kuroko/compiler.c

1916 lines
56 KiB
C
Raw Normal View History

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
2020-12-28 10:07:52 +03:00
#include <sys/types.h>
#include "kuroko.h"
#include "compiler.h"
#include "memory.h"
#include "scanner.h"
#include "object.h"
#include "debug.h"
#include "vm.h"
/**
* There's nothing really especially different here compared to the Lox
* compiler from Crafting Interpreters. A handful of additional pieces
* of functionality are added, and some work is done to make blocks use
* indentation instead of braces, but the basic layout and operation
* of the compiler are the same top-down Pratt parser.
*
* The parser error handling has been improved over the Lox compiler with
* the addition of column offsets and a printed copy of the original source
* line and the offending token.
*
* String parsing also includes escape sequence support, so you can print
* quotation marks properly, as well as escape sequences for terminals.
*
* One notable part of the compiler is the handling of list comprehensions.
* In order to support Python-style syntax, the parser has been set up to
* support rolling back to a previous state, so that when the compiler sees
* an expression with references to a variable that has yet to be defined it
* will first output the expression as if that variable was a global, then it
* will see the 'in', rewind, parse the rest of the list comprehension, and
* then output the expression as a loop body, with the correct local references.
*
* if/else and try/except blocks also have to similarly handle rollback cases
* as they can not peek forward to see if a statement after an indentation
* block is an else/except.
*/
typedef struct {
KrkToken current;
KrkToken previous;
int hadError;
int panicMode;
} Parser;
typedef enum {
PREC_NONE,
PREC_ASSIGNMENT, /* = */
PREC_OR, /* or */
PREC_AND, /* and */
2021-01-01 02:55:39 +03:00
PREC_BITOR, /* | */
PREC_BITXOR, /* ^ */
PREC_BITAND, /* & */
PREC_EQUALITY, /* == != in */
PREC_COMPARISON, /* < > <= >= */
2021-01-01 02:55:39 +03:00
PREC_SHIFT, /* << >> */
PREC_TERM, /* + - */
2021-01-01 02:55:39 +03:00
PREC_FACTOR, /* * / % */
PREC_UNARY, /* ! - not */
PREC_CALL, /* . () */
PREC_PRIMARY
} Precedence;
2020-12-26 10:53:15 +03:00
typedef void (*ParseFn)(int);
typedef struct {
const char * name;
ParseFn prefix;
ParseFn infix;
Precedence precedence;
} ParseRule;
typedef struct {
KrkToken name;
ssize_t depth;
int isCaptured;
} Local;
typedef struct {
size_t index;
int isLocal;
} Upvalue;
typedef enum {
TYPE_FUNCTION,
TYPE_MODULE,
TYPE_METHOD,
TYPE_INIT,
} FunctionType;
typedef struct Compiler {
struct Compiler * enclosing;
KrkFunction * function;
FunctionType type;
size_t localCount;
size_t scopeDepth;
size_t localsSpace;
Local * locals;
size_t upvaluesSpace;
Upvalue * upvalues;
2020-12-31 10:59:51 +03:00
size_t loopLocalCount;
size_t breakCount;
size_t breakSpace;
int * breaks;
size_t continueCount;
size_t continueSpace;
int * continues;
} Compiler;
typedef struct ClassCompiler {
struct ClassCompiler * enclosing;
KrkToken name;
int hasSuperClass;
} ClassCompiler;
2021-01-01 10:58:13 +03:00
static Parser parser;
static Compiler * current = NULL;
static ClassCompiler * currentClass = NULL;
static KrkChunk * currentChunk() {
return &current->function->chunk;
}
#define EMIT_CONSTANT_OP(opc, arg) do { if (arg < 256) { emitBytes(opc, arg); } \
else { emitBytes(opc ## _LONG, arg >> 16); emitBytes(arg >> 8, arg); } } while (0)
static void initCompiler(Compiler * compiler, FunctionType type) {
compiler->enclosing = current;
current = compiler;
compiler->function = NULL;
compiler->type = type;
compiler->scopeDepth = 0;
compiler->function = krk_newFunction();
2020-12-31 10:59:51 +03:00
compiler->localCount = 0;
compiler->localsSpace = 8;
compiler->locals = GROW_ARRAY(Local,NULL,0,8);
compiler->upvaluesSpace = 0;
compiler->upvalues = NULL;
compiler->breakCount = 0;
compiler->breakSpace = 0;
compiler->breaks = NULL;
compiler->continueCount = 0;
compiler->continueSpace = 0;
compiler->continues = NULL;
compiler->loopLocalCount = 0;
if (type != TYPE_MODULE) {
2020-12-28 05:11:50 +03:00
current->function->name = krk_copyString(parser.previous.start, parser.previous.length);
}
if (type == TYPE_INIT || type == TYPE_METHOD) {
Local * local = &current->locals[current->localCount++];
local->depth = 0;
local->isCaptured = 0;
local->name.start = "self";
local->name.length = 4;
}
}
static void parsePrecedence(Precedence precedence);
static ssize_t parseVariable(const char * errorMessage);
static void variable(int canAssign);
2020-12-26 10:53:15 +03:00
static void defineVariable(size_t global);
static ssize_t identifierConstant(KrkToken * name);
static ssize_t resolveLocal(Compiler * compiler, KrkToken * name);
static ParseRule * getRule(KrkTokenType type);
static void defDeclaration();
2020-12-26 10:53:15 +03:00
static void expression();
static void statement();
static void declaration();
2020-12-26 14:39:47 +03:00
static void or_(int canAssign);
static void and_(int canAssign);
2020-12-27 10:45:34 +03:00
static void classDeclaration();
static void declareVariable();
