misc/kforth: Part 0x04 - parsing numerics.

This commit is contained in:
Kyle Isom 2018-02-23 19:19:29 -08:00
parent 0ae7d49593
commit 505d71906c
12 changed files with 479 additions and 2 deletions

View File

@ -1,6 +1,7 @@
CXXSTD := c++14
CXXFLAGS := -std=$(CXXSTD) -Wall -Werror -g -O0
OBJS := linux/io.o \
io.o \
parser.o \
kforth.o
TARGET := kforth

3
defs.h
View File

@ -3,6 +3,9 @@
#ifdef __linux__
#include "linux/defs.h"
#else
typedef int KF_INT;
constexpr uint8_t STACK_SIZE = 16;
#endif
constexpr size_t MAX_TOKEN_LENGTH = 16;

View File

@ -9,6 +9,7 @@ Contents:
part-0x01
part-0x02
part-0x03
part-0x04
Indices and tables
==================

313
doc/part-0x04.rst Normal file
View File

@ -0,0 +1,313 @@
Write You a Forth, 0x04
-----------------------
:date: 2018-02-23 19:20
:tags: wyaf, forth
So, I lied about words being next. When I thought about it some more, what I
really need to do is start adding the stack in and adding support for parsing
numerics. I'll start with the stack, because it's pretty straightforward.
I've added a new definition: ``constexpr uint8_t STACK_SIZE = 128``. This goes
in the ``linux/defs.h``, and the ``#else`` in the top ``defs.h`` will set a
smaller stack size for other targets. I've also defined a type called ``KF_INT``
that, on Linux, is a ``uint32_t``::
index 4dcc540..e070d27 100644
--- a/defs.h
+++ b/defs.h
@@ -3,6 +3,9 @@
#ifdef __linux__
#include "linux/defs.h"
+#else
+typedef int KF_INT;
+constexpr uint8_t STACK_SIZE = 16;
#endif
constexpr size_t MAX_TOKEN_LENGTH = 16;
diff --git a/linux/defs.h b/linux/defs.h
index 57cdaeb..3740f5a 100644
--- a/linux/defs.h
+++ b/linux/defs.h
@@ -4,4 +4,7 @@
#include <stddef.h>
#include <stdint.h>
+typedef int32_t KF_INT;
+constexpr uint8_t STACK_SIZE = 128;
+
#endif
\ No newline at end of file
It seems useful to be able to adapt the kind of numbers supported; an AVR might do
better with 16-bit integers, for example.
``stack.h``
^^^^^^^^^^^
The stack is going to be templated, because we'll need a ``double`` stack later
for floating point and a return address stack later. This means everything will
go under ``stack.h``. This is a pretty simple implementation that's CS 101 material;
I've opted to have the interface return ``bool``\ s for everything to indicate stack
overflow and underflow and out of bounds::
#ifndef __KF_STACK_H__
#define __KF_STACK_H__
#include "defs.h"
template <typename T>
class Stack {
public:
bool push(T val);
bool pop(T &val);
bool get(size_t, T &);
size_t size(void) { return this->arrlen; };
private:
T arr[STACK_SIZE];
size_t arrlen;
};
// push returns false if there was a stack overflow.
template <typename T>
bool
Stack<T>::push(T val)
{
if ((this->arrlen + 1) > STACK_SIZE) {
return false;
}
this->arr[this->arrlen++] = val;
return true;
}
// pop returns false if there was a stack underflow.
template <typename T>
bool
Stack<T>::pop(T &val)
{
if (this->arrlen == 0) {
return false;
}
val = this->arr[this->arrlen - 1];
this->arrlen--;
}
// get returns false on invalid bounds.
template <typename T>
bool
Stack<T>::get(size_t i, T &val)
{
if (i > this->arrlen) {
return false;
}
val = this->arr[i];
return true;
}
#endif // __KF_STACK_H__
I'll put a ``Stack<KF_INT>`` in ``kforth.cc`` later on. For now, this gives me
an interface for the numeric parser to push a number onto the stack.
``parse_num``
^^^^^^^^^^^^^
It seems like the best place for this is in ``parser.cc`` --- though I might
move into a token processor later. The definition for this goes in ``parser.h``,
and the body is in ``parser.cc``::
// parse_num tries to parse the token as a signed base 10 number,
// pushing it onto the stack if needed.
bool
parse_num(struct Token *token, Stack<KF_INT> &s)
{
KF_INT n = 0;
uint8_t i = 0;
bool sign = false;
It turns out you can't parse a zero-length token as a number...
::
if (token->length == 0) {
return false;
}
I'll need to invert the number later if it's negative, but it's worth checking
the first character to see if it's negative.
::
if (token->token[i] == '-') {
i++;
sign = true;
}
Parsing is done by checking whether each character is within the range of the ASCII
