIABLE in the nucleus dictionary would be:
: VARIABLE  ;
Instead of allotting space, the value on the top of the stack is
compiled into the parameter area by <,> .
The definition of CONSTANT is:
: CONSTANT  @ ;
The compilation stage is identical to that of VARIABLE , but when the
word defined by CONSTANT is used, @ leaves the value on the stack.
70


We may also create single byte variables and constants by:
: CVARIABLE  ;
and
: CCONSTANT  C@ ;
These both expect a value in the range 0 to 255 on the stack to
initialise the value of the dictionary entry they create.
10.4 Arrays and Tables
The use of  to create new types of data structure can
be illustrated by the extension of FORTH to handle arrays.
10.4.1 0ne—dimensiona1 Arrays
A simple definition for a one—dimensional array is:
: ARRAY
 SWAP 2* +
A ten—element array of single precision variables is created by:
10 ARRAY VALUES
The words after  2DUP ( DUPLICATE INDEX AND PARAMETER ADDRESS )
@ OVER < ( CHECK IF INDEX TOO LARGE )
7l


SWAP 0< OR ( OR IF NEGATIVE )
5 ?ERROR ( ISSUE ERROR MESSAGE IF NEEDED )
2+ ( OTHERWISE STEP OVER MAXIMUM INDEX VALUE )
SWAP 2* + ( AND CONVERT INDEX TO ELEMENT ADDRESS )
7
If a more specific error message is required, the words 5 ?ERROR may
be replaced with, for example,
IF DROP CR
." RANGE ERROR - ARRAY INDEX = " .
QUIT
THEN
The inclusion of error checks, such as that given above, has the
disadvantage that it decreases the speed of execution. A solution to
this problem is to develop an application using full error checks
until it is working correctly. When it is certain that no errors can
occur, the words containing error checks can be replaced by simpler,
faster, versions. If an application is developed by use of the editing
facilities described in Chapter 8, it is a simple matter to change
these words as the remainder of the application is unchanged.
10.4.2 Two—Dimensiona1 Arrays
The following definition allows the creation of two—dimensional
arrays. The elements are single precision variables ·and time array
contents are not initialised. No index checking is done but error
checks could be added in a similar manner to that given for
one—dimensional arrays.
: 2ARRAY
 ROT ( GET FIRST INDEX TO TOP )
OVER @ * ( MULTIPLY BY STORED INDEX )
ROT + ( ADD SECOND INDEX )
2* ( CALCULATE BYTE OFFSET )
+ ( ADD TO BASE ADDRESS )
2+ ( STEP OVER STORED INDEX )
F
The use is C
10 5 2ARRAY RECTANGLE
to create a 10 by 5 array called RECTANGLE. In this example the array
indices may range from 0,0 to 9,4 inclusive. The address of, say, the
2,3 element is left on the stack by:
2 3 RECTANGLE
10.4.3 Tables
It may be necessary to create a table of values for which only the
starting address is needed. This type of structure can be implemented
72 t 2


very simply as follows:
: CTABLE

This, when used in the form
l0 CTABLE DATA
will create the word DATA with space for ten single-byte values. When
DATA is executed it will leave on the stack the starting address of
the data table.
The word LIST of Section 6.3 is an example of a table which leaves
both its start address and the number of l6—bit items it contains. It
may be created by use of the following definition of the wordx@ABLE .
: TABLE
 DUP 2+ ( get start address )
SWAP @ ( max number of items )
LIST is then created by
n TABLE LIST
where n is the required maximum number of items.
10.5 Strings
There are many ways of implementing string handling in FORTH. Two
examples are given in 'BYTE' magazine, in the August 1980 and February
1981 issues.
The following example shows a simple alternative method to handle
strings up to 255 characters in length.
: STRING ( max length ... )

l+ ( STEP OVER MAXIMUM LENGTH )
An empty string is then created by, for example:
l0 STRING WORDS
The string variable WORDS may now hold any character string up to 10
characters in length. A few additional words are required for input
and output of strings.
The definition of SIN uses the constant C/L , which gives the
number of characters per line in the display (i.e. 64). Remember also
that PAD returns the start address of the scratchpad area used for
text (Section 8J2.3) and for numeric conversion (Section 7.2.4).
73


: $IN ( ... addr\length )
HERE C/L 1+ BLANKS ( CLEAR MEMORY AT HERE )
l WORD ( INPUT STRING TO HERE )
( TERMINATED BY CARRIAGE RETURN )
HERE PAD C/L 1+ CMOVE ( MOVE STRING TO PAD )
PAD DUP C@ 1+ ( PREPARE TO MOVE STRING ... )
; ( INCLUDING COUNT BYTE )
: $1 ( from addr\length\to addr ... )
2DUP l — C@ 1+ > ( CHECK IF SPACE FOR STRING )
IF CR ." STRING OVERFLOW " ( IF NOT GIVE ERROR )
2DROP DROP QUIT ( CLEAR STACK & QUIT )
THEN
SWAP CMOVE ( OTHERWISE STORE STRING )
: $@ ~ ( addrl ... addr2\length )
COUNT ( PREPARE TO TYPE STRING )
The following shows how these words are used, assuming that the string
variable WORDS has been created as in the earlier example.
$IN HELLO OK
WORDS $1 QK
WORDS $@ TYPE SPACE HELLO OK
If the words LEFTS and RIGHTS of Section 7.3.2 are also defined, the
following examples can be tried.
WORDS 2 LEFTS TYPE SPACE HE OK
WORDS 3 RIGHTS TYPE SPACE LLO OK
10.6 A CASE Statement
10.6.1 Introduction
The conditional structure of Section 6.2 allows a two—way branch using
IF ... ELSE ... THEN
A CASE statement allows a branch to one of many possible word
sequences with a return to a common point. There are two basic methods
for the selection of the case to be executed. The first is a
'positional' case where the values to be tested are restricted to the
first n integers. The second method is a 'keyed' case where a value is
tested against a sequence of explicit values which need not be in
numerical order.
10.6.2 A Positional CASE
The following simple example ci E3 positional CASE will select the
words to be executed by means of an integer value on the stack. The
value must be in the range from zero to one less than the number of
cases available ixx the particular example. No error checks are made
for a number outside the permitted range. This CASE structure is used
in the graphics package to select the resolution mode, and has an
added check to restrict the choice to the integers from 0 to 3
inclusive.
74


Here is the definition of the defining word CASE: :
: CASE:
 SWAP 2*
+ @
EXECUTE
The word EXECUTE takes the execution (code field) address of a word
from the stack and executes the word's definition. Thus
' WARM ( get parameter field address of WARM )
CFA , ( convert to code field address )
EXECUTE
has the same effect as executing WARM directly from the keyboard.
To use the case structure it is first necessary to define each of
the possible actions, for example:
: NOTHING
." CASE O DOESN'T DO MUCH " ;
: BELL
." CASE l RINGS THE BELL "
7 EMIT`;
: HOME
." CASE 2 HOMES THE CURSOR "
30 EMIT ;
These actions are then included in a case structure for, say, the word
TEST .
CASE: TEST
NOTHING BELL HOME ;
When TEST is being created, SMUDGE ensures that the new entry will be
found in a dictionary search and ] then sets compilation mode, so that
the words following TEST will have their addresses compiled into the
dictionary entry.
When TEST is executed by:
O TEST A
1 TEST
or
2 TEST
the words following DOES> convert the case number to a pointer to the
address of the correct word in the list, and execute it.
Note that time CASE statements cnf many high—level languages are
based on GOTO—type control transfers, whereas this FORTH CASE has the
options compiled into the definition of the case word so that the
choice is fixed before execution. Basically, this is because it is not
easy to handle forward references, i.e. words that have not yet been
defined, in FORTH.
For a further discussion of a variety of possible forms for case
statements in FORTH see 'FORTH Dimensions' Vol.2 lk>;3 (1980) (see
Appendix E).
75


76


11.1 Fast Divide—by-Two
Division reqires a large number of operations and is usually very slow
in a microcomputer. The following routine is a machine code primitive
which will divide by two. Its action is identical with
2 /
but is about sixty times faster.
HEX
CREATE 2/ O
18 Cy   I   [   [
2DO , lF6 , 1FO , 38 C,
176 , 76 , 4C C, 2842 ,
SMUDGE
Assembly Listing
18 CLC
B5 01 I LDA l,X
10 09 BPL 2DlV
F6 00 INC 0,X
DO 02 ‘ BNE NOINC
F6 01 INC l,X
F0 Ol NOINC BEQ 2DIV -
38 SEC
76 Ol 2DIV ROR l,X
76 00 ROR 0,X
4C 42 28 JMP NEXT
11.2 Recursion
~A recursive routine is one which uses itself. In FORTH a dictionary
entry cannot find a reference to itself while it is being defined.
This problem is solved by defining the IMMEDIATE word MYSELF.
: MYSELF
LATEST ( get name field address of latest word )
PFA ( convert to parameter field address )
CFA ( convert this to code field address )
, ( compile the address )
; IMMEDIATE
11.2.1 Factorials
The following example shows its use to form a recursive definition to
calculate factorials.
77


: (FACT) ( nl\n2 ... n3 )
—DUP IF DUP ROT * SWAP
l - MYSELF
THEN
: FACT ( n ... )
DUP O< OVER 7 > OR 5 ?ERROR
1 SWAP (FACT) .
F
The calculation of the factorial is performed by (FACT), which leaves
the result on the stack. Attempting to calculate factorials of numbers
greater than 7 will cause an arithmetic overflow. Factorials of
negative numbers are not defined. Error checking is confined to FACT.
This then uses (FACT) to calculate the result, which is displayed.
The following, alternative, definition of FACT makes use of MD*,
defined in Section 4.4. The result is left as a double precision
number, allowing the calculation of factorials of numbers up to and
including 12.
: (FACT)
—DUP IF DUP 1 —
>R MD* R>
MYSELF
THEN
: FACT ( n ... )
DUP O< OVER 12 > OR 5 ?ERROR
1 O ROT (FACT) D.
11.2.2 Sorting an Array -
The word QUICKSORT ( nl\n2 ...) uses the quicksort algorithm to sort
elements n1 to n2 inclusive of the array NUMBERS into increasing
numerical Order.
Before typing in the following definitions you must first define
2/ ( Section 11.1 )
ARRAY ( Section 10.4.1 )
MYSELF ( Section 11.2 )
O VARIABLE TEMP ( TEMPORARY STORAGE )
256 ARRAY NUMBERS ( OR WHATEVER SIZE YOU WISH )
: EXCHANGE ( nl\n2...) ( EXCHANGE ELEMENTS nl & n2)
( OF NUMBERS >
NUMBERS DUP >R @ SWAP
NUMBERS DUP @ >R
1 R> R> !
: PARTITION ( nl\n2 ... n3\n4 ) ( SPLIT ARRAY INTO SMALLER SECTIONS,
EXCHANGING ELEMENTS WHERE NECESSARY )
BEGIN 2DUP > O= WHILE
SWAP BEGIN DUP NUMBERS @ TEMP @ < WHILE 1+ REPEAT
2DUP EXCHANGE SWAP 1+ SWAP 1 —
REPEAT
78


F
: QUICKSORT ( n1\n2 ... )
2DUP 2DUP + 2/ NUMBERS @ TEMP 1 PARTITION
>R ROT DUP R < R> SWAP
IF MYSELF ELSE 2DROP THEN 2DUP >
IF SWAP MYSELF ELSE 2DROP THEN
1 The routine can be checked by filling NUMBERS with integers in reverse
order by, for example
: NFILL ( n ... )
DUP 0 DO
DUP I NUMBERS ! 1 -
LOOP DROP
Then execute, for example,
100 NFILL
which should fill the first 100 elements of NUMBERS with reverse—order
integers. Executing
0 99 QUICKSORT
should then leave the elements of NUMBERS in ascending order. The
contents can be checked by, for example:
: NCHECK ( n ... ) 0 DO I NUMBERS ? LOOP ;
Then execute, for example:
100 NCHECK
11.3 A Screen Copying Utility
COpy1I’1g SCIEGHS, p&I't1CU.16.I'1y with 3 JC&p€··b£':lS€d V€I`S1OI`1, C8!} be VQIY
tedious. The following routines will allow the copying of up to five
screens at a time from one tape to another. The screen contents are
temporarily stored in the graphics memory from #8200 to #8BFF. The
screens may be renumbered during the copying process but the input and
output screens must be numbered consecutively. They require the prior
loading of the tape interface.
BASE @ ( SAVE CURRENT BASE )
HEX
: OFFKEY ( DISABLE KEYBOARD )
-1 20A +! ;
: ONKEY ( RE-ENABLE KEYBOARD )
l 20A +! ;
: PAUSE ( WAIT FOR A KEYPRESS )
KEY DROP CR ;
: INSCR ( first screen no.\no. of screens to input ... )
CR ." PLAY " PAUSE
79


