mirror/ioccc
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

force use of pandoc version >= 3.1.12.2

Browse Source 
The `bin/pandoc-wrapper.sh` tool will verify that the ssytem
has a recent enough version of `pandoc(1)`.

We explain why we insist on `pandoc(1)`. version >= 3.1.12.2
in comments where the $PANDOC_MIN_VERSION is set.

Added `--wrap=preserve` as a default argument to `pandoc(1)`.
While this change will impact every HTML file generated,
we believe this will improve the ability of someone to look
at the generated HTML and correlate with the original markdown.

Changed `bin/pandoc-wrapper.sh` VERSION from "1.4.2 2024-03-17"
to "1.4.3 2024-03-18".

Performed `make www` to achieve all of the above and to generate
HTML using the new  `--wrap=preserve` `pandoc(1)` mode.
This commit is contained in:
Landon Curt Noll committed 2024-03-18 01:26:39 -07:00
1 parent 9eb7a54e9f
commit b0b212acb1
402 files changed
+37061 -55132
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

1992/buzzard.2/buzzard.2.design.html
+208 -214
View File
@@ -122,62 +122,63 @@
<h1 id="first-third-almost-forth">FIRST &amp; THIRD (almost FORTH)</h1>
<p>FORTH is a language mostly familiar to users of “small” machines.
FORTH programs are small because they are interpreted–a function call in
FORTH takes two bytes. FORTH is an extendable language– built-in
primitives are indistinguishable from user-defined <em>words</em>. FORTH
interpreters are small because much of the system can be coded in
FORTH–only a small number of primitives need to be implemented. Some
FORTH interpreters can also compile defined words into machine code,
resulting in a fast system.</p>
FORTH programs are small because they are interpreted–a function
call in FORTH takes two bytes. FORTH is an extendable language–
built-in primitives are indistinguishable from user-defined
<em>words</em>. FORTH interpreters are small because much of the system
can be coded in FORTH–only a small number of primitives need to
be implemented. Some FORTH interpreters can also compile defined
words into machine code, resulting in a fast system.</p>
<p>FIRST is an incredibly small language which is sufficient for
defining the language THIRD, which is mostly like FORTH. There are some
differences, and THIRD is probably just enough like FORTH for those
differences to be disturbing to regular FORTH users.</p>
<p>The only existing FIRST interpreter is written in obfuscated C, and
rings in at under 800 bytes of source code, although through deletion of
both whitespace and obfuscation it can be brought to about 650
defining the language THIRD, which is mostly like FORTH. There are
some differences, and THIRD is probably just enough like FORTH for
those differences to be disturbing to regular FORTH users.</p>
<p>The only existing FIRST interpreter is written in obfuscated C,
and rings in at under 800 bytes of source code, although through
deletion of both whitespace and obfuscation it can be brought to about 650
bytes.</p>
<p>This document FIRST defines the FIRST environment and primitives,
with relevant design decision explanations. It secondly documents the
general strategies we will use to implement THIRD. The THIRD section
demonstrates how the complete THIRD system is built up using FIRST.</p>
with relevant design decision explanations. It secondly documents
the general strategies we will use to implement THIRD. The THIRD
section demonstrates how the complete THIRD system is built up
using FIRST.</p>
<h2 id="section-1-first">Section 1: FIRST</h2>
<h3 id="environment">Environment</h3>
<p>FIRST implements a virtual machine. The machine has three chunks of
memory: “main memory”, “the stack”, and “string storage”. When the
virtual machine wishes to do random memory accesses, they come out of
main memory–it cannot access the stack or string storage.</p>
<p>FIRST implements a virtual machine. The machine has three chunks
of memory: “main memory”, “the stack”, and “string storage”. When
the virtual machine wishes to do random memory accesses, they come
out of main memory–it cannot access the stack or string storage.</p>
<p>The stack is simply a standard LIFO data structure that is used