numeral values. Later on, I might add in separate functions for processing base 10
and base 16 numbers, and decide which to use based on a prefix (like ``0x``). If the
character is between those values, then the working number is multiplied by 10 and
the digit added.
::
while (i < token->length) {
if (token->token[i] < '0') {
return false;
}
if (token->token[i] > '9') {
return false;
}
n *= 10;
n += (uint8_t)(token->token[i] - '0');
i++;
}
If it was a negative number, then the working number has to be inverted::
if (sign) {
n *= -1;
}
Finally, return the result of pushing the number on the stack. One thing that
might come back to get me later is that this makes it impossible to tell if a
failure to parse the number is due to an invalid number or due to a stack
overflow. This will be a good candidate for revisiting later.
::
return s.push(n);
}
``io.cc``
^^^^^^^^^^
Conversely, it'll be useful to write a number to an ``IO`` interface. It
*seems* more useful right now to just provide a number → I/O function, but
that'll be easily adapted to a number → buffer function later. This will add
a real function to ``io.h``, which will require a corresponding ``io.cc``
(which also needs to be added to the ``Makefile``)::
#include "defs.h"
#include "io.h"
#include <string.h>
void
write_num(IO &interface, KF_INT n)
{
Through careful scientific study, I have determined that most number of digits
that a 32-bit integer needs is 10 bytes (sans the sign!). This will absolutely
need to be changed if ``KF_INT`` is ever moved to 64-bit (or larger!) numbers.
There's a TODO in the actual source code that notes this. ::
char buf[10];
uint8_t i = 10;
memset(buf, 0, 10);
Because this is going out to an I/O interface, I don't need to store the sign
in the buffer itself and can just print it and invert the number. Inverting is
important; I ran into a bug earlier where I didn't invert it and my subtractions
below were correspondingly off.
::
if (n < 0) {
interface.wrch('-');
n *= -1;
}
The buffer has to be filled from the end to the beginning to do the inverse of
the parsing method::
while (n != 0) {
char ch = (n % 10) + '0';
buf[i--] = ch;
n /= 10;
}
But then it can be just dumped to the interface::
interface.wrbuf(buf+i, 11-i);
}
``kforth.cc``
^^^^^^^^^^^^^^
And now I come to the fun part: adding the stack in. After including ``stack.h``,
I've added a stack implementation to the top of the file::
// dstack is the data stack.
static Stack<KF_INT> dstack;
It's kind of useful to be able to print the stack::
static void
write_dstack(IO &interface)
{
KF_INT tmp;
interface.wrch('<');
for (size_t i = 0; i < dstack.size(); i++) {
if (i > 0) {
interface.wrch(' ');
}
dstack.get(i, tmp);
write_num(interface, tmp);
}
interface.wrch('>');
}
Surrounding the stack in angle brackets is a cool stylish sort of thing, I
guess. All this is no good if the interpreter isn't actually hooked up to the
number parser::
// The new while loop in the parser function in kforth.cc:
while ((result = parse_next(buf, buflen, &offset, &token)) == PARSE_OK) {
interface.wrbuf((char *)"token: ", 7);
interface.wrbuf(token.token, token.length);
interface.wrln((char *)".", 1);
if (!parse_num(&token, dstack)) {
interface.wrln((char *)"failed to parse numeric", 23);
}
// Temporary hack until the interpreter is working further.
if (match_token(token.token, token.length, bye, 3)) {
interface.wrln((char *)"Goodbye!", 8);
exit(0);
}
}
But does it blend?
^^^^^^^^^^^^^^^^^^
Hopefully this works::
~/code/kforth (0) $ make
g++ -std=c++14 -Wall -Werror -g -O0 -c -o linux/io.o linux/io.cc
g++ -std=c++14 -Wall -Werror -g -O0 -c -o io.o io.cc
g++ -std=c++14 -Wall -Werror -g -O0 -c -o parser.o parser.cc
g++ -std=c++14 -Wall -Werror -g -O0 -c -o kforth.o kforth.cc
g++ -o kforth linux/io.o io.o parser.o kforth.o
~/code/kforth (0) $ ./kforth
kforth interpreter
<>
? 2 -2 30 1000 -1010
token: 2.
token: -2.
token: 30.
token: 1000.
token: -1010.
ok.
<2 -2 30 1000 -1010>
? bye
token: bye.
failed to parse numeric
Goodbye!
~/code/kforth (0) $
So there's that. Okay, next time *for real* I'll do a vocabulary thing.