OFFKEY
0 DO
DUP I + LIST ( LOAD AND LIST SCREEN )
FIRST 200 DUP
I * 8200 +
SWAP CMOVE ( MOVE TO GRAPHICS MEMORY )
LOOP DROP
ONKEY
: OUTSCR ( first screen no.\no. of screens to output ... )
ovmz sca 1 ( sm nsw SCREEN N0. )
." RECORD" PAUSE
OFFKEY
0 DO
I 200 * 8200 + ( TRANSFER FROM )
FIRST 200 CMOVE ( GRAPHICS MEMORY )
SAVE
1 SCR +! ( INCREMENT SCREEN NO. )
LOOP DROP
ONKEY
-1 SCR +! ( RESET FOR NEXT BATCH )
: COPY ( first input screen no.\first output screen no.\
no. of Screens ... )
5 MIN ( ENSURE NOT MORE THAN 5 SCREENS )
>R SWAP R ( KEEP NO. OF SCREENS ON RETURN STACK )
INSCR
R> ( RECOVER NO. OF SCREENS )
." NEW TAPE " PAUSE
OUTSCR
2
BASE ! ( RESTORE ORIGINAL BASE )
To copy screens 6, 7 and 8 to a new tape, numbered as 15, 16 and 17,
execute:
6 15 3 COPY
Separate use of INSCR and OUTSCR will allow intermediate editing of
the screens provided each screen is moved from the graphics memory to
the tape buffer, edited and then returned to the same area of the
graphics memory. In HEX,
n FIRST 200 CMOVE ( graphics —> tape buffer )
FIRST n 200 CMOVE ( tape buffer —> graphics )
where n may be
8200, 8400, 8600, 8800 or 8A00,
depending on which of the 5 screens is to be edited.
80


11.4 Use of the Operating System Monitor Routines
It may be necessary, from within FORTH, to use the cassette operating
system commands. The following definitions allow this to be done
either by direct execution from the keyboard or from within a
definition.
BASE @ HEX ‘
CREATE OSCLI ( call the OSCLI routine of the monitor )
8E86 , 20 C, FFF7 ,
8EA6 , 4C C, 2842 ,
SMUDGE
: (MONITOR)
1 WORD HERE PAD C/L 1+ CMOVE ( TRANSFER KEYED INPUT TO PAD )
PAD COUNT ( GET ADDRESS & LENGTH OF STRING )
2DUP + OD SWAP C! ( ADD 'RETURN' TO END )
1+ ( INCREASE STRING COUNT )
100 SWAP CMOVE CR ( TRANSFER TO COS BUFFER )
IN @ 60 IN 3 ( PUT RTS HERE FOR MONITOR )
OSCLI ( INTERPRET & EXECUTE )
IN 1 ( RESTORE VALUE OF IN )
: MONITOR
STATE @ IF ( COMPILING A DEFINITION )
COMPILE QUERY
COMPILE
THEN
(MoN1ToR)
; IMMEDIATE
BASE !
Note that THEN does not create a compiled address but only marks the
end of a conditional, for calculation of an offset. The second COMPILE
will, therefore, compile (MONITOR) and not THEN.
Any of the Cassette Operating System commands of Chapter 19 of
'Atomic Theory and Practice' pages 139 to 142 may then be used. Note
that the initial * of these commands should not be used. For example,
to use the COS command *CAT to obtain a catalogue of a tape, the
sequence:
MONITOR CAT
should be used (and ¤ot MONITOR *CAT).
If MONITOR is used in a definition, it will, on execution of the
definition, wait for the command to be entered from the keyboard. This
allows the command to be selected at execution time rather than being
fixed at the time of definition.
11.5 WAIT
The following example is an implementation of the WAIT instruction of
ATOM BASIC, which waits until the next 'tick' of the 60Hz sync signal.
Each tick is signalled by a zero in the most significant bit of Port C
of the 8255 PIA, at address #B002 in the ATOM.
A simple implementation is as follows:
81


HEX C
: WAIT
BEGIN BO02 C@ 80 < UNTIL
BEGIN B002 C@ 7F > UNTIL ;
DECIMAL
This is equivalent to the machine code subroutine at #FE66 in ATOM
BASIC.
11.6 Using the Internal Loudspeaker
11.6.1 Generating Tones
The simplest way of generating a tone from the internal loudspeaker is
to TOGGLE bit 2 of the output port at address #3002. This is done by
the word BLIP1 which is used by TONEI to generate a short note.
HEX
: BLIP1
B002 4 TOGGLE ;
: TONEI
100 0 DO BLIP1 LOOP ;
A second tone may be generated by toggling the speaker twice within a
definition.
: BLIP2
B002 4 2DUP TOGGLE TOGGLE ;
: TONE2
100 0 DO BLIP2 LOOP ;
The words BOP and BIP are useful for sound effects for paddle—type
games:
: BOP ’
30 0 DO BLIP1 LOOP ;
: BIP l
30 0 DO BLIP2 LOOP ;
Finally, we can produce a warble tone by:
: WARBLES
0 DO BIP BOP LOOP;
This is used by typing, for example:
l0 WARBLES
82


11.6.2 Music
The following definitions will allow the keyboard to be used to play
music via the internal speaker.
CREATE NOTE ( 0\FRE\LEN ... )
EAEA , EAEA , 6A0 , 88 C,
FDD0 , 18 C, 2B5 , 475 , _
495 , 3B5 , 575 , 595 ,
890 , 4A9 , 4D C, B002 ,
8D C, B002 , 5BO , EAEA ,
EAEA , EA C, B5 , 5D0 ,
1D6 , 18 C, 590 , EAEA ,
EAEA , EA C, D6 , C7D0 ,
IB5 , C6D0 , E8E8 , 4C C,
291E ,
SMUDGE
DECIMAL
: PITCH ( FREQUENCY ... FRE )
256 5 */ ;
: TONE ( FREQUENCY ... )
PITCH 0 SWAP 2500 NOTE ;
: TABLE ( BYTESIZE ... )
 ;
20 TABLE KEYS 40 TABLE FREQUENCIES
: CFILL ( BYTEVALS\TABLENAME\BYTELENGTH ... )
0 DO _
DUP I +
>R SWAP R> C!
LOOP DROP
32 93 91 64 59 80 76 75
73 74 85 72 89 71 70 82 68
69 83 65 KEYS 20 CFILL
: TFILL ( VALS\TABLENAME\LENGTH ...)
l 0 DO
DUP I 2* +
>R SWAP R> !
LOOP DROP
0 683 640 608 576 542 512 483
456 457 406 387 352 341 320 304
288 271 256 243 FREQUENCIES 20 TFILL
: KBD ( C\ADDR ... OFFSET )
0 BEGIN
>R 2DUP
R 19 > IF CR ." ?" 2DROP 1
ELSE R + c@ =
THEN R> 1+ SWAP
83


UNTIL x
1 -
;
: KEYBOARD
BEGIN
KEY KEYS KBD
>R 2DROP R>
2* FREQUENCIES + @ TONE
?ESC UNTIL
F
The tones produced by NOTE vary in timbre as well as pitch and are not
'pure' unless the value of FRE is an integral power of two. This is
because the speaker is only toggled when the value third on the stack
overflows. This occurs at slightly irregular intervals unless the
above condition is met. One effect is an apparent shift in the pitch
of some notes. The values in FREQUENCIES are therefore not exactly
those expected for a true scale, but are chosen for the best perceived
scale.
Executing KEYBOARD allows the notes to be played. Pressing the ESC
key will terminate execution. The arrangement of the keys used is as
follows:
EBI Y U I IEE
1\£3[]I?(3[§;II

Examples:
: DROPTEST
620 0 TIMER-ON DROP .TIME ;
DROPTEST 48 MICROSECONDS OK
(i.e. the word DROP executes in 48 microseconds)
: DUPTEST
620 TIMER-ON DUP .TIME DROP ;
DUPTEST 7l MICROSECONDS OK
: CONTEST
668 TIMER-ON 0 DROP .TIME ;
CONTEST 77 MICROSECONDS OK
This last example gives the execution time of a constant, in this case
the constant 0 . Its value must be dropped from the stack before
executing .TIME so the execution time of DROP is added to the time
adjustment, giving a total adjustment of 668 microseconds.
The following table gives execution times for some of the more
common words.
WORD: TYPE: EXECUTION TIME (MICROSECONDS) :
DROP CODE 48
DUP CODE 7l
OVER CODE 7l
ROT FORTH 424
PICK FORTH ll68
RoLL FoRTH 948l + l636(n-4) ( n = 4, 5, 6 ...>
(NOTE: l ROLL = NOOP, 2 ROLL = SWAP, 3 ROLL = ROT)
@ conn 80
! CODE 84
DO ... LOOP CODE 80 + ll6n (n = no. of loops)
I CODE 79
* CODE 759 to 1379
depending on the number of non—zero
bits in the multiplier
/ CODE 4230 to 4394
largely depending on the number of digits
in the result.
The words * and / are colon definitions but most of the execution time
is spent in the machine—code primitives U* and U/ .The relatively long
execution time for ROLL is due to it being (like PICK) a FORTH
definition, rather than a nmchine code definition, and including an
error check. It is particularly slow since it involves a fairly high
degree of movement of the stack contents.
85


11.8 Further Graphics
The following application will plot a figure similar to that shown on
the front cover of this manual. It uses recursive calls to 3SIDES to ·
plot three sides of a rectangle, their orientation being controlled by
a case statement (see Section 10.6.2).
The application requires the GRAPHICS package to have been loaded
previously. This contains the definition of CASE: . The variables X
and Y are defined in the FORTH vocabulary and are therefore distinct
from the X and Y in the GRAPHICS vocabulary. Note that this
duplication is not detected by the system since, when FORTH is the
CONTEXT vocabulary, no other vocabularies are searched.
FORTH DEFINITIONS DECIMAL
4 CONSTANT N 128 CONSTANT HO
O VARIABLE X O VARIABLE Y
O VARIABLE XO 0 VARIABLE YO
O VARIABLE H
: YSCALE ( SCALE Y VALUE TO FILL SCREEN )
3 2 */ 7
: XYLINE ( DRAW LINE TO X,Y*3/2 )
X @ Y @ YSCALE GRAPHICS LINE ;
: X+ ( LINE IN +VE X-DIRECTION )
H @ X +! XYLINE ;
: Y+ ( LINE IN +VE Y-DIRECTION )
H @ Y +! XYLINE ;
: X- ( LINE IN —VE X—DIRECTION )
H @ MINUS X +1 XYLINE ;
: Y- ( LINE IN —VE Y-DIRECTION )
H @ MINUS Y +! XYLINE ;
: BSIDES ( INDEX\CASE NO ... INDEX )
OVER DUP ( 2 COPIES OF INDEX )
IF ( NON—ZERO INDEX )
l - SWAP ( DECREMENT INDEX AND )
( BRING CASE NO. TO TOP )
[ HERE H I ] NOOP
( RESERVE SPACE FOR ORIENTATION )
( - DEFINED LATER — AND SAVE )
( ADDRESS IN H )
ELSE 2DROP ( CASE NO. AND INDEX )
THEN
: ORA ( THESE DETERMINE THE 4 ORIENTATIONS )
3 3SIDES X-
O 3SIDES Y-
O 3SIDES X+
1 3SIDES DROP
86 `


: ORB
2 3SIDES Y+
1 3SIDES x+
1 3SIDES Y-
O 3SIDES DROP
  Y
: ORC
1 3SIDES X+
2 3SIDES Y+
2 3SIDES X-
3 3SIDES DROP
: ORD
O 3SIDES Y-
3 3SIDES X-
3 3SIDES Y+
2 3SIDES DROP
7
CASE: ORIENTATION
ORA ORB ORC ORD
' ORIENTATION CFA H @ 1
( PLACE ADDRESS OF ORIENTATION IN 3SIDES )
( TO COMPLETE RECURSION )
: INITIALISE
HO DUP H 1
2 / DUP XO 1 YO 1
GRAPHICS 3 CLEAR WHITE
7
: XYSET ( START POSITION AND SIZE FOR EACH PLOT )
H @ 2 / DUP H 1
2 / DUP XO +1 YO +1
x0@Y0@
2DUP Y 1 X 1
YSCALE GRAPHICS MOVE
: PLOT—IT
INITIALISE
O BEGIN 1+ ( INCREMENT INDEX )
XYSET
O 3SIDES
DUP N =
UNTIL
DROP
KEY DROP 12 EMIT
When the application has been entered the figure is displayed by
executing PLOT—IT. The number of iterations is governed by the
constant N . Its value may be changed by, for example:
3 ' N 1
87


In graphics mode 3, the largest value of N for a clear display is 5.
However an interesting textured effect can be produced by changing the
WHITE in INITIALISE by, for example:
GRAPHICS ' INVERT CFA ( EXECUTION ADDRESS OF INVERT )
' XYSET NFA 4 — ( LOCATION OF WHITE IN INITIALISE )
!
and executing PLOT-IT with N = 6.
11.9 Non-destructive Stack Print
The following short application will allow the stack contents to be
displayed without destroying them. It is useful, for example, in the
development and testing of an application.
It requires the previous loading of 2/ (Section ll.l).
Alternatively the somewhat slower <2 /> may be used. In addition the
silent user variable S0 must be given a dictionary header by
6 USER S0
The definitions are as follows: P
: DEPTH ( ... n ) ( returns the number of stack items )
SP@ S0 @ SWAP — 2/ ;
: .S ( ... ) ( non-destructive stack display )
CR DEPTH
IF S0 @ 2 — SP@ SWAP
DO I ? -2 +LOOP
ELSE ." EMPTY "
THEN
The stack items are printed with the top stack item on the right.
88