implicitly by most of the FIRST primitives. The stack is made up of
<code>int</code>s, whatever size they are on the host machine.</p>
implicitly by most of the FIRST primitives. The stack is made up
of <code>int</code>s, whatever size they are on the host machine.</p>
<p>String storage is used to store the names of built-in and defined
primitives. Separate storage is used for these because it allows the C
code to use C string operations, reducing C source code size.</p>
primitives. Separate storage is used for these because it allows
the C code to use C string operations, reducing C source code size.</p>
<p>Main memory is a large array of <code>int</code>s. When we speak of
addresses, we actually mean indices into main memory. Main memory is
used for two things, primarily: the return stack and the dictionary.</p>
<p>The return stack is a LIFO data structure, independent of the
above-mentioned “stack”, which is used by FIRST to keep track of
function call return addresses.</p>
<p>The dictionary is a list of words. Each word contains a header and a
data field. In the header is the address of the previous word, an index
into the string storage indicating where the name of this word is
stored, and a “code pointer”. The code pointer is simply an integer
which names which “machine-language-primitive” implements this
instruction. For example, for defined words the code pointer names the
“run some code” primitive, which pushes the current program counter onto
the return stack and sets the counter to the address of the data field
for this word.</p>
<p>There are several important pointers into main memory. There is a
pointer to the most recently defined word, which is used to start
searches back through memory when compiling words. There is a pointer to
the top of the return stack. There is a pointer to the current end of
the dictionary, used while compiling.</p>
<p>For the last two pointers, namely the return stack pointer and the
dictionary pointer, there is an important distinction: the pointers
themselves are stored in main memory (in <code>FIRST</code>’s main
memory). This is critical, because it means FIRST programs can get at
them without any further primitives needing to be defined.</p>
addresses, we actually mean indices into main memory. Main memory
is used for two things, primarily: the return stack and the dictionary.</p>
<p>The return stack is a LIFO data structure, independent of
the above-mentioned “stack”, which is used by FIRST to keep
track of function call return addresses.</p>
<p>The dictionary is a list of words. Each word contains a header
and a data field. In the header is the address of the previous word,
an index into the string storage indicating where the name of this
word is stored, and a “code pointer”. The code pointer is simply
an integer which names which “machine-language-primitive” implements
this instruction. For example, for defined words the code pointer
names the “run some code” primitive, which pushes the current program
counter onto the return stack and sets the counter to the address of
the data field for this word.</p>
<p>There are several important pointers into main memory. There is
a pointer to the most recently defined word, which is used to start
searches back through memory when compiling words. There is a pointer
to the top of the return stack. There is a pointer to the current
end of the dictionary, used while compiling.</p>
<p>For the last two pointers, namely the return stack pointer and
the dictionary pointer, there is an important distinction: the pointers
themselves are stored in main memory (in <code>FIRST</code>’s main memory). This
is critical, because it means FIRST programs can get at them without
any further primitives needing to be defined.</p>
<h3 id="instructions">Instructions</h3>
<p>There are two kinds of FIRST instructions, normal instructions and
immediate instructions. Immediate instructions do something significant
@@ -196,8 +197,7 @@ means that by default FIRST simply compiles things.</p>
&lt;0 less than 0 pop top element of stack, push 1 if &lt; 0 else 0</code></pre>
<p>Note that we can synthesize addition and negation from binary minus,
but we cannot synthesize a time efficient divide or multiply from it.