static void namedVariable(KrkToken name, int canAssign);
static void addLocal(KrkToken name);
2020-12-30 10:59:21 +03:00
static void string(int canAssign);
2021-01-02 15:58:24 +03:00
static KrkToken decorator(size_t level, FunctionType type);
2021-01-03 11:59:50 +03:00
static void call(int canAssign);
static void errorAt(KrkToken * token, const char * message) {
if (parser.panicMode) return;
parser.panicMode = 1;
size_t i = (token->col - 1);
while (token->linePtr[i] && token->linePtr[i] != '\n') i++;
fprintf(stderr, "Parse error in \"%s\" on line %d col %d (%s): %s\n"
" %.*s\033[31m%.*s\033[39m%.*s\n"
" %-*s\033[31m^\033[39m\n",
currentChunk()->filename->chars,
(int)token->line,
(int)token->col,
getRule(token->type)->name,
message,
(int)(token->col - 1),
token->linePtr,
(int)(token->literalWidth),
token->linePtr + (token->col - 1),
(int)(i - (token->col - 1 + token->literalWidth)),
token->linePtr + (token->col - 1 + token->literalWidth),
(int)token->col-1,
""
);
parser.hadError = 1;
}
static void error(const char * message) {
errorAt(&parser.previous, message);
}
static void errorAtCurrent(const char * message) {
errorAt(&parser.current, message);
}
static void advance() {
parser.previous = parser.current;
for (;;) {
parser.current = krk_scanToken();
2020-12-28 04:54:25 +03:00
#ifdef ENABLE_SCAN_TRACING
if (vm.flags & KRK_ENABLE_SCAN_TRACING) {
fprintf(stderr, "[%s %d:%d '%.*s'] ",
getRule(parser.current.type)->name,
(int)parser.current.line,
(int)parser.current.col,
(int)parser.current.length,
parser.current.start);
2020-12-28 04:54:25 +03:00
}
#endif
if (parser.current.type == TOKEN_RETRY) continue;
if (parser.current.type != TOKEN_ERROR) break;
errorAtCurrent(parser.current.start);
}
}
static void consume(KrkTokenType type, const char * message) {
if (parser.current.type == type) {
advance();
return;
}
errorAtCurrent(message);
}
2020-12-26 10:53:15 +03:00
static int check(KrkTokenType type) {
return parser.current.type == type;
}
static int match(KrkTokenType type) {
if (!check(type)) return 0;
advance();
return 1;
}
static int identifiersEqual(KrkToken * a, KrkToken * b) {
return (a->length == b->length && memcmp(a->start, b->start, a->length) == 0);
}
static KrkToken syntheticToken(const char * text) {
KrkToken token;
token.start = text;
token.length = (int)strlen(text);
return token;
}
static void emitByte(uint8_t byte) {
krk_writeChunk(currentChunk(), byte, parser.previous.line);
}
static void emitBytes(uint8_t byte1, uint8_t byte2) {
emitByte(byte1);
emitByte(byte2);
}
static void emitReturn() {
if (current->type == TYPE_INIT) {
emitBytes(OP_GET_LOCAL, 0);
2020-12-28 04:54:25 +03:00
} else if (current->type == TYPE_MODULE) {
/* Un-pop the last stack value */
emitBytes(OP_GET_LOCAL, 0);
} else {
emitByte(OP_NONE);
}
emitByte(OP_RETURN);
}
static KrkFunction * endCompiler() {
emitReturn();
KrkFunction * function = current->function;
/* Attach contants for arguments */
for (int i = 0; i < function->requiredArgs; ++i) {
KrkValue value = OBJECT_VAL(krk_copyString(current->locals[i].name.start, current->locals[i].name.length));
krk_push(value);
krk_writeValueArray(&function->requiredArgNames, value);
krk_pop();
}
for (int i = 0; i < function->keywordArgs; ++i) {
KrkValue value = OBJECT_VAL(krk_copyString(current->locals[i+function->requiredArgs].name.start,
current->locals[i+function->requiredArgs].name.length));
krk_push(value);
krk_writeValueArray(&function->keywordArgNames, value);
krk_pop();
}
size_t args = current->function->requiredArgs + current->function->keywordArgs;
if (current->function->collectsArguments) {
KrkValue value = OBJECT_VAL(krk_copyString(current->locals[args].name.start,
current->locals[args].name.length));
krk_push(value);
krk_writeValueArray(&function->keywordArgNames, value);
krk_pop();
args++;
}
if (current->function->collectsKeywords) {
KrkValue value = OBJECT_VAL(krk_copyString(current->locals[args].name.start,
current->locals[args].name.length));
krk_push(value);
krk_writeValueArray(&function->keywordArgNames, value);
krk_pop();
args++;
}
2021-01-02 06:21:11 +03:00
#ifdef ENABLE_DISASSEMBLY
if ((vm.flags & KRK_ENABLE_DISASSEMBLY) && !parser.hadError) {
krk_disassembleChunk(currentChunk(), function->name ? function->name->chars : "<module>");
fprintf(stderr, "Function metadata: requiredArgs=%d keywordArgs=%d upvalueCount=%d\n",
function->requiredArgs, function->keywordArgs, (int)function->upvalueCount);
fprintf(stderr, "__doc__: \"%s\"\n", function->docstring ? function->docstring->chars : "");
fprintf(stderr, "Constants: ");
for (size_t i = 0; i < currentChunk()->constants.count; ++i) {
fprintf(stderr, "%d: ", (int)i);
krk_printValueSafe(stderr, currentChunk()->constants.values[i]);
if (i != currentChunk()->constants.count - 1) {
fprintf(stderr, ", ");
}
}
fprintf(stderr, "\nRequired arguments: ");
int i = 0;
for (; i < function->requiredArgs; ++i) {
fprintf(stderr, "%.*s%s",
(int)current->locals[i].name.length,
current->locals[i].name.start,
(i == function->requiredArgs - 1) ? "" : ", ");
}
fprintf(stderr, "\nKeyword arguments: ");
for (; i < function->requiredArgs + function->keywordArgs; ++i) {
fprintf(stderr, "%.*s=None%s",