Most detected errors in ATOM FORTH result in an error message of the
form:
? cccc MSG # n
where cccc is the word where FORTH thinks the error has occurred. The
general rule in error handling is that both the return and computation
stacks are cleared. The one major exception is error message 4,
indicating the redefinition of an existing word, when the message is
simply a warning. The message number is printed in the current base
so may not be immediately recognisable. In the following error
message list the message number is given in decimal and hex.
DECIMAL HEX Message
0 0 Unrecognised word or invalid character
l l Empty stack
2 2 Dictionary full
3 3 Has incorrect address mode (Assembler)
4 4 Not unique (warning only)
5 5 Index or parameter outside valid range
6 6 Tape/disc screen number out of range
7 7 Full stack
8 8 (Reserved for disc use)
9 9 )
l0 A ) N
ll B ) User definable
l2 C )
l3 D )
l4 E )
15 F (Reserved for disc use)
16 l0 (Reserved for disc use) ‘
l7 ll Compilation only
l8 l2 Execution only
l9 l3 Conditionals not paired
20 l4 · Definition not finished
2l l5 In protected dictionary
22 l6 Use only when loading
23 l7 Off current editing screen
24 l8 Declare vocabulary
89


90


This glossary contains all words present in ATOM FORTH. Each entry is
of the following form:
Word Stack Action Uses Leaves Status Pronunciation
followed by a description and in many cases a numerical example.
The computation (parameter) stack action is shown, where appropriate,
as a list of the values and their types before and after the execution
of the word, in the form:
(stack contents before ... stack contents after)
In all references to the stack, numbers to the right are at the top of
the stack. The notation nl\n2 is read as "nl is beneath n2". The
symbols used to represent the different stack value types include:
n l6—bit (single precision) signed number
u 16-bit (single precision) unsigned number
addr l6-bit address (unsigned)
nd 32-bit (double precision) signed number
ud 32-bit (double precision) unsigned number
b 8-bit one-byte number (unsigned)
c 7-bit ASCII character
count 6-bit string length count
f boolean flag: O = false, non-zero = true
ff boolean false flag = 0 _
tf boolean true flag = non-zero
The number of stack values that the word uses and leaves are also
shown. Some words have an additional letter indicating their status.
A only for the Acorn ATOM — not a standard FORTH word
C may only be used in a colon definition
E intended for execution only
P has precedence bit set; will execute even when in compile mode
Where not obvious, the standard pronunciation is given in square
brackets after the status.
The glossary contains all words that are immediately available to
the user when the basic system is ltmded. This includes headerless
entries (see Appendix C), which are listed with their code field
(execution) addresses. Since they have no names in the dictionary
their names are completely arbitrary, but the names given are those
preferred in a standard FORTH system.
9l


! (n\addr ...) 2 0 [store]
Stores the value n at the address addr.
before: 7 35 -1234 4128
after: 7 35
(-1234 is stored in the two bytes from address 4128)
!CSP 0 0 [store C—S-P]
Stores the stack pointer value in user variable CSP. Used as part
of the compiler security.
{ (ndl ... nd2) 2 2 [sharp]
Converts the least-significant digit (in the current base) of the
double-precision number ndl to the corresponding ASCII character,
which it then stores at PAD . The remaining part of the number is
left as nd2 for further conversions. # is used between <# and #> .
If BASE is DECIMAL
before: 9 32 1234567
after: 9 32 123456
(ASCII code #37 is stored in PAD )
{> (nd ... addr\count) 2 2 [sharp-greater]
Terminates numeric output conversion by dropping the double number
nd and leaving the address and character count of the converted
string in a form suitable for TYPE .
.Hbefore: 23 19 0 0
after: 23 19 4128 3
( TYPE would display the 3-character string starting at address
4128)
{S (ndl ... nd2) 2 2 [sharp-S]
Converts the double-precision number ndl into ASCII text by
repeated use of # , and stores the text at PAD . The
double-precision number nd2 is left on the stack, and has a value
of zero. #S is used beween <# and #> .
before: 27 5 1234567 ‘
after : 27 5 0000000
(ASCII codes #37, #36, #35, ... #31 are in consecutive memory
locations at PAD )
' (... addr) 0 1 P [tick]
(during execution)
0 0 `
(during compilation)
Used in the form ' nnnn and leaves the parameter field address of
dictionary word nnnn if in execution mode.
If used within a colon definition it will execute to compile the
address as a literal numerical value in the definition.
92


( P [paren]
Used in the form ( nnnn ) to insert a comment. All text nnnn up to
a right parenthesis on the same line is ignored. Since ( is a
FORTH word it must be followed by a space. A space is not
necessary before ) since it is only used as a delimiter for the
text.
) is pronounced "close—paren".
(+LOOP) (n ...) 1 0 [bracket-p1us—1oop]
Headerless code; execution address #28EO. The run—time procedure
compiled by +LOOP that increments the loop index by the signed
quantity n and tests for loop completion. See +LOOP .
(.') [bracket—dot—quote]
Headerless code; execution address #3lE4. The run-time procedure
compiled by ." that transmits the following in-line text to the
output device. See ." .
(;CODE) C
Headerless code; execution address #3146. The run—time procedure
that rewrites the code field address of the most—recently defined
word to point to the machine-code following (;CODE) . It is used
by the system defining words ( <:>, CONSTANT etc.) to define the
machine-code actions of dictionary entries using them.
This is, in a sense, a machine-code version of DOES> .
(ABORT) [bracket—abort]
Headerless code; execution address #347A. Executes after an error
when WARNING is -1. It normally causes the execution of ABORT but
the contents of the parameter area can be changed (with care) to
point to to a user—defined error—handling procedure.
(D0) c
Headerless code; execution address #2910. The run—time procedure
compiled by DO that moves the loop control parameters to the
return stack. See DO .
(ENTER) 0 0 [bracket—enter]
Headerless code; execution address #3AF5. Interprets the current
contents of the tape input buffer.
(FIND) (addr1\addr2 ... pfa\b\tf) 2 3 [bracket—find]
(found)
(addr1\addr2 ... ff) 2 1
(not found)
Headerless code; execution address #2955. Searches the dictionary
starting at the name field address addr2 for a match with the text
starting at addrl. For a successful match the parameter field
address and length byte of the name field plus a true flag are
left. If no match is found only a false flag is left.
93


(LOAD) (addr\f ...) 2 0 A [bracket-load]
Headerless code; execution address #3B69. The
implementation-dependent routine used by LOAD and —-> to load
screens from tape or disc. The addr is that of the zero-page data
required by the operating system. This data is completed by the
creation of aa 3-digit file name from the screen number and the
insertion of its address as the first item. If the rest of the
data is created by OSDATA then addr must be #62. The flag
determines the appearance of the prompt on the display. In all
cases an indication is given of the screen number for which a
search is being made. If the flag is false a further prompt PLAY
TAPE with a wait for a keypress is given. If the flag is true
(non-zero) these further actions do not occur.
(LOOP) [bracket-loop]
Headerless code; execution address #28BA. The run-time procedure
compiled by LOOP that increments the loop index by one and tests
for loop completion. See LOOP .
(NUMBER) (nd1\addrl ... nd2\addr2) 3 3 [bracket-number] .
Headerless code; execution address #33B7. Converts the ASCII text
beginning at addrl + 1 according to the current numeric conversion
base. The new number is accumulated into ndl, being left as nd2.
Addr2 is the address of the first non-convertable digit. (NUMBER)
is used by NUMBER .
* (nl\n2 ... n3) 2 1 [times]
Leaves as n3 the product of the two signed numbers nl and n2.
before: 7 -3 9
after: 7 -27
*/ (nl\n2\n3 ... n4) 3 1 [times-divide]
Leaves as n4 the value nl*n2/n3. The product nl*n2 is kept as a
double precision intermediate value, resulting in a more accurate
result than can be obtained by the sequence nl n2 * n3 / .
before: 7 3 17 5
after: 7 10
*/MOD (nl\n2\n3 ... n4\n5) 3 2 [times-divide-mod]
Leaves, as n4 and n5 respectively, the remainder and the integer
value of the result of nl*n2/n3. The product nl*n2 is kept as a
double precision intermediate value, resulting in a more accurate
result than can be obtained by the sequence nl n2 * n3 /MOD .
before: 7 3 17 5
after: 7 1 10
+ (nl\n2 ... n3) 2 1 [plus]
Leaves as n3 the sum of nl and n2.
before: 19 7 24
after: 19 31
94


+1 (n\addr ...) 2 0 [plus—store]
Adds n to the value at addr.
before: 25 -2 8427 (addr 8427 contains 29, for example)
after: 25 (addr 8427 now contains 27)
+— (nl\n2 ... n3) 2 1 [plus-minus]
Leaves as n3 the result of applying the sign of n2 to nl.
before: l7 4 -7
after: 17 -4
+LOOP (n ...) 1 0 P,C [plus-loop]
Used in colon definition in the form:
DO ... +LOOP
During execution +LOOP controls branching back to the
corresponding DO , dependent on the loop index and loop limit. The
loop index is incremented by n, which may be positive or negative.
Branching to DO will occur until
a) for positive n, the loop index is greater than or equal to the
loop limit, or
b) for negative n, the loop index is less than or equal to the
loop limit.
Execution then continues with the word following +LOOP .
+ORIGIN (n ... addr) l l [plus—origin]
Leaves the address of the nth byte after the start of the boot-up
parameter area. Used to access or modify the boot-up parameters.
, (n ...) 1 0 [comma]
Stores (compiles) n in the first two available bytes at the top of
the dictionary and increments the dictionary pointer by two.
— (nl\n2 ... n3) 2 1 [subtract]
Leaves as n3 the difference nl — n2.
—-> P [next screen]
Continues interpretation with the next screen of source code from
tape.
—DUP (ff ... ff) l l ( [dash—dup]
or (tf ... tf\tf) 1 2
Duplicates the top stack value if it is true (non-zero).
-FIND (... pfa\b\tf) 0 3 [dash—find]
(if found)
(... ff) O l
(if not found)
Used as —FIND nnnn . The CONTEXT and then the CURRENT vocabularies
are searched for the word nnnn . If found, the entry's parameter
field address, name length byte, and a true flag are left;
otherwise just a false flag is left. ”
95


—TRAILING (addr\n1 ... addr\n2) 2 2 [dash-trailing]
Changes the character count nl of the text string at addr so as
not to include any trailing blanks, and leaves the result as n2.
. (n ...) 1 0 [dot]
Prints the number n on the terminal device in the current numeric
base . The number is followed by one blank space.
.' P [dot—quote]
Used as ." cccc"
In a colon definition the literal string cccc is compiled together
with the execution address of a routine to transmit the text to
the terminal device.
In the execution mode the text up to the second " will be printed
immediately.
.R (nl\n2 ...) 2 0 [dot—R]
Print the number nl at the right—hand end of a field of n2 spaces.
Unlike <.> no following space is printed.
/ (nl\n2 ... n3) 2 0 [divide]
Leaves the value n3 = nl / n2.
before: I3 27 6
after: 13 4
/MOD (nl\n2 ... n3\n4) 2 2 [divide—mod]
Leaves the remainder n3 and quotient n4 of nl/n2. The remainder
has the sign of the dividend.
before: 13 27 6
after: 13 3 4
0,1,2 (... n) 0 1
These often—used numerical values are defined as constants in the
dictionary to save both time and dictionary space.
0< (n ... f) 1 1 [zero-less]
Leaves a true flag if n is less than zero, otherwise leaves a
false flag.
0= (n ... f) 1 1 [zero—equa1s]
Leaves a true flag if n is equal to zero, otherwise leaves a false
flag.
OBRANCH (f ...) 1 0 [zero—branch]
Headerless code; execution address #28A2. The run—time procedure
to cause a conditional branch. If f is false the following in—line
number is added to the interpretive pointer to cause a forward or
backward branch. It is compiled by IF , UNTIL and WHILE .
1+ (nl ... n2) 1 1 [one—p1us]
Increments nl by one to give n2.
96


2* (nl ... n2) 1 1 [two-times]
Multiplies nl by two to give n2. Faster in execution than 2 * .
2+ (nl ... n2) 1 l [two—plus]
Increments nl by two to give n2.
ZDROP (nd ...) 2 0 [two-drop]
Drops the double-precision number nd (or two single precision
numbers) from the stack.
2DUP (nd ... nd\nd) 2 4 [two—dup]
Duplicates the top double-precision number (or the top two
single-precision numbers) on the stack.
: P,E [colon]
Used to create a colon definition in the form
: cccc .... ;
Creates a dictionary entry for the word cccc as being equivalent
to the sequence of FORTH words until the next <;>. Each·word in
the sequence is compiled into the dictionary entry, unless its
precedence bit is set (P), in which case it is executed
immediately.
; P,C [semi—colon]
Terminates a colon definition and stops further compilation.
;S P [semico1on—S]
Stops interpretation of a screen from tape. It is also the word
compiled by <;> at the end of a colon definition to return
execution to the calling procedure.
< (nl\n2 ... f) 2 1 [less—than] .
Leaves a true flag if nl is less than n2, otherwise leaves a false
flag.
before: 15 2 17
after: 15 l (true)
<# [less—sharp]
Sets up for numeric output formatting. The conversion is performed
on a double number to produce text at PAD . See also # , #> , #S ,
SIGN .
 .... ;
to create a new defining word cccc
When cccc is executed in the form:
cccc nnnn
a new dictionary entry is created for nnnn with a name and a
parameter area produced by  .
97