<code>&lt;0</code> is synthesize-able, but only in a non-portable
way.</p>
<code>&lt;0</code> is synthesize-able, but only in a non-portable way.</p>
<h4 id="memory-operations">Memory Operations</h4>
<pre><code>SYMBOL NAME FUNCTION
@@ -216,16 +216,14 @@ _read read a space-delimited word, find it in the
dictionary, and compile a pointer to
that word&#39;s code pointer onto the
current end of the dictionary</code></pre>
<p>Although <code>_read</code> could be synthesized from key, we need
<code>_read</code> to be able to compile words to be able to start any
syntheses.</p>
<p>Although <code>_read</code> could be synthesized from key, we need <code>_read</code> to be able
to compile words to be able to start any syntheses.</p>
<h4 id="execution-operations">Execution Operations</h4>
<pre><code>NAME FUNCTION
exit leave the current function: pop the return stack
into the program counter</code></pre>
<h4 id="immediate-compilation-operations">Immediate (compilation)
Operations</h4>
<h4 id="immediate-compilation-operations">Immediate (compilation) Operations</h4>
<pre><code>SYMBOL NAME FUNCTION
: define read in the next space-delimited word, add it to
@@ -237,32 +235,31 @@ Operations</h4>
immediate immediate when used immediately after a name following a &#39;:&#39;,
it makes the word being defined run whenever
it is typed.</code></pre>
<p><code>:</code> cannot be synthesized, because we could not synthesize
anything. <code>immediate</code> has to be an immediate operation, so it
could not be synthesized unless by default operations were immediate;
but that would preclude our being able to do any useful compilation.</p>
<p><code>:</code> cannot be synthesized, because we could not synthesize anything.
<code>immediate</code> has to be an immediate operation, so it could not be
synthesized unless by default operations were immediate; but that
would preclude our being able to do any useful compilation.</p>
<h4 id="stack-operations">Stack Operations</h4>
<pre><code>NAME FUNCTION
pick pop top of stack, use as index into stack and copy up
that element</code></pre>
<p>If the data stack were stored in main memory, we could synthesize
<code>pick</code>; but putting the stack and stack pointer in main
memory would significantly increase the C source code size.</p>
<p>There are three more primitives, but they are “internal only”– they
have no names and no dictionary entries. The first is
<code>pushint</code>. It takes the next integer out of the instruction
stream and pushes it on the stack. This could be synthesized, but
probably not without using integer constants. It is generated by
<code>_read</code> when the input is not a known word. The second is
<code>compile me</code>. When this instruction is executed, a pointer to
the word’s data field is appended to the dictionary. The third is
<code>run me</code>–the word’s data field is taken to be a stream of
pointers to words, and is executed.</p>
<p>One last note about the environment: FIRST builds a very small word
internally that it executes as its main loop. This word calls
<code>_read</code> and then calls itself. Each time it calls itself, it
uses up a word on the return stack, so it will eventually trash things.
<p>If the data stack were stored in main memory, we could synthesize <code>pick</code>;
but putting the stack and stack pointer in main memory would significantly
increase the C source code size.</p>
<p>There are three more primitives, but they are “internal only”–
they have no names and no dictionary entries. The first is
<code>pushint</code>. It takes the next integer out of the instruction stream
and pushes it on the stack. This could be synthesized, but probably
not without using integer constants. It is generated by <code>_read</code> when
the input is not a known word. The second is <code>compile me</code>. When
this instruction is executed, a pointer to the word’s data field is
appended to the dictionary. The third is <code>run me</code>–the word’s data
field is taken to be a stream of pointers to words, and is executed.</p>
<p>One last note about the environment: FIRST builds a very small
word internally that it executes as its main loop. This word calls
<code>_read</code> and then calls itself. Each time it calls itself, it uses
up a word on the return stack, so it will eventually trash things.
This is discussed some more in section 2.</p>
<p>Here’s a handy summary of all the FIRST words:</p>
<pre><code>WORD DESCRIPTION
@@ -275,18 +272,18 @@ _read read a word from input and compile a pointer to it
exit stop running the current function
: compile the header of a definition
immediate modify the header to create an immediate word</code></pre>
<p>Here is a sample FIRST program. I’m assuming you’re using the ASCII
character set. FIRST does not depend upon ASCII, but since FIRST has no
syntax for character constants, one normally has to use decimal values.
This can be avoided using <code>getchar(3)</code>, though. Oh. One other
odd thing. FIRST initially builds its symbol table by calling
<code>:</code> several times, so it needs to get the names of the base
<p>Here is a sample FIRST program. I’m assuming you’re using
the ASCII character set. FIRST does not depend upon ASCII, but
since FIRST has no syntax for character constants, one normally has
to use decimal values. This can be avoided using <code>getchar(3)</code>, though.