(int)current->locals[i].name.length,
current->locals[i].name.start,
(i == function->keywordArgs - 1) ? "" : ", ");
}
fprintf(stderr, "\n");
}
#endif
current = current->enclosing;
return function;
}
static void freeCompiler(Compiler * compiler) {
FREE_ARRAY(Local,compiler->locals, compiler->localsSpace);
FREE_ARRAY(Upvalue,compiler->upvalues, compiler->upvaluesSpace);
2020-12-31 10:59:51 +03:00
FREE_ARRAY(int,compiler->breaks, compiler->breakSpace);
FREE_ARRAY(int,compiler->continues, compiler->continueSpace);
}
static size_t emitConstant(KrkValue value) {
return krk_writeConstant(currentChunk(), value, parser.previous.line);
}
2020-12-26 10:53:15 +03:00
static void number(int canAssign) {
const char * start = parser.previous.start;
int base = 10;
/* These special cases for hexadecimal, binary, octal values. */
if (start[0] == '0' && (start[1] == 'x' || start[1] == 'X')) {
base = 16;
start += 2;
} else if (start[0] == '0' && (start[1] == 'b' || start[1] == 'B')) {
base = 2;
start += 2;
} else if (start[0] == '0' && (start[1] == 'o' || start[1] == 'O')) {
base = 8;
start += 2;
}
/* If it wasn't a special base, it may be a floating point value. */
if (base == 10) {
for (size_t j = 0; j < parser.previous.length; ++j) {
if (parser.previous.start[j] == '.') {
double value = strtod(start, NULL);
emitConstant(FLOATING_VAL(value));
return;
}
}
}
/* If we got here, it's an integer of some sort. */
2020-12-30 05:26:07 +03:00
long value = strtol(start, NULL, base);
emitConstant(INTEGER_VAL(value));
}
2020-12-26 10:53:15 +03:00
static void binary(int canAssign) {
KrkTokenType operatorType = parser.previous.type;
ParseRule * rule = getRule(operatorType);
parsePrecedence((Precedence)(rule->precedence + 1));
switch (operatorType) {
case TOKEN_BANG_EQUAL: emitBytes(OP_EQUAL, OP_NOT); break;
case TOKEN_EQUAL_EQUAL: emitByte(OP_EQUAL); break;
case TOKEN_GREATER: emitByte(OP_GREATER); break;
case TOKEN_GREATER_EQUAL: emitBytes(OP_LESS, OP_NOT); break;
case TOKEN_LESS: emitByte(OP_LESS); break;
case TOKEN_LESS_EQUAL: emitBytes(OP_GREATER, OP_NOT); break;
2021-01-01 02:55:39 +03:00
case TOKEN_PIPE: emitByte(OP_BITOR); break;
case TOKEN_CARET: emitByte(OP_BITXOR); break;
case TOKEN_AMPERSAND: emitByte(OP_BITAND); break;
case TOKEN_LEFT_SHIFT: emitByte(OP_SHIFTLEFT); break;
case TOKEN_RIGHT_SHIFT: emitByte(OP_SHIFTRIGHT); break;
case TOKEN_PLUS: emitByte(OP_ADD); break;
case TOKEN_MINUS: emitByte(OP_SUBTRACT); break;
case TOKEN_ASTERISK: emitByte(OP_MULTIPLY); break;
case TOKEN_SOLIDUS: emitByte(OP_DIVIDE); break;
2020-12-28 10:32:27 +03:00
case TOKEN_MODULO: emitByte(OP_MODULO); break;
case TOKEN_IN: emitByte(OP_EQUAL); break;
default: return;
}
}
static int matchAssignment(void) {
return match(TOKEN_EQUAL) || match(TOKEN_PLUS_EQUAL) || match(TOKEN_MINUS_EQUAL) ||
match(TOKEN_PLUS_PLUS) || match(TOKEN_MINUS_MINUS);
}
static void assignmentValue(void) {
switch (parser.previous.type) {
case TOKEN_PLUS_EQUAL:
expression();
emitByte(OP_ADD);
break;
case TOKEN_MINUS_EQUAL:
expression();
emitByte(OP_SUBTRACT);
break;
case TOKEN_PLUS_PLUS:
emitConstant(INTEGER_VAL(1));
emitByte(OP_ADD);
break;
case TOKEN_MINUS_MINUS:
emitConstant(INTEGER_VAL(1));
emitByte(OP_SUBTRACT);
break;
default:
error("Unexpected operand in assignment");
break;
}
}
static void get_(int canAssign) {
int isSlice = 0;
if (match(TOKEN_COLON)) {
emitByte(OP_NONE);
isSlice = 1;
} else {
expression();
}
if (isSlice || match(TOKEN_COLON)) {
if (isSlice && match(TOKEN_COLON)) {
error("Step value not supported in slice.");
return;
}
if (match(TOKEN_RIGHT_SQUARE)) {
emitByte(OP_NONE);
} else {
expression();
consume(TOKEN_RIGHT_SQUARE, "Expected ending square bracket after slice.");
}
if (canAssign && matchAssignment()) {
error("Can not assign to slice.");
} else {
emitByte(OP_INVOKE_GETSLICE);
}
} else {
consume(TOKEN_RIGHT_SQUARE, "Expected ending square bracket after index.");
if (canAssign && match(TOKEN_EQUAL)) {
expression();
emitByte(OP_INVOKE_SETTER);
} else if (canAssign && matchAssignment()) {
emitByte(OP_SWAP);
emitBytes(OP_DUP, 1);
emitByte(OP_INVOKE_GETTER);
assignmentValue();
emitByte(OP_INVOKE_SETTER);
} else {
emitByte(OP_INVOKE_GETTER);
}
}
}
2020-12-27 10:45:34 +03:00
static void dot(int canAssign) {
consume(TOKEN_IDENTIFIER, "Expected property name");
2020-12-27 10:45:34 +03:00
size_t ind = identifierConstant(&parser.previous);
if (canAssign && match(TOKEN_EQUAL)) {
expression();
EMIT_CONSTANT_OP(OP_SET_PROPERTY, ind);
} else if (canAssign && matchAssignment()) {
emitBytes(OP_DUP, 0); /* Duplicate the object */
EMIT_CONSTANT_OP(OP_GET_PROPERTY, ind);
assignmentValue();
EMIT_CONSTANT_OP(OP_SET_PROPERTY, ind);
2020-12-27 10:45:34 +03:00
} else {
EMIT_CONSTANT_OP(OP_GET_PROPERTY, ind);
}
}
static void in_(int canAssign) {
parsePrecedence(PREC_COMPARISON);
KrkToken contains = syntheticToken("__contains__");
ssize_t ind = identifierConstant(&contains);
EMIT_CONSTANT_OP(OP_GET_PROPERTY, ind);
emitByte(OP_SWAP);
emitBytes(OP_CALL,1);
}
2020-12-26 10:53:15 +03:00
static void literal(int canAssign) {
switch (parser.previous.type) {
case TOKEN_FALSE: emitByte(OP_FALSE); break;