When nnnn is executed it has the address of its parameter area
(defined by  in cccc .
 (nl\n2 ... f) 2 l [greater than]
Leaves a true flag if nl is greater than n2, otherwise leaves a
false flag.
before: l2 0 -1
after: l2 l
>R (n ...) 1 0 C [to—R]
Removes a number from the computation stack and places it on the
return stack. Its use must be balanced with R> in the same
definition. It is used temporarily to remove a number from the
stack to access a lower number. See R> .
? (addr ...) 1 0 [questionemarkl
Prints the value contained in the two bytes starting at addr.
Equivalent to <@ .> .
?COMP [query—comp]
Issues an error message if not compiling.
?CSP [query-C—S—P]
Issues an error message if stack position differs from that saved
in CSP . Used as part of the compiler security.
?ERROR (f\n ...) 2 0 [query-error]
Issues error message number n if the boolean flag f is true. Uses
ERROR . The stack is always empty after an error message.
98


?ESC (... f) 0 1 [query-esc]
Tests the keyboard to see if the ESC key is depressed. A true
(non—zero) flag is returned if the key is down at the time of the
test, otherwise a false (zero) flag is returned. In many FORTH
systems this function is carried out by the word ?TERMINAL which
may test for any key being pressed.
?EXEC [query—exec]
Issues an error message if not executing.
?LOADING [query-loading]
Issues an error message if not loading from tape.
?PAIRS (n1\n2 ...) 2 0 [query—pairs]
Issues an error message if nl does not equal n2. The message
indicates that compiled conditionals (IF ... ELSE ... THEN or
BEGIN ... UNTIL etc.) do not match. It is part of the compiler
security. The error message is given if, for example, the sequence
IF ... UNTIL is found during compilation of a dictionary entry.
?STACK [query-stack]
Issues an error message if the stack is out of bounds.
?TERMINAL
See ?ESC .
@ (addr ... n) 1 1 [fetch]
Leaves on the stack the l6—bit value s found at addr.
before: ll 4123
after: ll 375
(assuming the value 375 was stored in the two bytes from address
4123).
ABORT
Clears both stacks, enters the execution state, prints the
start—up message on the terminal device and returns control to the
keyboard.
ABS (n ... u) 1 1
Leaves u as the absolute value of n.
before: 12 -17
after: l2 l7
AGAIN P,C
Used in a colon definition in the form
BEGIN ... AGAIN
During execution of a word containing this sequence, AGAIN forces
a branch back to the corresponding BEGIN to create a endless loop.
99


    OCC) 1- 0
The value of n is added to the dictionary pointer to reserve n
bytes of dictionary space. The dictionary pointer may be moved
backwards by use of a negative n but this should be used with
caution to avoid losing essential dictionary content.
AND (nl\n2 ... n3) 2 1
Leaves as n3 the bit-by—bit logical AND of nl and n2.
(assuming binary)
before: 1101 1010 1100
after: 1101 1000
BACK (addr ...) 1 0
Calculates the backward branch offset from HERE to addr and
compile into the next available dictionary memory address. Used in
the compilation of conditionals ( AGAIN UNTIL etc.)
A user variable containing the current number base used for input
and output conversion.
BEGIN 0 1 P,C
Used in a colon definition in the forms:
BEGIN ... AGAIN
BEGIN ... UNTIL
BEGIN ... WHILE ... REPEAT
BEGIN marks the start of a sequence that may be executed
repeatedly. It acts as a return point from the corresponding AGAIN
, UNTIL or REPEAT .
B1. (... c) 0 1 [B-L]
A constant that leaves the ASCII value for 'blank' or 'space' (Hex
20).
BLANKS (addr\n ...) 2 0
Fill n bytes of memory starting at addr with blanks.
BLK (... addr) 0 1 [B—L-K]
A user variable indicating the input source. If BLK is zero, input
is taken from the keyboard. If it is non-zero input is taken from
the tape input buffer area. I
BRANCH d
Headerless code; execution address #288D. The run—time procedure
to cause an unconditional branch. The following in-line value is
added to the interpretive pointer to cause a forward or backward
branch. It is compiled by ELSE , AGAIN and REPEAT .
100


C! (b\addr ...) 2 0 [C—store]
Stores byte b (8 bits) at addr.
before: 53 29 3127
after: 53
(29 is stored in the single byte at address 3127)
C, (b ...) 1 0 [C-com a]
Stores (compiles) k>;h1 the next available dictionary byte,
advancing the dictionary pointer by one.
C/L (... n) 0 1 [C—slash-L]
A constant containing the number of characters per line. This is
normally 64, so a full FORTH 'line' will occupy two lines of the
VDU display.
CG (addr ... b) 1 1 [C—fetch]
Leaves b as the 8-bit contents of addr.
CFA (pfa ... cfa) l 1 [C-F-A]
Converts the parameter field address of a word to its code field
(execution) address.
CLIT (... n) 0 1 [C—lit]
Headerless code; execution address #285B. Within a
colon-definition CLIT can be compiled before an 8-bit literal
value. When the word containing CLIT is later executed the 8-bit
value (in the range 0-255) is pushed to the stack as a
single-precision (16-bit) number with its most—significant part
set to zero.
This word is used by a number of system words but is not available
to the user via the keyboard interpreter, which uses only LIT .
CMOVE (fr m\to\count ...) 3 0
Moves 'count' bytes, starting at 'from' to the bdock of memory
starting at 'to'. The byte at 'from' is moved first and the
transfer proceeds towards high memory. No check is made as to
whether the destination area overlaps the source area.
COLD
The cold start procedure used on first entry to the system. The
dictionary pointer and user variables are initialised from the
boot—up parameters and the system re-started via ABORT . It may be
called from the keyboard to remove all application programs and
restart with the nucleus dictionary alone.
COMPILE
COMPILE acts during the execution of the word containing it. The
code field (execution) address of the word following COMPILE is
compiled into the dictionary instead of executing, cf. [COMPILE] .
- 101


CONSTANT (n ...) 1 0
A defining word used in the form
n CONSTANT cccc
to create a word cccc , with the value n contained in its
parameter field. When cccc is executed the value n will be left on
the stack.
CONTEXT (... addr) 0 l
A user variable leaving the address of a pointer to the VOCABULARY
in which a dictionary search will start.
COUNT (addrl ... addr2\n) 1 2
Leaves the address addr2 and byte count n of a text string
starting at addrl, in a form suitable for use by TYPE . It is
assumed that the text string has its count byte at addrl and that
the actual character string starts at addrl + l.
(assuming that the text string 3 65 66 67 starts at 6124),
before: 47 6124
after: 47 6125 3
( TYPE would then display the 3 characters ABC)
CR [c-R1
Transmit a carriage return and line feed to the terminal output
device.
CREATE V
A defining word used in the form
CREATE cccc
to create a dictionary header for the word cccc with its code
field pointing to the first byte of the parameter field.
One common use is, with the aid of <,>, to compile machine code
into the parameter area to produce a machine code primitive. This
does not need an assembler vocabulary.
CSP (... addr) 0 1 [C—S—P]
A user variable used for temporary storage of the stack pointer in
checking of compilation errors.
CURRENT (... addr) 0 1
A user variable containing a pointer to the vocabulary into which
new definitions will be placed. As soon as a definition is made in
the CURRENT vocabulary, it automatically becomes also the CONTEXT
vocabulary.
D+ (ndl\nd2 ... nd3) 4 2 [D—plus]
Leaves as nd3 the double number sum of double number ndl and nd2.
102


D+- (nd1\n ... nd2) 3 2 [D—p1us—minus]
Applies the sign of single number n to the double number ndl,
leaving the result nd2. See +— .
D. (nd ...) 2 0 [D-dot]
Prints the signed double number nd according to the current BASE.
One blank is printed after the number. See <.>.
D.R (nd\n ...) 3 0 [D-d0t—R]
Prints a signed double number nd on the right of a field n
characters wide. See .R . No trailing blank is printed.
¤ABs (nd   ud) 2 2
Leaves the absolute value ud of a signed double number nd. See
ABS .
DECIMAL
Sets BASE to decimal numeric conversion for input and output.
DEFINITIONS
Sets the CURRENT vocabulary to the CONTEXT vocabulary. If used in
the form
cccc DEFINITIONS
where cccc is a VOCABULARY word, all subsequent definitions will
be placed in the vocabulary cccc .
DIGIT (c\nl ... n2\tf) 2 2
(valid)
(c\n1 ... ff) 2 1
(invalid)
Converts ASCII character c, wih base nl, to its binary equivalent
n2 and a true flag. If c is not a valid character in base nl, then
only a false flag is left.
In hexadecimal base, but displaying stack in binary.
a) Valid
Hex character 'D' has ASCII code #44, or 01000100 in binary,
and represents the value 1101 in binary
before: 00010110 01000100 00010000
after: 00010110 00001101 00000001
b) Invalid
Character 'G' has ASCII code #47, or 01000111 in binary, and
does not represent a hexadecimal value
before: 00010110 01000111 00010000
after: 00010110 00000000
103


DLITERAL (nd ...) 2 0 P
(compiling)
In the compiling state a double number nd is compiled as a double
literal number in the dictionary. Later execution of the word
including this literal number will replace nd on the stack.
In the execution mode DLITERAL has no effect.
nuiuus (ndl ... nd2) 2 2
Change the sign of ndl, leaving it as nd2.
no (n1\n2 ...) 2 0 P,c
May only be used within a colon definition in the forms
nl n2 DO ... LOOP
nl n2 DO ... +LOOP g
This is the equivalent of a FOR ... NEXT loop in BASIC, repeating
a sequence of operations a fixed number of times. The value of nl
is the loop limit and n2 is the initial value of the loop index.
The loop terminates when the loop index equals or exceeds the
limit. The sequence of operations in the loop will always be
executed at least once. See I , LOOP , +LOOP , LEAVE .
DOES> [does]
Used with  in cccc . At the start of this sequence
the address of the parameter field of nnnn will be put on the
stack so that the execution can refer to the particular values
associated with nnnn . See 

ELSE P,C
Used in a colon definition in the form
IF ... ELSE ... THEN ,
During execution ELSE causes a branch to the words after THEN if
the flag tested by IF was true, and is the destination of the
branch taken at IF if the flag was false. See IF .
Burr (c ...) 1 0
Transmits ASCII character c to the output device. The contents of
OUT are incremented for each character output.
before: 23 65
after: 23
(A is displayed on the output).
ENCLOSE (addr\c ... addr\n1\n2\n3) 2 4
Headerless code; execution address #29Bl. The text—scanning
primitive used by WORD .
The text starting at addr is searched, ignoring leading
occurrences of the delimiter c, until the first non-delimiter
character is found. The offset from addr to this character is left
as nl. The search continues from this point until the first
delimiter after the text is found. The offsets from addr to this
delimiter and to the first character not included in the scan are
left as n2 and n3 respectively. The search will, regardless of the
value of c, stop on encountering an ASCII null (00) which is
regarded as an unconditional delimiter. The null is never included
in the scan. ‘
Examples:
Text at addr nl n2 n3
ccABCDcc 2 6 7
ABCDcc 0 4 5
ABCOcc 0 3 3
Occc 0 1 O
ERASE (addr\n ...) 2 0
Sets n bytes of memory starting at addr to contain zeros.
muon (n ...) 1 0
Gives an error notification. The value in WARNING is examined and
Lf it is -1 a system ABORT is executed, via (ABORT) which is a
headerless dictionary entry. Otherwise an error message number n
is given, the stacks are emptied and control is returned to the
keyboard. In the system as supplied, WARNING is set to zero so the
error message is given. Changing WARNING to -1 and also the vector
in (ABORT) allows the user to define his own error response.
EXECUTE (addr ...) 1 0
Executes the definition whose code field (execution) address is on
the stack.
105


EXPECT (addr\count ...) 2 0
Transfers characters from the keyboard to the memory starting at
addr until a RETURN (#0D) is found or until the maximum count of
characters has been received. Backspace deletes characters from
both the display and the memory area but will not move past the
starting point at addr. One or more nulls are added at the end of
the text.
FENCE (addr ...) 1 0
A user variable containing an address below which the user is not
allows to FORGET . In order to use FORGET on an entry below this
point it is necessary to alter the contents of FENCE . In the
ATOM, changing the contents of FENCE to a value less than #8000
(the start of the upper block of RAM) may produce unpredictable
results for FORGET .
FILL (addr\n\b ...) 3 0
Fills n bytes of memory starting at addr with the value b.
FIRST (... n) 0 1
A constant that leaves the address n of the first byte of the tape
input/output buffer area.
FORGET _ E
Used in the form
FORGET cccc
to delete the definition with name cccc and all dictionary entries
following it. An error message is given if the CURRENT vocabulary
is not the same as the CONTEXT vocabulary, i.e. if the entry cccc
is not in the vocabulary that is searched first. An error message
is also given if cccc is in the protected area of the dictionary,
below FENCE . Regardless of the value stored in FENCE the nucleus
dictionary, all of which is necessary for the correct operation of
the system, is protected against FORGET .
FORTH P
The name of the primary vocabulary. Execution makes FORTH the
CONTEXT vocabulary. It is IMMEDIATE so it will execute if used
during the creation of aa colon definition. Until other
vocabularies are defined, all new words become a part of FORTH .
All other vocabularies ultimately link to the FORTH vocabulary.
HERE (... addr) 0 1
Leaves the contents of DP i.e. the address of the first unused
byte in the dictionary.
HEX
Sets the numeric conversion BASE to sixteen (hexadecimal).
HLD (... addr) 0 1 [H-L-D]
A user variable containing the address of the latest character of
text produced during numeric output conversion (by # ).
106