Oh. One other odd thing. FIRST initially builds its symbol table
by calling <code>:</code> several times, so it needs to get the names of the base
symbols as its first 13 words of input. You could even name them
differently if you wanted.</p>
<p>These FIRST programs have FORTH comments in them: they are contained
inside parentheses. FIRST programs cannot have FORTH comments; but I
need some device to indicate what’s going on. (THIRD programs are an
entirely different subject.)</p>
inside parentheses. FIRST programs cannot have FORTH comments; but I need
some device to indicate what’s going on. (THIRD programs are an entirely
different subject.)</p>
<pre><code> ( Our first line gives the symbols for the built-ins )
: immediate _read @ ! - * / &lt;0 exit echo key _pick
@@ -315,73 +312,68 @@ test
Hello
)</code></pre>
<h2 id="section-2-motivating-third">Section 2: Motivating THIRD</h2>
<p>What is missing from FIRST? There are a large number of important
primitives that aren’t implemented, but which are easy to implement.
<code>drop</code>, which throws away the top of the stack, can be
implemented as <code>{ 0 * + }</code> – that is, multiply the top of the
stack by <code>0</code> (which turns the top of the stack into a
<code>0</code>), and then add the top two elements of the stack.</p>
<p>What is missing from FIRST? There are a large number of
important primitives that aren’t implemented, but which are
easy to implement. <code>drop</code>, which throws away the top of the
stack, can be implemented as <code>{ 0 * + }</code> – that is, multiply
the top of the stack by <code>0</code> (which turns the top of the stack
into a <code>0</code>), and then add the top two elements of the stack.</p>
<p><code>dup</code>, which copies the top of the stack, can be easily
implemented using temporary storage locations. Conveniently, FIRST
leaves memory locations <code>3</code>, <code>4</code>, and
<code>5</code> unused. So we can implement <code>dup</code> by writing
the top of stack into <code>3</code>, and then reading it out twice:
<code>{ 3 ! 3 @ 3 @ }</code>.</p>
<p>We will never use the FIRST primitive <code>pick</code> in building
THIRD, just to show that it can be done; <code>pick</code> is only
provided because <code>pick</code> itself cannot be built out of the
rest of <code>FIRST</code>’s building blocks.</p>
<p>So, instead of worrying about stack primitives and the like, what
else is missing from FIRST? We get recursion, but no control flow–no
conditional operations. We cannot at the moment write a looping routine
which terminates.</p>
<p>Another glaring dissimilarity between FIRST and FORTH is that there
is no “command mode”–you cannot be outside of a <code>:</code>
definition and issue some straight commands to be executed. Also, as we
noted above, we cannot do comments.</p>
<p>FORTH also provides a system for defining new data types, using the
words (in one version of FORTH) <code>builds and does</code>. We would
like to implement these words as well.</p>
<p>As the highest priority thing, we will build control flow structures
first. Once we have control structures, we can write recursive routines
that terminate, and we are ready to tackle tasks like parsing, and the
building of a command mode.</p>
<p>By the way, location <code>0</code> holds the dictionary pointer,
location <code>1</code> holds the return stack pointer, and location
<code>2</code> should always be <code>0</code>–it’s a fake dictionary
entry that means <code>pushint</code>.</p>
implemented using temporary storage locations. Conveniently,
FIRST leaves memory locations <code>3</code>, <code>4</code>, and <code>5</code> unused. So we can
implement <code>dup</code> by writing the top of stack into <code>3</code>, and then
reading it out twice: <code>{ 3 ! 3 @ 3 @ }</code>.</p>
<p>We will never use the FIRST primitive <code>pick</code> in building THIRD,
just to show that it can be done; <code>pick</code> is only provided because
<code>pick</code> itself cannot be built out of the rest of <code>FIRST</code>’s building
blocks.</p>
<p>So, instead of worrying about stack primitives and the
like, what else is missing from FIRST? We get recursion, but
no control flow–no conditional operations. We cannot at the
moment write a looping routine which terminates.</p>
<p>Another glaring dissimilarity between FIRST and FORTH is
that there is no “command mode”–you cannot be outside of a
<code>:</code> definition and issue some straight commands to be executed.