case TOKEN_NONE: emitByte(OP_NONE); break;
case TOKEN_TRUE: emitByte(OP_TRUE); break;
default: return;
}
}
static void expression() {
parsePrecedence(PREC_ASSIGNMENT);
}
2020-12-26 10:53:15 +03:00
static void varDeclaration() {
ssize_t ind = parseVariable("Expected variable name.");
2020-12-26 10:53:15 +03:00
if (match(TOKEN_EQUAL)) {
expression();
} else {
emitByte(OP_NONE);
}
defineVariable(ind);
2020-12-26 10:53:15 +03:00
}
static void printStatement() {
int argCount = 0;
do {
expression();
argCount++;
} while (match(TOKEN_COMMA));
EMIT_CONSTANT_OP(OP_PRINT, argCount);
2020-12-26 10:53:15 +03:00
}
static void synchronize() {
parser.panicMode = 0;
while (parser.current.type != TOKEN_EOF) {
if (parser.previous.type == TOKEN_EOL) return;
switch (parser.current.type) {
case TOKEN_CLASS:
case TOKEN_DEF:
case TOKEN_LET:
case TOKEN_FOR:
case TOKEN_IF:
case TOKEN_WHILE:
case TOKEN_PRINT:
case TOKEN_RETURN:
return;
default: break;
}
advance();
}
}
static void declaration() {
if (check(TOKEN_DEF)) {
defDeclaration();
} else if (match(TOKEN_LET)) {
2020-12-26 10:53:15 +03:00
varDeclaration();
if (!match(TOKEN_EOL) && !match(TOKEN_EOF)) {
error("Expected EOL after variable declaration.\n");
}
2020-12-27 10:45:34 +03:00
} else if (check(TOKEN_CLASS)) {
classDeclaration();
2021-01-02 15:58:24 +03:00
} else if (check(TOKEN_AT)) {
decorator(0, TYPE_FUNCTION);
} else if (match(TOKEN_EOL) || match(TOKEN_EOF)) {
2020-12-26 14:39:47 +03:00
return;
} else if (check(TOKEN_INDENTATION)) {
return;
2020-12-26 10:53:15 +03:00
} else {
statement();
}
if (parser.panicMode) synchronize();
}
static void expressionStatement() {
expression();
emitByte(OP_POP);
}
static void beginScope() {
current->scopeDepth++;
}
static void endScope() {
current->scopeDepth--;
while (current->localCount > 0 &&
2020-12-28 05:11:50 +03:00
current->locals[current->localCount - 1].depth > (ssize_t)current->scopeDepth) {
if (current->locals[current->localCount - 1].isCaptured) {
emitByte(OP_CLOSE_UPVALUE);
} else {
emitByte(OP_POP);
}
current->localCount--;
}
}
static int emitJump(uint8_t opcode) {
emitByte(opcode);
emitBytes(0xFF, 0xFF);
return currentChunk()->count - 2;
}
static void patchJump(int offset) {
int jump = currentChunk()->count - offset - 2;
if (jump > 0xFFFF) {
error("Unsupported far jump (we'll get there)");
}
currentChunk()->code[offset] = (jump >> 8) & 0xFF;
currentChunk()->code[offset + 1] = (jump) & 0xFF;
}
static void block(size_t indentation, const char * blockName) {
if (match(TOKEN_EOL)) {
if (check(TOKEN_INDENTATION)) {
size_t currentIndentation = parser.current.length;
2020-12-30 10:59:21 +03:00
if (currentIndentation <= indentation) return;
advance();
if (!strcmp(blockName,"def") && match(TOKEN_STRING)) {
size_t before = currentChunk()->count;
string(0);
/* That wrote to the chunk, rewind it; this should only ever go back two bytes
* because this should only happen as the first thing in a function definition,
* and thus this _should_ be the first constant and thus opcode + one-byte operand
* to OP_CONSTANT, but just to be safe we'll actually use the previous offset... */
currentChunk()->count = before;
/* Retreive the docstring from the constant table */
current->function->docstring = AS_STRING(currentChunk()->constants.values[currentChunk()->constants.count-1]);
consume(TOKEN_EOL,"Garbage after docstring defintion");
if (!check(TOKEN_INDENTATION) || parser.current.length != currentIndentation) {
error("Expected at least one statement in function with docstring.");
}
advance();
2020-12-26 14:39:47 +03:00
}
2020-12-30 10:59:21 +03:00
declaration();
while (check(TOKEN_INDENTATION)) {
2020-12-28 06:16:44 +03:00
if (parser.current.length < currentIndentation) break;
2020-12-30 10:59:21 +03:00
advance();
declaration();
2020-12-30 10:59:21 +03:00
};
2021-01-02 06:21:11 +03:00
#ifdef ENABLE_SCAN_TRACING
if (vm.flags & KRK_ENABLE_SCAN_TRACING) {
fprintf(stderr, "finished with block %s (ind=%d) on line %d, sitting on a %s (len=%d)\n",
2020-12-30 10:59:21 +03:00
blockName, (int)indentation, (int)parser.current.line,
getRule(parser.current.type)->name, (int)parser.current.length);
2020-12-28 07:35:02 +03:00
}
#endif
}
} else {
statement();
}
}
2020-12-28 05:11:50 +03:00
static void function(FunctionType type, size_t blockWidth) {
Compiler compiler;
initCompiler(&compiler, type);
2020-12-28 04:54:25 +03:00
compiler.function->chunk.filename = compiler.enclosing->function->chunk.filename;
beginScope();
if (type == TYPE_METHOD || type == TYPE_INIT) current->function->requiredArgs = 1;
int hasCollectors = 0;
consume(TOKEN_LEFT_PAREN, "Expected start of parameter list after function name.");
if (!check(TOKEN_RIGHT_PAREN)) {
do {
if (match(TOKEN_SELF)) {
if (type != TYPE_INIT && type != TYPE_METHOD) {
error("Invalid use of `self` as a function paramenter.");
}
continue;
}
if (match(TOKEN_ASTERISK)) {
if (match(TOKEN_ASTERISK)) {
if (hasCollectors == 2) {
error("Duplicate ** in parameter list.");
return;
}
hasCollectors = 2;
current->function->collectsKeywords = 1;
} else {
if (hasCollectors) {
error("Syntax error.");
return;
}
hasCollectors = 1;
current->function->collectsArguments = 1;