    OOO) 1 0
Used between <# and #> to insert an ASCII character c into a
converted numeric string. 2E (hex) HOLD will place a decimal point
in the string.
I (OOC   0 1 C
Used in a DO ... LOOP to place the current value of the loop index
on the stack. It must be used at the same level of nesting as the
DO ... LOOP , i.e. it will not operate correctly if included in a
colon definition word beween DO and LOOP .
ID. (addr ...) 1 0 [I-D—dot]
Prints the name of a word from its name field address on the
stack.
IF (15 ...) 1 0 1>,c
Used in a colon definition in the forms
a) IF (true) ... THEN
b) IF (true) ... ELSE (false) ... THEN
If the flag f is true the sequence of words after IF is executed
and execution ij; then transferred tx> the word immediately
following THEN . If f is false execution transfers
a) to the word following THEN , or
b) to the sequence of words following ELSE and subsequently to the
first word after THEN .
IMMEDIATE
Sets the precedence bit of the most recently defined word so that
it will execute rather than being compiled during the compilation
of a word definition. See [COMPILE] .
IN (... addr) 0 l
A user variable containing the byte offset to the present position
in the input buffer (terminal or tape) from where the next text
will be accepted.
INTERPRET
The outer text interpreter which either executes or compiles a
text sequence, depending on STATE , from the current input buffer
(terminal or tape). If the word name cannot be found after a
search of the CONTEXT and then the CURRENT vocabularies, it is
converted to a number using the current base. If this conversion
also fails an error message is given.
If a decimal point is found as part of a number the position of
the decimal pointer will be stored in DPL and a double number will
be left on the stack. The number itself will not contain any
reference to the decimal point.
KEY (... c) 0 1
Leaves the ASCII value of the next terminal key pressed.
LATEST (... addr) 0 l
Leave the name field address of the most recently defined word in
the CURRENT vocabulary.
107


LEAVE C
Forces the termination of a.IX) ... LOOP at the first following
time that LOOP or +LOOP is reached. This is done by setting the
loop limit equal to the current value of the loop index, which is
not changed. Execution will continue normally until reaching LOOP
or +LOOP .
LFA (pfa ... lfa) 1 l [L-F-A]
Convert the parameter field address, pfa, to its link field
address, lfa.
LIMIT (... addr) 0 1
A constant leaving the address of the first byte after the highest
memory available for the tape I/O buffer.
LIT 0 l C (
Within a colon definition, LIT is automatically compiled before
each 16-bit literal number encountered in the input text. Later
execution of LIT causes the contents of the following two bytes to
be pushed onto the stack.
    OOO)  
During compilation the stack value n is compiled into the
dictionary entry as a l6—bit (single-precision) number.
A possible use is
: nnnn ... [calculate a value] LITERAL ... ;
Compilation is suspended (by [ ) for a value to be calculated and
then resumed (by ] ) for LITERAL to compile the value into the
definition of nnnn.
Loma (n ...) 1 0
Searches the tape file for screen n and, when found, loads it into
the tape buffer. The PLAY TAPE response is as described in *LOAD,
page l40 of "Atomic Theory and Practice", with the addition that
the message '>n' is given, indicating that a search is being made
for screen n. The contents of the buffer is then interpreted.
Interpretation will end at the end of the screen or at ;S unless
-—> is found, in which case loading will continue with screen n+l.
LOOP P,C
Used in a colon definition in the form
DO ... LOOP
During execution LOOP controls branching back to the corresponding
DO , dependent on the loop index and loop limit. The loop index is
incremented by one and tested against the loop limit. Branching to
DO continues until the index is equal to or greater than the limit
when execution continues with the word following LOOP .
M* (n1\n2 ... nd) 2 2 [M-times]
Leaves as the double—precision number nd the signed product of the
two signed single—precision numbers nl and n2.
108


M/ (nd\n1 ... n2\n3) 3 2 [M—divide]
Leaves, as n2 and n3 respectively, the signed remainder and signed
quotient from the division of the double number dividend nd by the
single divisor nl. The sign of the remainder is that of the
dividend.
M/MOD (ud1\u2 ... u3\ud4) 3 3 [M—divide—mod]
Leaves, as ud4 and u3 respectively, the double quotient and
remainder from the division of the double dividend udl by the
divisor u2. All are unsigned integers.
MAX (nl\n2 ... max) 2 1
Leaves as max the larger of nl and n2.
MESSAGE (n ...) 1 0
Prints on the output device error message number n.
MIN (nl\n2 ... min) 2 1
Leaves as min the smaller of nl and n2.
MINUS (nl ... n2) 1 1
Changes the sign of nl and leaves the result as n2. The sign is
changed by forming the two's complement.
Leaves as mod the remainder of nl/n2 with the sign of nl.
NFA (pfa ... nfa) 1 1 [N—F—A]
Converts the parameter field address of a definition to its name
field address.
uuc1·o1> (... addr) 0 1 A
Headerless code, execution address #2822. A constant containing
the address of the top of the nucleus dictionary. Used by FORGET
to prevent accidental forgetting of the nucleus dictionary, all of
which is required for the correct operation of the system.
NUMBER (addr ... nd) 1 2
Converts the character string starting at addr with a character
count byte, to the signed double number nd using the current
numeric base. If the string contains a decimal point its position
will be given in IKEA. If a valid numeric conversion is not
possible an error message will be given. Used by INTERPRET .
NOOP [no—op]
A no—operation in FORTH. One possible use is to reserve address
space in a colon—definition for later over—writing by the
execution address of a subsequent definition.
OR (nl\n2 ... or) 2 1
Leaves as or the bit-by-bit logical OR of nl and n2.
109


OSDATA (addr1\n ... addr2) 0 1 A [O—S—data]
Headerless code; execution address #3B4B. The
implementation—dependent routine that creates the zero-page data
required by the operating system load or save, with the exception
of the pointer to the filename (see "Atomic Theory and Practice"
p.l9l). The beginning of this data, left as addr, is at #62.
OSLOAD (addr\f ...> 2 0 A ~ [0-s—1¤ad]
Headerless code; execution address #3B24. The
implementation—dependent machine code used by (LOAD) to make the
appropriate call to the operating system load routines. The
meanings of addr and f are as for (LOAD) .
OUT (... addr) 0 1
A user variable containing a value that is incremented by EMIT .
It may be examined and changed by the user to control display
formats.
OVER (nl\n2 ... nl\n2\nl) 2 3
Copies the second stack item over the top item.
Before: 15 23 17
after: 15 23 17 23
PAD (... addr) 0 1
Leaves the address of the text output buffer. This is always 68
bytes above HERE . Numeric output characters are stored downwards
from PAD , character text is stored upwards.
PFA (nfa ... pfa) 1 1 [P-F—A]
Converts the name field address, nfa, of a dictionary entry to its
parameter field address, pfa.
QUERY
Inputs up tx> 80 characters terminated by RETURN (#0D) from the
keyboard. The text is stored in the terminal input buffer whose
address is given by TIB. The value of IN is set to zero (in
preparation for interpretation by INTERPRET ).
QUIT
Clears the return stack, stops compilation and returns control to
the keyboard. No message is given.
R (... n) 0 1—
Copy the top cu? the return stack to the computation stack. The
action is identical to that of I .
R0 (... addr) 0 1 [R-zero]
A silent user variable (no dictionary entry) containing the
initial address of the top of the return stack. It may be given a
header in the dictionary by:
8 USER R0
110 i


R# (... addr) 0 1 [R-sharp]
A user variable which contains the location of the editing cursor
for the Editor.
R> (... n) 0 1 [R—from]
Removes the top value from the return stack and leaves it on the
computation stack . See >R .
REPEAT P,C
Used in a colon definition in the form
BEGIN ... WHILE ... REPEAT
In execution REPEAT forces an unconditional branch back to BEGIN .
1201.1. (n ...) 11
Rotates the top nl stack items so that the nth item is moved to
the top.
1 ROLL has no effect
2 ROLL is equivalent to SWAP
3 ROLL is equivalent to ROT
An error message is given if n is less than l.
ROT (n1\n2\n3 ... n2\n3\n1) 3 3
Rotates the top three items on the stack, bringing the third item
to the top
before: 92 17 28 12
after: 92 28 12 17
RP! [R—P-store]
Initialises the return stack pointer.
RP@ (... addr) 0 1 [R-P—fetch]
Leaves the address of the return stack pointer. Note that this
points one byte below the last return stack value.
S->D (n ... nd) 1 2 [S—t0-D]
Leaves as nd the signed single-precision number n converted to the
form of a signed double-precision number (with unchanged value).
S0 (... addr) 0 1 [S-zero]
A silent user variable (no dictionary entry) containing the
address which marks the initial top of the computation stack. It
may be given a header in the dictionary by:
6 USER SO
SCR (... addr) 0 1 i· [S-C-R]
A user variable containing the number of the most recently
referenced source text screen.
lll


SIGN (n\nd ... nd) 3 2
Stores an ASCII '-' sign in the converted numeric output string at
PAD if n is negative. The sign of n is usually that of the double
number to be converted. Although n is discarded the double number
nd is kept either for further conversion or to be dropped by #> .
SIGN may only be used between <# and #> .
SMUDGE
TOGGLEs the 'smudge bit' in the name header of the most recently
created definition in the CURRENT vocabulary. This switches
between enabling and disabling the finding of the entry during a
dictionary search.
The name field is smudged during the definition of a word to
prevent the incomplete definition from being found, and then
smudged again on completion.
SP! 0 0 [S—P—store]
Initialises the computation stack pointer.
SPG (... addr) 0 1 [S-P—fetch]
Leaves the value of the stack pointer on the stack. The value
corresponds to the state of the stack before the operation.
before: 1 2
(address 58 56 54) .
after: l 2 56
SPACE
Transmits an ASCII blank to the output device.
SPACES (n ...1 1 0
Transmits n ASCII blanks to the output device.
STATE (... addr) 0 1
A user variable indicating the state of compilation. A zero value
indicates execution and a non—zero value indicates compilation.
SWAP (nl\n2 ... n2\nl) 2 2
Exchanges the top two items on the stack.
before: 17 23 59
after: 17 59 23
THEN P,C
Used in a colon definition in the forms
IF ... THEN
IF ... ELSE ... THEN
Marks the destination of forward branches from IF or ELSE as the
conclusion of the conditional structure . See IF .
TIB (... addr) 5 0 1
A user variable containing the address of the terminal input
buffer.
112 Q


TRAVERSE (addr1\n ... addr2) 2 1
Headerless code; execution address #2FB8. Moves across the name
field of a dictionary entry. If n=l, addrl should be the address
of the name length byte (i.e. the NFA of the word) and the
movement is towards high memory. If n=-l, addrl should be the last
letter of the name and the movement is towards low memory. The
addr2 that is left is the address of the other end of the name.
TOGGLE (addr\b ...) 2 0
Complements the contents of addr by the bit pattern b.
before: b = 00110000, contents of addr = 01101010
after: contents of addr = 01011010
TYPE (addr\count ...) 2 0
Transmits count characters of ai string starting at addr to the
output device.
U* (u1\u2 ... ud) 2 2 [U—times]
Leaves the unsigned double-precision product of two unsigned
numbers.
U. (n ...) l 0 ( [U-d t]
Transmits the 16-bit value n to the output device. n is
represented as an unsigned integer in the current numeric
conversion base. A trailing space is printed.
Leaves the unsigned remainder u2 and unsigned quotient u3 from the
division of the unsigned double dividend ud by the unsigned
divisor ul.
u< (un1\un2   f) 2 1
Unsigned comparision. Leaves a true flag if unl is less than un2,
otherwise leave a false flag. For correct operation the difference
between unl and un2 should not exceed 32767.
UNTIL (f ...) 1 0 p,c
Used in a colon definition in the form
BEGIN ... UNTIL
If f is false execution branches back to the corresponding BEGIN .
If f is true execution continues with the next word after UNTIL .
usmz (n ...) 1 0
A defining word used in the form
n USER cccc
to create a user variable cccc . Execution of cccc leaves the
address, in the user area, of the value of cccc . The value of n
is the offset from the start of the user variable area to the
memory location (2 bytes) in which the value is stored. The value
is not initialised.
113