27
io.cc Normal file
View File

@ -0,0 +1,27 @@
#include "defs.h"
#include "io.h"
#include <string.h>
void
write_num(IO &interface, KF_INT n)
{
// TODO(kyle): make the size of the buffer depend on the size of
// KF_INT.
char buf[10];
uint8_t i = 10;
memset(buf, 0, i);
if (n < 0) {
interface.wrch('-');
n *= -1;
}
while (n != 0) {
char ch = (n % 10) + '0';
buf[i--] = ch;
n /= 10;
}
interface.wrbuf(buf+i, 11-i);
}

3
io.h
View File

@ -21,4 +21,7 @@ public:
virtual void wrln(char *buf, size_t len) = 0;
};
void write_num(IO &, KF_INT);
#endif // __KF_IO_H__

View File

@ -1,7 +1,9 @@
#include "io.h"
#include "parser.h"
#include "stack.h"
#include <stdlib.h>
#include <string.h>
#ifdef __linux__
#include "linux.h"
@ -10,6 +12,26 @@
static char ok[] = "ok.\n";
static char bye[] = "bye";
// dstack is the data stack.
static Stack<KF_INT> dstack;
static void
write_dstack(IO &interface)
{
KF_INT tmp;
interface.wrch('<');
for (size_t i = 0; i < dstack.size(); i++) {
if (i > 0) {
interface.wrch(' ');
}
dstack.get(i, tmp);
write_num(interface, tmp);
}
interface.wrch('>');
}
static bool
parser(IO &interface, const char *buf, const size_t buflen)
{
@ -28,6 +50,10 @@ parser(IO &interface, const char *buf, const size_t buflen)
interface.wrbuf(token.token, token.length);
interface.wrln((char *)".", 1);
if (!parse_num(&token, dstack)) {
interface.wrln((char *)"failed to parse numeric", 23);
}
// Temporary hack until the interpreter is working further.
if (match_token(token.token, token.length, bye, 3)) {
interface.wrln((char *)"Goodbye!", 8);
@ -58,6 +84,8 @@ interpreter(IO &interface)
static char linebuf[81];
while (true) {
write_dstack(interface);
interface.wrch('\n');
interface.wrch('?');
interface.wrch(' ');
buflen = interface.rdbuf(linebuf, 80, true, '\n');

View File

@ -6,7 +6,5 @@
// build support for linux
#include "linux/io.h"
constexpr uint8_t STACK_SIZE = 128;
#endif // __KF_LINUX_H__

View File

@ -4,4 +4,7 @@
#include <stddef.h>
#include <stdint.h>
typedef int32_t KF_INT;
constexpr uint8_t STACK_SIZE = 128;
#endif

View File

@ -1,5 +1,6 @@
#include "defs.h"
#include "parser.h"
#include "stack.h"
#include <string.h>
@ -68,4 +69,41 @@ parse_next(const char *buf, const size_t length, size_t *offset,
*offset = cursor;
return PARSE_OK;
}
bool
parse_num(struct Token *token, Stack<KF_INT> &s)
{
KF_INT n = 0;
uint8_t i = 0;
bool sign = false;
if (token->length == 0) {
return false;
}
if (token->token[i] == '-') {
i++;
sign = true;
}
while (i < token->length) {
if (token->token[i] < '0') {
return false;
}
if (token->token[i] > '9') {
return false;
}
n *= 10;
n += (uint8_t)(token->token[i] - '0');
i++;
}
if (sign) {
n *= -1;
}
return s.push(n);
}

View File

@ -2,6 +2,7 @@
#define __KF_PARSER_H__
#include "defs.h"
#include "stack.h"
struct Token {
char *token;
@ -18,5 +19,9 @@ typedef enum _PARSE_RESULT_ : uint8_t {
bool match_token(const char *, const size_t, const char *, const size_t);
PARSE_RESULT parse_next(const char *, const size_t, size_t *, struct Token *);
// TODO(kyle): investigate a better return value, e.g. to differentiate between
// stack failures and parse failures.
bool parse_num(struct Token *, Stack<KF_INT> &);
#endif // __KF_PARSER_H__

57
stack.h Normal file
View File

@ -0,0 +1,57 @@
#ifndef __KF_STACK_H__
#define __KF_STACK_H__
#include "defs.h"
template <typename T>
class Stack {
public:
bool push(T val);
bool pop(T &val);
bool get(size_t, T &);
size_t size(void) { return this->arrlen; };
private:
T arr[STACK_SIZE];
size_t arrlen;
};
// push returns false if there was a stack overflow.
template <typename T>
bool
Stack<T>::push(T val)
{
if ((this->arrlen + 1) > STACK_SIZE) {
return false;
}
this->arr[this->arrlen++] = val;
return true;
}
// pop returns false if there was a stack underflow.
template <typename T>
bool
Stack<T>::pop(T &val)
{
if (this->arrlen == 0) {
return false;
}
val = this->arr[this->arrlen - 1];
this->arrlen--;
}
// get returns false on invalid bounds.
template <typename T>
bool
Stack<T>::get(size_t i, T &val)
{
if (i > this->arrlen) {
return false;
}
val = this->arr[i];
return true;
}
#endif // __KF_STACK_H__