}
/* Collect a name, specifically "args" or "kwargs" are commont */
ssize_t paramConstant = parseVariable("Expect parameter name.");
defineVariable(paramConstant);
/* Make that a valid local for this function */
size_t myLocal = current->localCount - 1;
EMIT_CONSTANT_OP(OP_GET_LOCAL, myLocal);
/* Check if it's equal to the unset-kwarg-sentinel value */
emitConstant(KWARGS_VAL(0));
emitByte(OP_EQUAL);
int jumpIndex = emitJump(OP_JUMP_IF_FALSE);
/* And if it is, set it to the appropriate type */
beginScope();
KrkToken synth = syntheticToken(hasCollectors == 1 ? "listOf" : "dictOf");
namedVariable(synth, 0);
emitBytes(OP_CALL, 0);
EMIT_CONSTANT_OP(OP_SET_LOCAL, myLocal);
endScope();
/* Otherwise pop the comparison. */
patchJump(jumpIndex);
emitByte(OP_POP); /* comparison value */
continue;
}
ssize_t paramConstant = parseVariable("Expect parameter name.");
defineVariable(paramConstant);
if (match(TOKEN_EQUAL)) {
/*
* We inline default arguments by checking if they are equal
* to a sentinel value and replacing them with the requested
* argument. This allows us to send None (useful) to override
* defaults that are something else. This essentially ends
* up as the following at the top of the function:
* if param == KWARGS_SENTINEL:
* param = EXPRESSION
*/
size_t myLocal = current->localCount - 1;
EMIT_CONSTANT_OP(OP_GET_LOCAL, myLocal);
emitConstant(KWARGS_VAL(0));
emitByte(OP_EQUAL);
int jumpIndex = emitJump(OP_JUMP_IF_FALSE);
beginScope();
expression(); /* Read expression */
EMIT_CONSTANT_OP(OP_SET_LOCAL, myLocal);
endScope();
patchJump(jumpIndex);
emitByte(OP_POP);
current->function->keywordArgs++;
} else {
current->function->requiredArgs++;
}
} while (match(TOKEN_COMMA));
}
consume(TOKEN_RIGHT_PAREN, "Expected end of parameter list.");
consume(TOKEN_COLON, "Expected colon after function signature.");
block(blockWidth,"def");
KrkFunction * function = endCompiler();
size_t ind = krk_addConstant(currentChunk(), OBJECT_VAL(function));
EMIT_CONSTANT_OP(OP_CLOSURE, ind);
for (size_t i = 0; i < function->upvalueCount; ++i) {
/* TODO: if the maximum count changes, fix the sizes for this */
emitByte(compiler.upvalues[i].isLocal ? 1 : 0);
if (i > 255) {
emitByte((compiler.upvalues[i].index >> 16) & 0xFF);
emitByte((compiler.upvalues[i].index >> 8) & 0xFF);
}
emitByte((compiler.upvalues[i].index) & 0xFF);
}
freeCompiler(&compiler);
}
2020-12-28 05:11:50 +03:00
static void method(size_t blockWidth) {
/* This is actually "inside of a class definition", and that might mean
* arbitrary blank lines we need to accept... Sorry. */
if (match(TOKEN_EOL)) {
return;
}
/* def method(...): - just like functions; unlike Python, I'm just always
* going to assign `self` because Lox always assigns `this`; it should not
* show up in the initializer list; I may add support for it being there
* as a redundant thing, just to make more Python stuff work with changes. */
if (check(TOKEN_AT)) {
decorator(0, TYPE_METHOD);
} else {
consume(TOKEN_DEF, "expected a definition, got nothing");
consume(TOKEN_IDENTIFIER, "expected method name");
size_t ind = identifierConstant(&parser.previous);
FunctionType type = TYPE_METHOD;
if (parser.previous.length == 8 && memcmp(parser.previous.start, "__init__", 8) == 0) {
type = TYPE_INIT;
}
function(type, blockWidth);
EMIT_CONSTANT_OP(OP_METHOD, ind);
}
}
2020-12-27 10:45:34 +03:00
static void classDeclaration() {
2020-12-28 05:11:50 +03:00
size_t blockWidth = (parser.previous.type == TOKEN_INDENTATION) ? parser.previous.length : 0;
2020-12-27 10:45:34 +03:00
advance(); /* Collect the `class` */
consume(TOKEN_IDENTIFIER, "Expected class name.");
KrkToken className = parser.previous;
2020-12-27 10:45:34 +03:00
size_t constInd = identifierConstant(&parser.previous);
declareVariable();
EMIT_CONSTANT_OP(OP_CLASS, constInd);
defineVariable(constInd);
ClassCompiler classCompiler;
classCompiler.name = parser.previous;
classCompiler.hasSuperClass = 0;
classCompiler.enclosing = currentClass;
currentClass = &classCompiler;
if (match(TOKEN_LEFT_PAREN)) {
if (match(TOKEN_IDENTIFIER)) {
variable(0);
if (identifiersEqual(&className, &parser.previous)) {
error("A class can not inherit from itself.");
}
beginScope();
addLocal(syntheticToken("super"));
defineVariable(0);
namedVariable(className, 0);
emitByte(OP_INHERIT);
classCompiler.hasSuperClass = 1;
}
consume(TOKEN_RIGHT_PAREN, "Expected closing brace after superclass.");
}
namedVariable(className, 0);
2020-12-27 10:45:34 +03:00
consume(TOKEN_COLON, "Expected colon after class");
if (match(TOKEN_EOL)) {
if (check(TOKEN_INDENTATION)) {
size_t currentIndentation = parser.current.length;
if (currentIndentation <= blockWidth) {
errorAtCurrent("Unexpected indentation level for class");
}
2020-12-30 10:59:21 +03:00
advance();
if (match(TOKEN_STRING)) {
string(0);
emitByte(OP_DOCSTRING);
consume(TOKEN_EOL,"Garbage after docstring defintion");
if (!check(TOKEN_INDENTATION) || parser.current.length != currentIndentation) {
goto _pop_class;
}
advance();
}
method(currentIndentation);
while (check(TOKEN_INDENTATION)) {
2020-12-28 06:16:44 +03:00
if (parser.current.length < currentIndentation) break;