VARIABLE (n ...) 1 0
A defining word used in the form
n VARIABLE cccc
to create a variable cccc with initial value n. Execution of cccc
leaves the address, in the parameter area of cccc , containing the
value of cccc .
VLIST [V-list]
Display, on the output device, a list of the names of all words in
the CONTEXT vocabulary and any other vocabulary from which the
CONTEXT vocabulary is chained. All VLIST's will therefore include
a listing of words in the FORTH vocabulary. The listing be be
interrupted by pressing the ESC key and resumed by pressing the
space bar. If, after interruption, The ESC key (or any other key
except the space bar) is pressed, the listing will be aborted.
VOC-LINK (... addr) 0 l
A user variable containing the address of a vocabulary link field
in the word which defines the most recently created vocabulary.
All vocabularies are linked through these fields in their defining
words.
VOCABULARY E
A defining word used in the form
VOCABULARY cccc
to create a defining word for a vocabulary with name cccc .
Execution of cccc makes it the CONTEXT vocabulary in which a
dictionary search will start. Execution of the sequence:
cccc DEFINITIONS
will make cccc the CURRENT vocabulary into which new definitions
are placed. Vocabulary cccc is so linked that a dictionary search
will also find all words in the vocabulary in which cccc was
originally defined. All vocabularies, therefore, ultimately link
to FORTH .
By convention all vocabulary defining words are declared
IMMEDIATE .
WARNING (... addr) 0 l
A user variable whose value determines the action on detection of
an error. If WARNING contains -l an error causes a system ABORT
which may, by changing a pointer in (ABORT) , be altered to a
user—defined response. If WARNING contains zero then an error
message number is given. In the system provided, WARNING is set to
zero on initialisation . See ERROR .
WHILE (f ...) l 0 P,C
Used in a colon definition in the form
BEGIN ... WHILE ... REPEAT
WHILE tests the top value on the stack. If it is true execution
continues to REPEAT which forces a branch back to BEGIN . If f is
false execution skips to the first word after REPEAT . See BEGIN .
ll4


WIDTH (... addr) 0 1
A user variable containing the maximum number of letters saved
during the compilation of a definition's name. It must be a value
between l and 3l inclusive and has a default value of 3l. The
value may be changed at any time provided it is kept within the
above limits.
wom) (c ...) 1 0
Accepts text characters from the input buffer (terminal or tape)
until a delimiter character c is found. The string starting with a
length count byte, is then placed in the WORD buffer starting at
HERE and two or more banks are added to the end. The choice of
input buffer is determined by BLK . See BLK , IN .
X
This is a pseudonym for the dictionary entry whose name is one
character of ASCII null (00). It is the procedure to terminate
interpretathmm of text from the input buffer, since both input
buffers have at least one null character at the end.
XOR (nl\n2 ... xor) 2 l
Leaves the bit-by—bit logical exclusive—OR of nl and n2.
[ P [left—bracket]
Used in the creation of a colon definition in the form
: nnnn ... [ ... ] ... ;
to suspend compilation of the definition and allow words to
execute . See ] .
[COMPILE] P,C [bracket—compi1e]
Used in the creation of a colon definition to force the
compilation of an IMMEDIATE word which would otherwise execute.
The most frequent use is with vocabulary words e.g.
[COMPILE] FORTH
to delay the change of the CONTEXT vocabulary to FORTH until the
word containing the above sequence executes.
] [right bracket]
Used during the creation of a colon definition, to resume
compilation after the suspension of compilation by [ .
` ll5


116


’ _ O O
Two s - Complement Ar1thmet1c
In unsigned arithmetic using 16-bit numbers, the lowest value that can
be represented is zero, appearing as binary notation as
O 0 O 0 0 0 0 0 0 0 O 0 0 O 0 0 ,
and the highest number appears as
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
which represents the decimal value 65535. There are therefore,
including zero, 65536 different numbers.
To understand the operations on signed numbers, consider what
happens if one is added to the highest unsigned value, 65535. In
binary notation this sum appears as
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
+ 0 0 O O 0 0 0 0 0 0 0 O O 0 0 1
(l) 0 0 O 0 0 O 0 0 0 O 0 0 0 0 O 0
In a computer, working to 16-bit accuracy, the one in the 17th place
is lost and the value stored as the result will be zero. If we add one
to a number and find the result is zero, it is natural to interpret
the original number as having a value of -1.
Thus, for signed arithmetic, the number
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
can be used to represent -1.
In general the number —x is represented by the value which gives a
zero result when +x is added to it (ignoring any overflow into the
17th place). The signed values -2 , -23 and -32768 are therefore
represented by
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 O
1 1 1 1 1 1 1 1 1 1 1 0 1 0 0 1
and 1 0 0 O 0 0 0 O 0 0 0 0 0 0 0 0 respectively.
All negative values are represented by binary numbers whose most
significant (16th) bit is an one. Accordingly, the highest positive
number that can be represented is:
0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
or +32767, and the most negative number is -32768, shown earlier.
The range for a signed number is thus from -32768 to +32767 which,
including zero, gives a total of 65536 different values (as for
unsigned numbers).
Whether a number is interpreted as a signed or an unsigned value
is entirely a matter of context; the binary number
1 1 l 1 1 1 1 1 1 1 1 1 1 1 1 1
117


may represent either +65535 or -l depending on the conversion routine
used.
The above discussion has been confined to 16-bit numbers but
similar considerations apply to any precision. In all cases the most
significant bit of the number will be zero for a positive value and
one for a negative value. It may, therefore, be regarded as a 'sign
bit'.
In general the binary representation of a negative number may be
found by writing down the binary representation of the corresponding
positive number, inverting all the bits and adding one. This is shown
in the following example to find the two's-complement representation
of -4 (in 8-bit precision):
O 0 0 O 0 I O 0 (+4)
invert all bits (form the one's-complement):
l l l l l 0 l l
add l (form the two's-complement):
11111100 (-4)
ll8


FORTH for the Acorn Systems 3 and 4 is almost identical to ATOM
FORTH, except for the changes in the memory map. FORGET is changed
to work in the range #3C00 and #8000 and no memory is needed above
#8000.
Pointer Contents Address
/ / / / / / / / / / / / / 6000
USER VARIABLES
UP , LIMIT --—- > 5FC4
DISC/TAPE BUFFER
FIRST -———-—-- 9 5DC0
TEXT BUFFER
PAD ----——-— > -—-——--—--——--
NUMERIC CONVERSION BUFFER
WORD BUFFER
DP -—------ > — + —---——-——---
APPLICATIONS DICTIONARY
DP0 ———-———- 9 3C00
NUCLEUS DICTIONARY
( As for ATOM FORTH )
0000
The system can be modified to use memory up to #8000 by changing
UP, LIMIT and FIRST, for example:
HEX
LIMIT 2000 + ' LIMIT I
FIRST 2000 + ' FIRST 1
l0 +ORIGIN @ 2000 + 10 +ORIGIN ! ( change UP )
COLD ( initialise )
The system should then be resaved:
SAVE FORTHA 240 400 2EB
SAVE FORTHB 2800 3C00 2EB


120


All dictionary entries in FORTH have the same general form:
NFA Name length (l byte), Name
Characters of the name field
(up to 31 bytes)
LFA Link pointer to Link
previous NFA (2 bytes) field
CFA Pointer to machine Code
code to execute (2 bytes) field
PFA Parameter
field
The name length byte contains, in its least significant five bits, the
number of characters in the name of the word (maximum 3l characters).
The sixth bit is the 'smudge' bit which, when set to l, will prevent
the dictionary entry from being found on a dictionary search (except
by VLIST ). This is mainly used to prevent the finding and use of a
partly—completed dictionary entry. The seventh bit is known as the
precedence bit and marks a word as being IMMEDIATE when set to l. The
eighth or most significant (sign) bit is always set to l, as is the
sign bit of the last character of the name. This is to allow the
operation of TRAVERSE , which will move, in either direction, across
the name field of the word.
The link field contains the address of the start of the name field
of the preceding dictionary entry to allow a dictionary search to be
made. A link field containing zero marks the end of the dictionary.
The various types of dictionary entry differ only in the contents
of their code fields and parameter fields. The code field always
contains a pointer to the start of an executable machine code routine,
and the different possibilities are given in the following list.
a) Machine-code primitives
The machine code is placed in the parameter field of
the entry and the code field points to its start.
b) Constants
The value of the constant is contained in a two-byte
parameter field and the code field points to a
machine-code routine to place the contents of the
parameter field on the stack.
c) Variables
The value of the variable is contained in a two-byte
parameter field and the code field contains a pointer
to a machine-code routine which places the address of
the parameter area on the stack.
l2l


d) User Variables
The offset from the start of the user variable area to
the address where the variable is stored is contained in a
one-byte parameter area. The code field points to a
machine-code routine which adds the offset to the address
of the start of the user variable area and places the result
on the stack.
e) Colon definitions
The code field contains a pointer to a machine-code
routine which interprets the contents of the parameter
field as a list of addresses of other FORTH words to be
executed.
f) Words constructed using 
The first two bytes of the parameter area contain the
address of the words following DOES> in the creating word.
The remainder of the parameter area contains a series of
values placed there by the words (if any) following
 ).
Saving Dictionary Space
The maximum length of the name of a dictionary entry is contained in
the user variable WIDTH . This may at any time be reduced from its
default value of 3l characters, with a consequent saving of space in
the name field of a dictionary entry. If, for example, the value of
WIDTH is reduced to 3 by
3 WIDTH I
then any new dictionary entry will be with the actual length of its
name, but only the first three characters saved. All words must then
be uniquely determined by their length and their first three
characters (i.e. LOOK and LOOP will not be distinguished). The use of
a value of WIDTH less than 3 is not recommended but, as this
demonstrates,
IT IS VER- EAS- TO REA- FOR-- IF YOU ONL— HAVE- THE FIR-- THR--
LET ---- AND THE LEN--- OF THE WOR-
The value of WIDTH may be increased or decreased at any time (subject
to its remaining in the range 3 to 3l inclusive) and will not have any
effect on previously defined words.
l22


Headerless Code
Some dictionary entries in ATOM FORTH are 'headerless'. This means
that their heads do not include name and link fields. They cannot,
therefore, be found by a dictionary search, or included in a new
definition unless their code field (execution) addresses are known to
the programmer. The glossary gives this address for each headerless
entry.
Creating headerless code is a very efficient way of saving memory,
but it does reduce the flexibility of the system since headerless
entries are relatively difficult to use.
The main use of headerless code is in producing a stand—alone
system whose action is fixed, such as a dedicated control system.
123


124


Appe11d1x D
Memory A11oc:et1on
Pointer Contents Address
/ / / / / / / / / / / / /
9800
USER VARIABLES
UP , LIMIT —--- > 97C4
TAPE I/O BUFFER
FIRST —-—--——— > 95C0
- _ _ —TEXT BUFFER- - - - _
PAD ———-——-— > --————---——-—-
NUMERIC CONVERSION BUFFER
WORD BUFFER
DP ———--——— 9 -—---—--—-----
APPLICATIONS DICTIONARY
DP0 —-——-—-- 9 8C00
GRAPHICS MODE 3
---—-—-—--———— 8600
GRAPHICS MODE 2
-————-————-—-— 8400
GRAPHICS MODE 1
-—-———-——--—-- 8200
VDU/GRAPHICS MODE 0
8000
/ / / / / /
/ / / / / / / / / / / / /
/ / / / / / / / / / / / / / 3000
NUCLEUS DICTIONARY
`§56§ZE`i§§§§§£§`
ORIGIN ————-——- > 2800
/ / / / / / / / / / / / /
/ / / / / / / / / / / / / 0400
BLOCK ZERO
0000
125


Block zero Memory Map
Pointer Contents Address
////////////// 0400
GRAPHICS PLOT VECTOR
03FE
LOWER DICTIONARY AREA
0240
OPERATING SYSTEM
VECTORS
0200
R0 -——--——- 9
RETURN STACK
RP —-————-— 9
01A4
TERMINAL INPUT BUFFER A
TIB -—--———- 9 0150
FREE (ECONET)
0140
COS\DOS INPUT BUFFER
0100
TOPO]?  
Zero Page RESERVED FOR COS\DOS
0098
FORTH SCRATCHPAD
AND POINTERS
0087
0086
FREE
006C
006B
TAPE INTERFACE
WORKSPACE
0062
GRAPHICS WORKSPACE
005A
S0 —--—-——— 9
COMPUTATION
STACK
SP ————-——— 9
0000
126