Also, as we noted above, we cannot do comments.</p>
<p>FORTH also provides a system for defining new data types,
using the words (in one version of FORTH) <code>builds and does</code>.
We would like to implement these words as well.</p>
<p>As the highest priority thing, we will build control flow
structures first. Once we have control structures, we can
write recursive routines that terminate, and we are ready to
tackle tasks like parsing, and the building of a command mode.</p>
<p>By the way, location <code>0</code> holds the dictionary pointer, location
<code>1</code> holds the return stack pointer, and location <code>2</code> should always
be <code>0</code>–it’s a fake dictionary entry that means <code>pushint</code>.</p>
<h2 id="section-3-building-third">Section 3: Building THIRD</h2>
<p>In this section we’ll eventually have real comments.</p>
<p>The first thing we have to do is give the symbols for our
built-ins.</p>
<p>The first thing we have to do is give the symbols for our built-ins.</p>
<pre><code>: immediate _read @ ! - * / &lt; exit echo key _pick</code></pre>
<p>Next we want to be mildly self commenting, so we define the word
<code>r</code> to push the <em>address of the return stack pointer</em>
onto the stack–NOT the value of the return stack pointer. (In fact, when
we run <code>r</code>, the value of the return stack pointer is
temporarily changed.)</p>
<p>Next we want to be mildly self commenting, so we define the word <code>r</code> to push
the <em>address of the return stack pointer</em> onto the stack–NOT the value of the
return stack pointer. (In fact, when we run <code>r</code>, the value of the return stack
pointer is temporarily changed.)</p>
<pre><code>: r 1 exit</code></pre>
<p>Next, we’re currently executing a short loop that contains
<code>_read</code> and recursion, which is slowly blowing up the return
stack. So let’s define a new word, from which you can never return. What
it does is drops the top value off the return stack, calls
<code>_read</code>, then calls itself. Because it kills the top of the
return stack, it can recurse indefinitely.</p>
<p>Next, we’re currently executing a short loop that contains <code>_read</code> and
recursion, which is slowly blowing up the return stack. So let’s define a new
word, from which you can never return. What it does is drops the top value off
the return stack, calls <code>_read</code>, then calls itself. Because it kills the top of
the return stack, it can recurse indefinitely.</p>
<pre><code>: ]
r @ Get the value of the return stack pointer
1 - Subtract one
r ! Store it back into the return stack pointer
_read Read and compile one word
] Start over</code></pre>
<p>Notice that we don’t need to exit, since we never come back. Also,
it’s possible that an immediate word may get run during
<code>_read</code>, and that <code>_read</code> will never return!</p>
<p>Notice that we don’t need to exit, since we never come back. Also, it’s
possible that an immediate word may get run during <code>_read</code>, and that <code>_read</code>
will never return!</p>
<p>Now let’s get compile running.</p>
<pre><code>: main immediate ]
main</code></pre>
<p>Next off, I’m going to do this the easy but non-portable way, and put
some character constant definitions in. I wanted them at the top of the
file, but that would have burned too much of the return stack.</p>
<p>Next off, I’m going to do this the easy but non-portable way, and put some
character constant definitions in. I wanted them at the top of the file, but
that would have burned too much of the return stack.</p>
<pre><code>: &#39;&quot;&#39; 34 exit