advance(); /* Pass the indentation */
method(currentIndentation);
2020-12-30 10:59:21 +03:00
}
#ifdef ENABLE_SCAN_TRACING
if (vm.flags & KRK_ENABLE_SCAN_TRACING) fprintf(stderr, "Exiting from class definition on %s\n", getRule(parser.current.type)->name);
#endif
/* Exit from block */
}
} /* else empty class (and at end of file?) we'll allow it for now... */
2020-12-30 10:59:21 +03:00
_pop_class:
emitByte(OP_POP);
if (classCompiler.hasSuperClass) {
endScope();
}
currentClass = currentClass->enclosing;
2020-12-27 10:45:34 +03:00
}
static void markInitialized() {
if (current->scopeDepth == 0) return;
current->locals[current->localCount - 1].depth = current->scopeDepth;
}
static void defDeclaration() {
2020-12-28 05:11:50 +03:00
size_t blockWidth = (parser.previous.type == TOKEN_INDENTATION) ? parser.previous.length : 0;
advance(); /* Collect the `def` */
ssize_t global = parseVariable("Expected function name.");
markInitialized();
function(TYPE_FUNCTION, blockWidth);
defineVariable(global);
}
2021-01-02 15:58:24 +03:00
static KrkToken decorator(size_t level, FunctionType type) {
size_t blockWidth = (parser.previous.type == TOKEN_INDENTATION) ? parser.previous.length : 0;
advance(); /* Collect the `@` */
/* Collect an identifier */
expression();
2021-01-02 15:58:24 +03:00
consume(TOKEN_EOL, "Expected line feed after decorator.");
if (blockWidth) {
consume(TOKEN_INDENTATION, "Expected next line after decorator to have same indentation.");
if (parser.previous.length != blockWidth) error("Expected next line after decorator to have same indentation.");
}
KrkToken funcName;
if (check(TOKEN_DEF)) {
/* We already checked for block level */
advance();
consume(TOKEN_IDENTIFIER, "Expected function name.");
funcName = parser.previous;
if (type == TYPE_METHOD && funcName.length == 8 && !memcmp(funcName.start,"__init__",8)) {
type = TYPE_INIT;
}
function(type, blockWidth);
} else if (check(TOKEN_AT)) {
funcName = decorator(level+1, type);
} else {
error("Expected a function declaration or another decorator.");
}
emitBytes(OP_CALL, 1);
2021-01-02 15:58:24 +03:00
if (level == 0) {
if (type == TYPE_FUNCTION) {
parser.previous = funcName;
declareVariable();
size_t ind = (current->scopeDepth > 0) ? 0 : identifierConstant(&funcName);
defineVariable(ind);
} else {
size_t ind = identifierConstant(&funcName);
EMIT_CONSTANT_OP(OP_METHOD, ind);
}
}
return funcName;
}
2020-12-26 14:39:47 +03:00
static void emitLoop(int loopStart) {
2020-12-31 10:59:51 +03:00
/* Patch continue statements to point to here, before the loop operation (yes that's silly) */
while (current->continueCount > 0 && current->continues[current->continueCount-1] > loopStart) {
patchJump(current->continues[current->continueCount-1]);
current->continueCount--;
}
2020-12-26 14:39:47 +03:00
emitByte(OP_LOOP);
2020-12-31 10:59:51 +03:00
2020-12-26 14:39:47 +03:00
int offset = currentChunk()->count - loopStart + 2;
if (offset > 0xFFFF) error("offset too big");
emitBytes(offset >> 8, offset);
2020-12-31 10:59:51 +03:00
/* Patch break statements */
2020-12-26 14:39:47 +03:00
}
static void ifStatement() {
/* Figure out what block level contains us so we can match our partner else */
2020-12-28 05:11:50 +03:00
size_t blockWidth = (parser.previous.type == TOKEN_INDENTATION) ? parser.previous.length : 0;
KrkToken myPrevious = parser.previous;
2020-12-26 14:39:47 +03:00
/* Collect the if token that started this statement */
advance();
/* Collect condition expression */
expression();
/* if EXPR: */
consume(TOKEN_COLON, "Expect ':' after condition.");
int thenJump = emitJump(OP_JUMP_IF_FALSE);
emitByte(OP_POP);
/* Start a new scope and enter a block */
beginScope();
block(blockWidth,"if");
2020-12-26 14:39:47 +03:00
endScope();
int elseJump = emitJump(OP_JUMP);
patchJump(thenJump);
emitByte(OP_POP);
/* See if we have a matching else block */
if (blockWidth == 0 || (check(TOKEN_INDENTATION) && (parser.current.length == blockWidth))) {
2020-12-28 06:16:44 +03:00
/* This is complicated */
KrkToken previous;
if (blockWidth) {
previous = parser.previous;
advance();
}
2020-12-26 14:39:47 +03:00
if (match(TOKEN_ELSE)) {
if (check(TOKEN_IF)) {
parser.previous = myPrevious;
ifStatement(); /* Keep nesting */
} else {
consume(TOKEN_COLON, "Expect ':' after else.");
beginScope();
block(blockWidth,"else");
endScope();
}
} else {
if (!check(TOKEN_EOF) && !check(TOKEN_EOL)) {
krk_ungetToken(parser.current);
parser.current = parser.previous;
if (blockWidth) {
parser.previous = previous;
}
2020-12-28 06:16:44 +03:00
}
2020-12-26 14:39:47 +03:00
}
}
patchJump(elseJump);
}
2020-12-31 10:59:51 +03:00
static void patchBreaks(int loopStart) {
/* Patch break statements to go here, after the loop operation and operand. */
while (current->breakCount > 0 && current->breaks[current->breakCount-1] > loopStart) {
patchJump(current->breaks[current->breakCount-1]);
current->breakCount--;
}
}
static void breakStatement() {
if (current->breakSpace < current->breakCount + 1) {
size_t old = current->breakSpace;
current->breakSpace = GROW_CAPACITY(old);
current->breaks = GROW_ARRAY(int,current->breaks,old,current->breakSpace);
}