Relocation of the Applications Dictionary
The applications dictionary, tape buffer and user variable area
normally reside in memory between #8200 and #97FF. All addresses in
this range are, if treated as signed single-precision integers,
'negative'. This has two important consequences for relocation of the
applications dictionary in the address range below #8000:
l. Before compiling a dictionary entry, FORTH checks whether there is
sufficient room between the top of the dictionary (contents of DP)
and the start of the tape buffer (value of FIRST). A 'positive'
value in DP and a 'negative' value of FIRST will cause the test to
fail and give error message 2 (dictionary full). In order to avoid
extensive changes to the system, relocation of the applications
dictionary below #8000 should be accompanied by a corresponding
relocation of the tape buffer.
2. The word FORGET performs an address validation check before
allowing part of the dictionary to be discarded. This check assumes
that valid addresses are 'negative', and they will therefore fail
in the relocated applications dictionary. The definition of FORGET
must be modified for the relocated system.
The modifications given here also relocate the user variables.
This is not strictly necessary but will leave the upper RAM completely
free so that mode 4 graphics can be used. The changes are made
permanent so that the new system can be saved on tape. The areas to be
saved are #240 to #400 and #2800 to #3C00, both with an execution
address of #2800. A minimwm of 2K of RAM is required in the region
between #3C00 and #7FFF inclusive. In the code, XXXX represents the
RAM start address and YYYY the end address + l.
COLD ( start on an empty applications dictionary )
HEX
XXXX DUP DP 1 ( relocate applications dictionary )
DUP lE +ORIGIN 1 ( change DP boot-up parameter )
DUP lC +ORIGIN 1 FENCE 1 ( FENCE and its boot—up parameter )
YYYY 3C - ( start of new user variable area )
DUP l0 +ORIGIN 1 ( change user variable boot—up parameter )
DUP ' LIMIT 1 ( relocate end of tape buffer )
204 — ' FIRST 1 ( and the beginning )
: TEMP ( a temporary definition of the new FORGET )
CURRENT @ CONTEXT @
— l8 ?ERROR ( CONTEXT = CURRENT? )
[COMPILE] ·
DUP FENcE @
[ 2822 . ] ( top of nucleus )
MAX < l5 ?ERROR ( below FENCE or in nucleus? )
DUP NFA DP 1
LFA @ CURRENT @ 1 ;
If this definition is compiled successfully, all is well so far.
' TEMP HERE OVER — ( PFA and parameter field length )
' FORGET SWAP CMOVE ( overwrite FORGET )
( the new version is shorter, so OK )
l27


FORGET TEMP
( If TEMP is forgotten successfully, everything is OK )
COLD ( Finally, check out modified boot—up parameters )
128


1. The FORTH Interest Group in the U.S.A. supply many documents
relating to FORTH, including assembly listings for· many different
microprocessors, a language model, reprints of "BYTE" magazine
articles and a bi-monthly magazine entitled "FORTH Dimensions".
For details of current costs for membership and their publications
write (with an SAE please) to:
FORTH Interest Group,
P.O. Box 1105,
San Carlos,
Ca. 94070.
2. The FORTH Interest Group U.K. is the British branch of the U.S.A.
group. At present it meets on the first Thursday of every month at 7
p.m. at the Polytechnic of the South Bank in London. Membership
includes a bi-monthly newsletter entitled “FORTHWRITE". Like its
parent group, F.I.G. U.K. exists to promote interest in and the use of
the FORTH language and its members are prepared to help with any
difficulties that may be encountered. For further details contact
(S.A.E. please):
The Honorary Secretary,
F.I.G. U.K.,
15, St. Albans Mansions,
Kensington Court Palace,
London W8 5QH
3. A number of articles on FORTH have appeared in "BYTE" magazine:
August, 1980 A FORTH language 'special'
February, 1981 Stacking Strings in FORTH
March, 1981 A Coding Sheet for FORTH
The August 1980 issue is now unobtainable, but reprints of a11 the
BYTE articles are available from the U.S.A. FORTH Interest Group (Ref.
1).
4. "FORTH for Microcomputers" Dr. Dobb's Journal No. 25 (May 1978).
A brief review of the external and internal workings, with a variety
of examples.
129


5. "Starting FORTH" L. Brodie.
published by Prentice—Ha11 (Nov. 1981).
Available from: Computer Solutions Ltd.,
Treway House,
Hanworth Lane,
Chertsey.
Tel: Chertsey (09328) 65292
The author is from FORTH Inc., the company started by Charles Moore,
the inventor of FORTH. This is a very good introduction to the
language, with lots of examples.
6. "Threaded Interpretive Languages" R. G. Loeliger,
published by Byte Books (McGraw-Hill) (1981).
A good, clear, description of the internal workings of FORTH-like
languages, based on an implementation for the Z—80 microprocessor. Not
for the complete beginner, but try it in a couple of month's time!
130


I (FORTH word) 21, 92
1CSP (FORTH word) 92
“ character 96
£ (FORTH word) 53, 92
£> (FORTH word) 53, 92
ES (FORTH word) 53, 92
$1 examp-e 74
$@ example 74
$IN example 51, 73, 74
' (FORTH word) 92
( (FORTH word) 60, 93
(+LOOP) (FORTH word) 93
(.") (FORTH word) 93
(;CODE) (FORTH word) 93
(ABORT) (FORTH word) 46, 93
(DO) (FORTH word) 93
(ENTER) (FORTH word) 93
(FIND) (FORTH word) 93
(LINE) (graphics word) 67
(LOAD) (FORTH word) 94
(LOOP) (FORTH word) 94
(NUMBER) (FORTH word) 46, 47, 94
(PLOT) (graphics word) 66
) character 60, 93
* (FORTH word) 12, 94
*/ (FORTH word) 12, 94
*/MOD (FORTH word) 12, 94
+ (FORTH word) 12, 94
+1 (FORTH word) 22, 95
+— (FORTH word) 12, 95
+LOOP (FORTH word) 39, 95
+ORIGIN (FORTH word) 95
, (FORTH word) 24, 95
— (FORTH word) 12, 95
—-> (FORTH word) 60, 95
-DUP (FORTH word) 13, 37, 95
—FIND (FORTH word) 95
—TRAILING (FORTH word) 45, 51, 96
. (FORTH word) ll, 45, 51, 96
." (FORTH word) 45, 96
.POUNDS example 53
.R (FORTH word) 51, 96
.REAL example 54


.S example 88
.TIME example 84
/ (FORTH word) l2, 96
/MOD (FORTH word) 12, 96
0 (FORTH word) 96
0< (FORTH word) 15, 96
0= (FORTH word) 15, 96
OBRANCH (FORTH word) 96
l (FORTH word) 96
1+ (FORTH word) 12, 96
2 (FORTH word) 96
2* (FORTH word) l2, 97
2+ (FORTH word) 12, 97
2/ example 77
ZARRAY example 72
2DROP (FORTH word) 17, 97
2DUP (FORTH word) 17, 97
3—COUNT example 39
: (FORTH word) 20, 97
; (FORTH word) 20, 97
;S (FORTH word) 61, 97
< (FORTH word) 15, 97
<£ (FORTH wotd) 53, 97
 (FORTH word) 15, 98
>R (FORTH word) 13, 98
? (FORTH word) 22, 98
?COMP (FORTH word) 98
?CSP (FORTH word) 98
?ERROR (FORTH word) 98
?ESC (FORTH word) 42, 99
?EXEC (FORTH word) 99
?LOADING (FORTH word) 99
?PAIRS (FORTH word) 99
?STACK (FORTH word) 99
?TERMINAL (FORTH word) 99
@ (FORTH word) 22, 99
ABORT (FORTH word) 99


ABS (FORTH word) 12, 99
AGAIN (FORTH word) 42, 99
allocation memory 125
ALLOT (FORTH word) 70, 100
AND (FORTH word) 15, 100
animated graphics 66
applications separating 21
arithmetic 10
doub1e—precision 16
single-precision 11
two's—comp1ement 117
arrays 71
one—dimensiona1 71
two-dimensional 72
with index check 71
B (editor command) 63
BACK (FORTH word) 100
BACKWARDS example 39
BASE (FORTH word) 23, 49, 100
base BASE—36 50
BINARY 50
conversion example 50
BASE-36 base 50
bases numeric 49
BEGIN (FORTH word) 42, 100
BINARY base 50
BL (FORTH word) 49, 100
BLACK (graphics word) 66
BLANKS (FORTH word) 49, 100
BLK (FORTH word) 23, 100
blocks of memory manipulating 48
BRANCH (FORTH word) 100
branches 35
conditional 35
nested 37
BREAK key 7
C (editor command) 62
C! (FORTH word) 101
C, (FORTH word) 24, 101
C/L (FORTH word) 73, 101
C@ (FORTH word) 101
CASE example 74, 75
CFA (FORTH word) 20, 101
character input 45
output 50
CLEAR (FORTH word) 65
clear graphics 67
screen 67
CLIT (FORTH word) 101
CMOVE (FORTH word) 48, 101
intelligent 49
code field address 19
code headerless 123
coding example 4
COLD (FORTH word) 8, 101
· cold start 8
colon-definitions 20
compilation of 30


form of 20
comments 60
compilation of co1on—definitions
30
IMMEDIATE words 31
numbers 32
COMPILE (FORTH word) 31, 101
example 32
compiler security 31
computation stack 10
conditional branches 35
CONSTANT (FORTH word) 21, 70, 102
CONTEXT (FORTH word) 23, 28, 102
conversion numeric 47
copying screens 79
COUNT (FORTH word) 45, 51, 102
COUNTER example 38
COUNTS example 38
cover figure 86
CR (FORTH word) 102
CREATE (FORTH word) 24, 102
+ example 27
AND example 27
RP@ example 28
CSP (FORTH word) 23, 102
CTABLE example 73
CURRENT (FORTH word) 23, 28, 102
cursor editing 62
CVARIABLE example 71, 71
D (editor command) 61
D+ (FORTH word) 16, 102
D+- (FORTH word) 17, 103
D—>H example 50
D. (FORTH word) 51, 103
D.R (FORTH word) 52, 103
DABS (FORTH word) 17, 103
DATA example 73
DECIMAL (FORTH word) 50, 103
decimal point 16
definite loops 37
definitions 19
DEFINITIONS (FORTH word) 29, 103
DELAYS example 38
DELETE (editor command) 63
deleting lines 61
DELTAX (graphics word) 67
DELTAY (graphics word) 67
demonstration graphics 86
DIAGONAL example 66
dictionary entries 19, 121
entry types 121
relocation 127
dictionary space extra 65
saving 122
DIGIT (FORTH word) 103
discs 64
DLITERAL (FORTH word) 32, 104
DMINUS (FORTH word) 17, 104
DO (FORTH word) 37, 104


DO—IT-LATER example 31
DO-IT-NOW example 30
double—precision arithmetic 16
numbers 16
operators 16
DP (FORTH word) 23, 104
DPL (FORTH word) 16, 23, 46, 47,
104
DROP (FORTH word) 13, 104
DUMP example 52
DUP (FORTH word) 13, 104
E (editor command) 61
econet 64
editing cursor 62
example 58, 59, 63
lines 61
strings 62
editor 58
editor commands:
B 63
C 62
D 61
DELETE 63
E 61
F 62
H 61
I 61
M 63
MATCH 63
N 63
P 60, 61
R 61
S 61
T 61
TILL 62
WHERE 57
X 62
editor loading 55
EDITOR vocabulary 58 ‘
ELSE (FORTH word) 105
EMIT (FORTH word) 50, 105
EMPTY—BUFFERS (FORTH word) 58
ENCLOSE (FORTH word) 105
ERASE (FORTH word) 49, 105
erasing lines 61
ERR (graphics word) 67
ERROR (FORTH word) 105
error message from arrays 72
messages 91
errors 91
examples:
$174
$@ 74
$IN 51, 73, 74
.POUNDS 53
.REAL`54
.S 88
.TIME 84
2/ 77


ZARRAY 72
3-COUNT 39
BACKWARDS 39
base conversion 50
CASE 74, 75
coding 4
COMPILE 32
COUNTER 38
COUNTS 38
CREATE + 27
CREATE AND 27
CREATE RP@ 28
CTABLE 73
CVARIABLE 71, 71
D->H 50
DATA 73
DELAYS 38
DIAGONAL 66
DO—IT-LATER 31
DO—IT-NOW 30
DUMP 52
editing 58, 59, 63
FACT 77, 78
factorials 77
FAMILY 69
fast 2 / 77
FLASH 66
GCD 43
HELL—FREEZES—OVER 43
IMMEDIATE 30, 32
INPUT 47
INSCR 80
INVERT editing 62
JTEST 41
KEYBOARD 83
LEFT$ 51
LIST 73
LOOK—UP 40, 41
MD* 17
MEMBER 69
MONITOR 81
NUMIN 46
OUTSCR 80
PAUSE 42
PLOT-IT 86
quadratic 12, 14
QUICKSORT 78
RECTANGLE 72
recursion 77
RIGHTS 51
RND 59
SEQUENCE 39
SHOWASCII 45
SIZE 70
sort 78
STARS 7
STRING 73
STRINGS 51
TENCOUNT 37
tones 82


TRIANGLE 67
VALUES 71
WAIT 81
WASHING 4
WORDS 73
[COMPILE] 31, 32
EXECUTE (FORTH word) 105
execution address 19
execution time of words 85
EXPECT (FORTH word) 106
extension memory 67, 127
extra dictionary space 65
F (editor command) 62
FACT example 77, 78
factorials example 77
FAMILY example 69
fast 2 / example 77
FENCE (FORTH word) 23, 106
FILL (FORTH word) 49, 106
FIRST (FORTH word) 106
FLASH example 66
FORGET (FORTH word) 21, 28, 106
form of colon-definitions 20
FORTH (FORTH word) 29, 106
FORTH words:
1 21, 92
ECSP 92
E 53, 92
E> 53, 92
ES 53, 92
' 92
( 60, 93
(+LOOP) 93
(.") 93
(;CODE) 93
(ABORT) 46, 93
(DO) 93
(ENTER) 93
(FIND) 93
(LOAD) 94
(LOOP) 94
(NUMBER) 46, 47, 94
* 12, 94
*/ 12, 94
*/MOD 12, 94
+ 12, 94
+1 22, 95
+- 12, 95
+LOOP 39, 95
+ORIGIN 95
, 24, 95
— 12, 95
——> 60, 95
—DUP 13, 37, 95
—FIND 95
—TRAILING 45, 51, 96
. 11, 45, 51, 96
." 45, 96
.R 51, 96