: &#39;)&#39; 41 exit
: &#39;\n&#39; 10 exit
@@ -390,29 +382,28 @@ file, but that would have burned too much of the return stack.</p>
: &#39;-&#39; 45 exit
: cr &#39;\n&#39; echo exit</code></pre>
<p>Next, we want to define some temporary variables for locations
<code>3</code>, <code>4</code>, and <code>5</code>, since this’ll make
our code look clearer.</p>
<p>Next, we want to define some temporary variables for locations <code>3</code>, <code>4</code>, and
<code>5</code>, since this’ll make our code look clearer.</p>
<pre><code>: _x 3 @ exit
: _x! 3 ! exit
: _y 4 @ exit
: _y! 4 ! exit
</code></pre>
<p>OK. Now, we want to make THIRD look vaguely like FORTH, so we’re
going to define <code>;</code>. What <code>;</code> ought to do is
terminate a compilation, and turn control over to the command-mode
handler. We don’t have one, so all we want <code>;</code> to do for now
is compile <code>exit</code> at the end of the current word. To do this
we’ll need several other words.</p>
<p>Swap by writing out the top two elements into temps, and then reading
them back in the other order.</p>
<p>OK. Now, we want to make THIRD look vaguely like FORTH,
so we’re going to define <code>;</code>. What <code>;</code> ought to do is
terminate a compilation, and turn control over to the
command-mode handler. We don’t have one, so all we want
<code>;</code> to do for now is compile <code>exit</code> at the end of the
current word. To do this we’ll need several other words.</p>
<p>Swap by writing out the top two elements into temps, and
then reading them back in the other order.</p>
<pre><code>: swap _x! _y! _x _y exit</code></pre>
<p>Take another look and make sure you see why that works, since it
LOOKS like I’m reading them back in the same order–in fact, it not only
looks like it, but I AM!</p>
<p>Addition might be nice to have. To add, we need to negate the top
element of the stack, and then subtract. To negate, we subtract from
<code>0</code>.</p>
<p>Take another look and make sure you see why that works,
since it LOOKS like I’m reading them back in the same
order–in fact, it not only looks like it, but I AM!</p>
<p>Addition might be nice to have. To add, we need to
negate the top element of the stack, and then subtract.
To negate, we subtract from <code>0</code>.</p>
<pre><code>: +
0 swap -
-
@@ -420,12 +411,13 @@ element of the stack, and then subtract. To negate, we subtract from
</code></pre>
<p>Create a copy of the top of stack.</p>
<pre><code>: dup _x! _x _x exit</code></pre>
<p>Get a mnemonic name for our dictionary pointer–we need to compile
stuff, so it goes through this.</p>
<p>Get a mnemonic name for our dictionary pointer–we need
to compile stuff, so it goes through this.</p>
<pre><code>: h 0 exit</code></pre>
<p>We’re going to need to advance that pointer, so let’s make a generic
pointer-advancing function. Given a pointer to a memory location,
increment the value at that memory location.</p>
<p>We’re going to need to advance that pointer, so let’s
make a generic pointer-advancing function.
Given a pointer to a memory location, increment the value
at that memory location.</p>
<pre><code>: inc
dup @ Get another copy of the address, and get the value
so now we have value, address on top of stack.
@@ -441,12 +433,13 @@ stack into the dictionary, and advance the pointer</p>
h inc And increment the dictionary pointer
exit</code></pre>
<p><code>'</code> is a standard FORTH word. It should push the address
of the word that follows it onto the stack. We could do this by making
<code>'</code> immediate, but then it’d need to parse the next word.