for (size_t i = current->loopLocalCount; i < current->localCount; ++i) {
emitByte(OP_POP);
}
current->breaks[current->breakCount++] = emitJump(OP_JUMP);
}
static void continueStatement() {
if (current->continueSpace < current->continueCount + 1) {
size_t old = current->continueSpace;
current->continueSpace = GROW_CAPACITY(old);
current->continues = GROW_ARRAY(int,current->continues,old,current->continueSpace);
}
for (size_t i = current->loopLocalCount; i < current->localCount; ++i) {
emitByte(OP_POP);
}
current->continues[current->continueCount++] = emitJump(OP_JUMP);
}
2020-12-26 14:39:47 +03:00
static void whileStatement() {
2020-12-28 05:11:50 +03:00
size_t blockWidth = (parser.previous.type == TOKEN_INDENTATION) ? parser.previous.length : 0;
2020-12-26 14:39:47 +03:00
advance();
int loopStart = currentChunk()->count;
expression();
consume(TOKEN_COLON, "Expect ':' after condition.");
int exitJump = emitJump(OP_JUMP_IF_FALSE);
emitByte(OP_POP);
2020-12-31 10:59:51 +03:00
int oldLocalCount = current->loopLocalCount;
current->loopLocalCount = current->localCount;
2020-12-26 14:39:47 +03:00
beginScope();
block(blockWidth,"while");
2020-12-26 14:39:47 +03:00
endScope();
2020-12-31 10:59:51 +03:00
current->loopLocalCount = oldLocalCount;
2020-12-26 14:39:47 +03:00
emitLoop(loopStart);
patchJump(exitJump);
emitByte(OP_POP);
2020-12-31 10:59:51 +03:00
patchBreaks(loopStart);
2020-12-26 14:39:47 +03:00
}
static void forStatement() {
/* I'm not sure if I want this to be more like Python or C/Lox/etc. */
2020-12-28 05:11:50 +03:00
size_t blockWidth = (parser.previous.type == TOKEN_INDENTATION) ? parser.previous.length : 0;
2020-12-26 14:39:47 +03:00
advance();
/* For now this is going to be kinda broken */
beginScope();
2020-12-28 07:35:02 +03:00
ssize_t loopInd = current->localCount;
2020-12-26 14:39:47 +03:00
varDeclaration();
2020-12-28 07:35:02 +03:00
int loopStart;
int exitJump;
2020-12-26 14:39:47 +03:00
2020-12-28 07:35:02 +03:00
if (match(TOKEN_IN)) {
defineVariable(loopInd);
2020-12-26 14:39:47 +03:00
/* ITERABLE.__iter__() */
beginScope();
expression();
endScope();
2020-12-28 07:35:02 +03:00
KrkToken _it = syntheticToken("__loop_iter");
2020-12-28 05:11:50 +03:00
size_t indLoopIter = current->localCount;
2020-12-28 07:35:02 +03:00
addLocal(_it);
defineVariable(indLoopIter);
KrkToken _iter = syntheticToken("__iter__");
2020-12-28 07:35:02 +03:00
ssize_t ind = identifierConstant(&_iter);
EMIT_CONSTANT_OP(OP_GET_PROPERTY, ind);
emitBytes(OP_CALL, 0);
/* assign */
EMIT_CONSTANT_OP(OP_SET_LOCAL, indLoopIter);
/* LOOP STARTS HERE */
loopStart = currentChunk()->count;
/* Call the iterator */
EMIT_CONSTANT_OP(OP_GET_LOCAL, indLoopIter);
emitBytes(OP_CALL, 0);
/* Assign the result to our loop index */
EMIT_CONSTANT_OP(OP_SET_LOCAL, loopInd);
/* Get the loop iterator again */
EMIT_CONSTANT_OP(OP_GET_LOCAL, indLoopIter);
emitBytes(OP_EQUAL, OP_NOT);
2020-12-28 07:35:02 +03:00
exitJump = emitJump(OP_JUMP_IF_FALSE);
emitByte(OP_POP);
} else {
consume(TOKEN_COMMA,"expect ,");
loopStart = currentChunk()->count;
beginScope();
2020-12-28 07:35:02 +03:00
expression(); /* condition */
endScope();
2020-12-28 07:35:02 +03:00
exitJump = emitJump(OP_JUMP_IF_FALSE);
2020-12-26 14:39:47 +03:00
emitByte(OP_POP);
2020-12-28 07:35:02 +03:00
if (check(TOKEN_COMMA)) {
advance();
int bodyJump = emitJump(OP_JUMP);
int incrementStart = currentChunk()->count;
beginScope();
2020-12-28 07:35:02 +03:00
expression();
endScope();
2020-12-28 07:35:02 +03:00
emitByte(OP_POP);
emitLoop(loopStart);
loopStart = incrementStart;
patchJump(bodyJump);
}
2020-12-26 14:39:47 +03:00
}
consume(TOKEN_COLON,"expect :");
2020-12-31 10:59:51 +03:00
int oldLocalCount = current->loopLocalCount;
current->loopLocalCount = current->localCount;
beginScope();
block(blockWidth,"for");
endScope();
2020-12-26 14:39:47 +03:00
2020-12-31 10:59:51 +03:00
current->loopLocalCount = oldLocalCount;
2020-12-26 14:39:47 +03:00
emitLoop(loopStart);
patchJump(exitJump);
emitByte(OP_POP);
2020-12-31 10:59:51 +03:00
patchBreaks(loopStart);
2020-12-26 14:39:47 +03:00
endScope();
}
static void returnStatement() {
if (check(TOKEN_EOL) || check(TOKEN_EOF)) {
emitReturn();
} else {
if (current->type == TYPE_INIT) {
error("Can not return values from __init__");
}
expression();
emitByte(OP_RETURN);
}
}
2020-12-29 05:00:12 +03:00
static void tryStatement() {
size_t blockWidth = (parser.previous.type == TOKEN_INDENTATION) ? parser.previous.length : 0;
advance();
consume(TOKEN_COLON, "Expect ':' after try.");
/* Make sure we are in a local scope so this ends up on the stack */
beginScope();
int tryJump = emitJump(OP_PUSH_TRY);
addLocal(syntheticToken("exception"));
defineVariable(0);
beginScope();
block(blockWidth,"try");
2020-12-29 05:00:12 +03:00
endScope();
int successJump = emitJump(OP_JUMP);
patchJump(tryJump);
if (blockWidth == 0 || (check(TOKEN_INDENTATION) && (parser.current.length == blockWidth))) {
KrkToken previous;
if (blockWidth) {
previous = parser.previous;
advance();
}
if (match(TOKEN_EXCEPT)) {
consume(TOKEN_COLON, "Expect ':' after except.");
beginScope();
block(blockWidth,"except");
2020-12-29 05:00:12 +03:00
endScope();