/ 12, 96
/MOD 12, 96
0 96
0< 15, 96
0= 15, 96
OBRANCH 96
1 96
1+ 12, 96
2 96
2* 12, 97
2+ 12, 97
2DROP 17, 97
2DUP 17, 97
: 20, 97
; 20, 97
;S 61, 97
< 15, 97
<£ 53, 97
 15, 98
>R 13, 98
? 22, 98
?COMP 98
?CSP 98
? 98
?ESC 42, 99
?EXEC 99
?LOADING 99
?PAIRS 99
?STACK 99
?TERMINAL 99
@ 22, 99
ABORT 99
ABS 12, 99
AGAIN 42, 99
ALLOT 70, 100
AND 15, 100
BACK 100
BASE 23, 49, 100
BEGIN 42, 100
BL 49, 100
BLANKS 49, 100
BLK 23, 100
BRANCH 100
C! 101
C, 24, 101
C/L 73, 101
C@ 101
CFA 20, 101
CLEAR 65
CLIT 101
CMOVE 48, 101
COLD 8, 101
COMPILE 31, 101
CONSTANT 21, 70, 102
CONTEXT 23, 28, 102
COUNT 45, 51, 102


CR 102
CREATE 24, 102
CURRENT 23, 28, 102
D+ 16, 102
D+- 17, 103
D. 51, 103
D.R 52, 103
DABS 17, 103
DECIMAL 50, 103
DEFINITIONS 29, 103
DIGIT 103
DLITERAL 32, 104
DMINUS 17, 104
DO 37, 104
DOES> 69, 70, 104
DP 23, 104
DPL 16, 23, 46, 47, 104
DROP 13, 104
DUP 13, 104
editor commands 62
ELSE 105
EMIT 50, 105
EMPTY-BUFFERS 58
ENCLOSE 105
ERASE 49, 105
ERROR 105
EXECUTE 105
execution time of 85
EXPECT 106
FENCE 23, 106
FILL 49, 106
FIRST 106
FORGET 21, 28, 106
FORTH 29, 106
HERE 45, 51, 106
HEX 50, 106
HLD 23, 106
HOLD 53, 107
I 38, 107
ID. 107
IF 35, 107
IMMEDIATE 30, 107
IN 23, 107
in text 1
INTERPRET 107
J 41
K 42
KEY 107
LATEST 77, 107
LEAVE 40, 108
LFA 20, 108
LIMIT 108
LIST 56
LIT 108
LITERAL 32, 108
LOAD 56, 108
LOOP 37, 108
M* 16, 108
M/ 17, 109
M/MOD 17, 109


MAX 15, 109
MESSAGE 109
MIN 15, 109
MINUS 12, 109
MOD 12, 109
names of 89, 122
NFA 20, 109
NOOP 109
NUCTOP 109
NUMBER 46, 109
OR 15, 109
OSDATA 110
OSLOAD 110
OUT 23, 110
OVER 13, 110
PAD 54, 61, 110
PFA 20, 110
PICK 13
PROGRAM 58
pronunciation of 89
QUERY 45, 110
QUIT 110
R 13, 110
RE 23, 111
R0 23, 110
R> 13, 111
REPEAT 42, 43, 111
ROLL 13, 111
ROT 13, 111
RP! 111
RP@ 111
S->D 111
S0 23, 111
SCR 23, 57, 111
SIGN 53, 112
SMUDGE 16, 24, 112
A SP! 112
SP@ 112
SPACE 112
SPACES 112
stack actions of 89
STATE 23, 112
status of 89
SWAP 13, 112
THEN 35, 112
TIB 23, 112
TOGGLE 15, 113
TRAVERSE 113
TYPE 45, 50, 51, 113
U* 16, 113
U. 11, 52, 113
U/ 17, 113
U< 15, 113
UNTIL 42, 113·
USER 22, 113
validity of 4
VARIABLE 22, 114
VLIST 7, 114
VOC—LINK 23, 114
VOCABULARY 19, 28, 114


WARM 8
WARNING 23, 46, 114
WIDTH 23, 115, 122
WORD 45, 115
X 115
XOR 15, 115
[ 30, 115
[COMPILE] 30, 115
] 30, 115
further reading 129
GCD example 43
glossary of FORTH words 89
GOTO 35
graphics 65, 86
animated 66
clear 67
demonstration 86
line-drawing 66
memory 65
mode 4 67
modes 65
package 65
point—plotting 66
relative plotting 67
resolution 65
GRAPHICS vocabulary 66
graphics words:
(LINE) 67
(PLOT) 66
BLACK 66
DELTAX 67
DELTAY 67
ERR 67
INVERT 66
LINE 67
MOVE 67
PLOT 66
REL 67
RLINE 67
RMOVE 67
RPLOT 67
SETXY 67
WHITE 66
XDIR 67
YDIR 67
H (editor command) 61
headerless code 123
HELL—FREEZES—OVER example 43
HERE (FORTH word) 45, 51, 106
HEX (FORTH word) 50, 106
HLD (FORTH word) 23, 106
HOLD (FORTH word) 53, 107
I (editor command) 61
(FORTH word) 38, 107
ID. (FORTH word) 107
IF (FORTH word) 35, 107
IMMEDIATE (FORTH word) 30, 107


example 30, 32
words compilation of 31
IN (FORTH word) 23, 107
indefinite loops 42
index check for arrays 71
index to manual 113
indirect threaded code 4
input 45
character 45
INPUT example 47
( input numeric 46
text 45
INSCR example 80
integers printing 11
intelligent CMOVE 49
INTERPRET (FORTH word) 107
introduction to FORTH 2
to manual 1
INVERT (graphics word) 66
INVERT editing example 62
IP location 25, 26
J (FORTH word) 41
JTEST example 41
K (FORTH word) 42
KEY (FORTH word) 107
KEYBOARD example 83
LATEST (FORTH word) 77, 107
LEAVE (FORTH word) 40, 108
LEFT$ example 51
LFA (FORTH word) 20, 108
LIMIT (FORTH word) 108
LINE (graphics word) 67
line-drawing graphics 66
lines deleting 61
editing 61
erasing 61
replacing 61
link field address 19
LIST (FORTH word) 56
example 73
LIT (FORTH word) 108
LITERAL (FORTH word) 32, 108
LOAD (FORTH word) 56, 108
loading editor 55
FORTH 7
screens 56
tape interface 55
locations:
IP 25, 26
N 25
UP 25
W 25, 26
XSAVE 25, 25
zero-page 25
logical operators 15
LOOK—UP example 40, 41
LOOP (FORTH word) 37, 108


loop index 37, 38
limit 37
loops indefinite 42
nested 39
loudspeaker 82
M (editor command) 63
M* (FORTH word) 16, 108
M/ (FORTH word) 17, 109
M/MOD (FORTH word) 17, 109
machine-code 24
manipulating blocks of memory 48
MATCH (editor command) 63
MAX (FORTH word) 15, 109
MD* example 17
MEMBER example 69
memory allocation 125
extension 67, 127
graphics 65
memory map ATOM FORTH 125, 126
System FORTH 119
memory usage 65
MESSAGE (FORTH word) 109
meta—FORTH 69
MIN (FORTH word) 15, 109
MINUS (FORTH word) 12, 109
mixed—precision operators 16, 17
MOD (FORTH word) 12, 109
mode 4 graphics 67
modes graphics 65
monitor commands 81
MONITOR example 81
MOVE (graphics word) 67
music 83
N (editor command) 63
location 25
name field address 19
names of FORTH words 89, 122 1
nested branches 37
loops 39
NFA (FORTH word) 20, 109
NOOP (FORTH word) 109
notation postfix 11
reverse-Polish ll
NUCTOP (FORTH word) 109
NUMBER (FORTH word) 46, 109
numbers compilation of 32
doub1e—precision 16
single—precision 11
numeric bases 49
conversion 47
input 46
output 51
output formatting 53
NUMIN example 46
OCTAL 50
one—dimensional arrays 71
operators:


doub1e—precision 16
logical 15
mixed-precision 16, 17
relational 15
single—precision 12
stack 13, 17
OR (FORTH word) 15, 109
OS commands 81
OSDATA (FORTH word) 110
OSLOAD (FORTH word) 110
OUT (FORTH word) 23, 110
output 50
character 50
formatting numeric 53
numeric 51
text 50
OUTSCR example 80
OVER (FORTH word) 13, 110
P (editor command) 60, 61
PAD (FORTH word) 54, 61, 110
parameter field 19
stack 10
PAUSE example 42
PFA (FORTH word) 20, 110
PICK (FORTH word) 13
PLOT (graphics word) 66
PLOT-IT example 86
point-plotting graphics 66
POPTWO 24
postfix notation 11
printing integers ll
PROGRAM (FORTH word) 58
pronunciation of FORTH words 89
quadratic example 12, 14
QUERY (FORTH word) 45, 110
QUICKSORT example 78
QUIT (FORTH word) 110
R (editor command) 61
(FORTH word) 13, 110
RE (FORTH word) 23, 111
RO (FORTH word) 23, 110
R> (FORTH word) 13, 111
reading further 129
RECTANGLE example 72
recursion 77
example 77
references 129
REL (graphics word) 67
relational operators 15
relative plotting graphics 67
relocation dictionary 127
tape buffer 127
user variable area 127
REPEAT (FORTH word) 42, 43, 111
replacing lines 61
resolution graphics 65
return stack 10, 13


reverse-Polish notation ll
RIGHTS example 51
RMOVE (graphics word) 67
RND example 59
ROLL (FORTH word) 13, lll
ROT (FORTH word) 13, 111
RP! (FORTH word) lll
RP@ (FORTH word) 111
RPLOT (graphics word) 67
S (editor command) 61
S—>D (FORTH word) lll
S0 (FORTH word) 23, lll
SAVE (tape interface word) 57
saving dictionary space 122
SCR (FORTH word) 23, 57, lll
screen clear 67
copying utility 79
numbering 63
screens 55, 58
loading 56
security compiler 31
separating applications 21
SEQUENCE example 39
SETXY (graphics word) 67
SHOWASCII example 45
SIGN (FORTH word) 53, 112
single-precision arithmetic 11
numbers 11
operators 12
SIZE example 70
SMUDGE (FORTH word) 16, 24, 112
sort example 78
sound 82
SP! (FORTH word) 112
SP@ (FORTH word) 112
SPACE (FORTH word) 112
SPACES (FORTH word) 112
stack actions of FORTH words 89
stack computation 10
operators 13, 17
overflow 38
parameter 10,
print 88
return 10, 13
transfers 13
stacks 9
STARS example 7
STATE (FORTH word) 23, 112
status of FORTH words 89
storage of strings 50, 51
STRING example 73
string handling 51
strings 73
editing 62
STRINGS example 51
strings storage of 50, 51
SWAP (FORTH word) 13, 112
System FORTH 119


T (editor command) 61
tables 72
tape buffer relocation 127
tape interface 56
loading 55
TENCOUNT example 37
terminating jumps 25
text FORTH words in 1
input 45
input delimiter 46
output 50
THEN (FORTH word) 35, 112
threaded code 4 4
TIB (FORTH word) 23, 112
TILL (editor command) 62
timer 84
TOGGLE (FORTH word) 15, 113
tones example 82
TRAVERSE (FORTH word) 113
TRIANGLE example 67
two's—comp1ement arithmetic 117
two-dimensional arrays 72
TYPE (FORTH word) 45, 50, 51, 113
U* (FORTH word) 16, 113
U. (FORTH word) 11, 52, 113
U/ (FORTH word) 17, 113
U< (FORTH word) 15, 113
UNTIL (FORTH word) 42, 113
UP location 25
USER (FORTH word) 22, 113
user variable area relocation 127
user variables 23
utilities:
MONITOR 81
screen copying 79
validity of FORTH words 4
VALUES example 71
VARIABLE (FORTH word) 22, 114
example 70
VIA timer 84
VLIST (FORTH word) 7, 114
VOC—LINK (FORTH word) 23, 114
vocabularies:
EDITOR 58
FORTH 29
GRAPHICS 66
VOCABULARY (FORTH word) 19, 28,
114
W location 25, 26
WAIT example 81
WARM (FORTH word) 8
warm start 7
warning 91
WARNING (FORTH word) 23, 46, 114
WASHING example 4
WHERE (editor command) 57
WHILE (FORTH word) 42, 43, 114


WHITE (graphics word) 66
WIDTH (FORTH word) 23, 115, 122
WORDS example 73
X (FORTH word) 115
(editor command) 62
X-register 25
XDIR (graphics word) 67
XOR (FORTH word) 15, 115
XSAVE location 25, 25


YDIR (graphics word) 67
zero—page location 25
[ (FORTH word) 30, 115
[COMPILE] (FORTH word) 30, 115
example 31, 32
] (FORTH word) 30, 115