Instead, we compile the next word as normal. When <code>'</code> is
executed, the top of the return stack will point into the instruction
stream immediately after the <code>'</code>. We push the word there, and
advance the return stack pointer so that we don’t execute it.</p>
of the word that follows it onto the stack. We could
do this by making <code>'</code> immediate, but then it’d need to
parse the next word. Instead, we compile the next word
as normal. When <code>'</code> is executed, the top of the return
stack will point into the instruction stream immediately
after the <code>'</code>. We push the word there, and advance the
return stack pointer so that we don’t execute it.</p>
<pre><code>: &#39;
r @ Get the address of the top of return stack
We currently have a pointer to the top of return stack
@@ -469,18 +462,19 @@ the address of <code>exit</code> onto the stack, then writes it out.</p>
exit And we should return</code></pre>
<p>Now let’s test out <code>;</code> by defining a useful word:</p>
<pre><code>: drop 0 * + ;</code></pre>
<p>Since we have <code>inc</code>, we ought to make
<code>dec</code>:</p>
<p>Since we have <code>inc</code>, we ought to make <code>dec</code>:</p>
<pre><code>: dec dup @ 1 - swap ! ;</code></pre>
<p>Our next goal, now that we have <code>;</code>, is to implement
<code>if-then</code>. To do this, we’ll need to play fast and loose with
the return stack, so let’s make some words to save us some effort.</p>
<p>First we want a word that pops off the top of the normal stack and
pushes it on top of the return stack. We’ll call this <code>tor</code>,
for TO-Return-stack. It sounds easy, but when <code>tor</code> is
running, there’s an extra value on the return stack–<code>tor</code>’s
return address! So we have to pop that off first… We better just bite
the bullet and code it out–but we can’t really break it into smaller
<code>if-then</code>. To do this, we’ll need to play fast and
loose with the return stack, so let’s make some
words to save us some effort.</p>
<p>First we want a word that pops off the top of the normal
stack and pushes it on top of the return stack. We’ll
call this <code>tor</code>, for TO-Return-stack. It sounds easy,
but when <code>tor</code> is running, there’s an extra value on the
return stack–<code>tor</code>’s return address! So we have to pop
that off first… We better just bite the bullet and
code it out–but we can’t really break it into smaller
words, because that’ll trash the return stack.</p>
<pre><code>: tor
r @ @ Get the value off the top of the return stack
@@ -489,8 +483,8 @@ words, because that’ll trash the return stack.</p>
r @ 1 + r ! Increment the return stack pointer--but can&#39;t use inc
r @ ! Store our return address back on the return stack
;</code></pre>
<p>Next we want the opposite routine, which pops the top of the return
stack, and puts it on the normal stack.</p>
<p>Next we want the opposite routine, which pops the top
of the return stack, and puts it on the normal stack.</p>
<pre><code>: fromr
r @ @ Save old value
r @ 1 - r ! Decrement pointer
@@ -498,25 +492,26 @@ stack, and puts it on the normal stack.</p>
swap Bring return address to top
r @ ! Store it and return
;</code></pre>
<p>Now, if we have a routine that’s recursing, and we want to be polite
about the return stack, right before we recurse we can run
<code>{ fromr drop }</code> so the stack won’t blow up. This means,
though, that the first time we enter this recursive routine, we blow our
<em>real</em> return address–so when we’re done, we’ll return up two
levels. To save a little, we make <code>tail</code> mean
<code>{ fromr drop }</code>; however, it’s more complex since there’s a
new value on top of the return stack.</p>
<p>Now, if we have a routine that’s recursing, and we
want to be polite about the return stack, right before
we recurse we can run <code>{ fromr drop }</code> so the stack won’t
blow up. This means, though, that the first time we
enter this recursive routine, we blow our <em>real</em> return
address–so when we’re done, we’ll return up two levels.
To save a little, we make <code>tail</code> mean <code>{ fromr drop }</code>;
however, it’s more complex since there’s a new value on
top of the return stack.</p>
<pre><code>: tail fromr fromr drop tor ;</code></pre>
<p>Now, we want to do <code>if</code>. To do this, we need to convert
values to boolean values. The next few words set this up.</p>
values to boolean values. The next few words set this
up.</p>
<p><code>minus</code> gives us unary negation.</p>
<pre><code>: minus 0 swap - ;</code></pre>
<p>If top of stack is boolean, <code>bnot</code> gives us inverse</p>
<pre><code>: bnot 1 swap - ;</code></pre>
<p>To compare two numbers, subtract and compare to <code>0</code>.</p>
<pre><code>: &lt; - &lt;0 ;</code></pre>
<p><code>logical</code> turns the top of stack into either
<code>0</code> or <code>1</code>.</p>
<p><code>logical</code> turns the top of stack into either <code>0</code> or <code>1</code>.</p>
<pre><code>: logical
dup Get two copies of it
0 &lt; 1 if &lt; 0, 0 otherwise
@@ -526,14 +521,13 @@ values to boolean values. The next few words set this up.</p>
;</code></pre>
<p><code>not</code> returns 1 if top of stack is 0, and 0 otherwise</p>
<pre><code>: not logical bnot ;</code></pre>
<p>We can test equality by subtracting and comparing to
<code>0</code>.</p>
<p>We can test equality by subtracting and comparing to <code>0</code>.</p>
<pre><code>: = - not ;</code></pre>
<p>Just to show how you compute a branch: Suppose you’ve compiled a call
to branch, and immediately after it is an integer constant with the
offset of how far to branch. To branch, we use the return stack to read
the offset, and add that on to the top of the return stack, and
return.</p>
<p>Just to show how you compute a branch: Suppose you’ve
compiled a call to branch, and immediately after it is
an integer constant with the offset of how far to branch.
To branch, we use the return stack to read the offset, and
add that on to the top of the return stack, and return.</p>
<pre><code>: branch
r @ Address of top of return stack
@ Our return address
@@ -542,12 +536,12 @@ return.</p>
+ The address we want to execute at
r @ ! Store it back onto the return stack
;</code></pre>
<p>For conditional branches, we want to branch by a certain amount if
true, otherwise we want to skip over the branch offset constant–that is,
branch by one. Assuming that the top of the stack is the branch offset,
and the second on the stack is <code>1</code> if we should branch, and
<code>0</code> if not, the following computes the correct branch
offset.</p>
<p>For conditional branches, we want to branch by a certain
amount if true, otherwise we want to skip over the branch
offset constant–that is, branch by one. Assuming that
the top of the stack is the branch offset, and the second
on the stack is <code>1</code> if we should branch, and <code>0</code> if not, the
following computes the correct branch offset.</p>
<pre><code>: computebranch 1 - * 1 + ;</code></pre>
<p>Branch if the value on top of the stack is <code>0</code>.</p>
<pre><code>: notbranch
@@ -558,31 +552,31 @@ offset.</p>
r @ ! Store it
;</code></pre>
<p><code>here</code> is a standard FORTH word which returns a pointer to
the current dictionary address–that is, the value of the dictionary
pointer.</p>
the current dictionary address–that is, the value of
the dictionary pointer.</p>
<pre><code>: here h @ ;</code></pre>
<p>We’re ALL SET to compile <code>if...else...then</code> constructs!
Here’s what we do. When we get <code>if</code>, we compile a call to
<code>notbranch</code>, and then compile a dummy offset, because we
don’t know where the <code>then</code> will be. On the <em>stack</em> we
leave the address where we compiled the dummy offset. <code>then</code>
will calculate the offset and fill it in for us.</p>
Here’s what we do. When we get <code>if</code>, we compile a call
to <code>notbranch</code>, and then compile a dummy offset, because
we don’t know where the <code>then</code> will be. On the <em>stack</em>
we leave the address where we compiled the dummy offset.
<code>then</code> will calculate the offset and fill it in for us.</p>
<pre><code>: if immediate
&#39; notbranch , Compile notbranch
here Save the current dictionary address
0 , Compile a dummy value
;</code></pre>
<p><code>then</code> expects the address to <code>fixup</code> to be on
the stack.</p>
<p><code>then</code> expects the address to <code>fixup</code> to be on the stack.</p>
<pre><code>: then immediate
dup Make another copy of the address
here Find the current location, where to branch to
swap - Calculate the difference between them
swap ! Bring the address to the top, and store it.
;</code></pre>
<p>Now that we can do <code>if...then</code> statements, we can do some
parsing! Let’s introduce real FORTH comments. <code>find-)</code> will
scan the input until it finds a <code>)</code>, and exit.</p>
<p>Now that we can do <code>if...then</code> statements, we can do
some parsing! Let’s introduce real FORTH comments.
<code>find-)</code> will scan the input until it finds a <code>)</code>, and
exit.</p>
<pre><code>: find-)
key Read in a character
&#39;)&#39; = Compare it to close parentheses