} else if (!check(TOKEN_EOL) && !check(TOKEN_EOF)) {
2020-12-29 05:00:12 +03:00
krk_ungetToken(parser.current);
parser.current = parser.previous;
if (blockWidth) {
parser.previous = previous;
}
}
}
patchJump(successJump);
endScope(); /* will pop the exception handler */
}
static void raiseStatement() {
expression();
emitByte(OP_RAISE);
}
static void importStatement() {
consume(TOKEN_IDENTIFIER, "Expected module name");
size_t ind = identifierConstant(&parser.previous);
EMIT_CONSTANT_OP(OP_IMPORT, ind);
2021-01-04 01:39:00 +03:00
if (match(TOKEN_AS)) {
consume(TOKEN_IDENTIFIER, "Expected identifier after `as`");
ind = identifierConstant(&parser.previous);
}
declareVariable();
defineVariable(ind);
}
2020-12-28 13:25:33 +03:00
static void exportStatement() {
do {
consume(TOKEN_IDENTIFIER, "only named variable may be exported to the global namespace");
namedVariable(parser.previous, 0);
namedVariable(parser.previous, 0);
size_t ind = identifierConstant(&parser.previous);
EMIT_CONSTANT_OP(OP_DEFINE_GLOBAL, ind);
EMIT_CONSTANT_OP(OP_SET_GLOBAL, ind);
emitByte(OP_POP);
} while (match(TOKEN_COMMA));
2020-12-28 13:25:33 +03:00
}
2020-12-26 10:53:15 +03:00
static void statement() {
if (match(TOKEN_EOL) || match(TOKEN_EOF)) {
2020-12-26 14:39:47 +03:00
return; /* Meaningless blank line */
}
if (check(TOKEN_IF)) {
2020-12-26 14:39:47 +03:00
ifStatement();
} else if (check(TOKEN_WHILE)) {
whileStatement();
} else if (check(TOKEN_FOR)) {
forStatement();
2020-12-29 05:00:12 +03:00
} else if (check(TOKEN_TRY)) {
tryStatement();
2020-12-26 10:53:15 +03:00
} else {
if (match(TOKEN_PRINT)) {
printStatement();
} else if (match(TOKEN_EXPORT)) {
exportStatement();
} else if (match(TOKEN_RAISE)) {
raiseStatement();
} else if (match(TOKEN_RETURN)) {
returnStatement();
} else if (match(TOKEN_IMPORT)) {
importStatement();
2020-12-31 10:59:51 +03:00
} else if (match(TOKEN_BREAK)) {
breakStatement();
} else if (match(TOKEN_CONTINUE)) {
continueStatement();
} else {
expressionStatement();
}
if (!match(TOKEN_EOL) && !match(TOKEN_EOF)) {
errorAtCurrent("Unexpected token after statement.");
}
2020-12-26 10:53:15 +03:00
}
}
static void grouping(int canAssign) {
expression();
consume(TOKEN_RIGHT_PAREN, "Expect ')' after expression.");
}
2020-12-26 10:53:15 +03:00
static void unary(int canAssign) {
KrkTokenType operatorType = parser.previous.type;
parsePrecedence(PREC_UNARY);
switch (operatorType) {
case TOKEN_MINUS: emitByte(OP_NEGATE); break;
2021-01-01 02:55:39 +03:00
case TOKEN_TILDE: emitByte(OP_BITNEGATE); break;
/* These are equivalent */
case TOKEN_BANG:
case TOKEN_NOT:
emitByte(OP_NOT);
break;
default: return;
}
}
static void string(int type) {
2020-12-29 10:19:17 +03:00
/* We'll just build with a flexible array like everything else. */
size_t stringCapacity = 0;
size_t stringLength = 0;
char * stringBytes = 0;
#define PUSH_CHAR(c) do { if (stringCapacity < stringLength + 1) { \
size_t old = stringCapacity; stringCapacity = GROW_CAPACITY(old); \
stringBytes = GROW_ARRAY(char, stringBytes, old, stringCapacity); \
} stringBytes[stringLength++] = c; } while (0)
/* This should capture everything but the quotes. */
do {
int type = parser.previous.type == TOKEN_BIG_STRING ? 3 : 1;
const char * c = parser.previous.start + type;
while (c < parser.previous.start + parser.previous.length - type) {
if (*c == '\\') {
switch (c[1]) {
case 'n': PUSH_CHAR('\n'); break;
case 'r': PUSH_CHAR('\r'); break;
case 't': PUSH_CHAR('\t'); break;
case '[': PUSH_CHAR('\033'); break;
case '\n': break;
default: PUSH_CHAR(c[1]); break;
}
c += 2;
} else {
PUSH_CHAR(*c);
c++;
2020-12-29 10:19:17 +03:00
}
}
} while (match(TOKEN_STRING) || match(TOKEN_BIG_STRING));
2020-12-29 10:19:17 +03:00
emitConstant(OBJECT_VAL(krk_copyString(stringBytes,stringLength)));
FREE_ARRAY(char,stringBytes,stringCapacity);
}
static size_t addUpvalue(Compiler * compiler, ssize_t index, int isLocal) {
size_t upvalueCount = compiler->function->upvalueCount;
for (size_t i = 0; i < upvalueCount; ++i) {
Upvalue * upvalue = &compiler->upvalues[i];
2020-12-28 05:11:50 +03:00
if ((ssize_t)upvalue->index == index && upvalue->isLocal == isLocal) {
return i;
}
}
if (upvalueCount + 1 > current->upvaluesSpace) {
size_t old = current->upvaluesSpace;
current->upvaluesSpace = GROW_CAPACITY(old);
current->upvalues = GROW_ARRAY(Upvalue,current->upvalues,old,current->upvaluesSpace);
}
compiler->upvalues[upvalueCount].isLocal = isLocal;
compiler->upvalues[upvalueCount].index = index;
return compiler->function->upvalueCount++;
}
static ssize_t resolveUpvalue(Compiler * compiler, KrkToken * name) {
if (compiler->enclosing == NULL) return -1;
ssize_t local = resolveLocal(compiler->enclosing, name);
if (local != -1) {
compiler->enclosing->locals[local].isCaptured = 1;
return addUpvalue(compiler, local, 1);
}
ssize_t upvalue = resolveUpvalue(compiler->enclosing, name);
if (upvalue != -1) {
return addUpvalue(compiler, upvalue, 0);
}
return -1;
}
#define DO_VARIABLE(opset,opget) do { \
if (canAssign && match(TOKEN_EQUAL)) { \
expression(); \
EMIT_CONSTANT_OP(opset, arg); \
} else if (canAssign && matchAssignment()) { \
EMIT_CONSTANT_OP(opget, arg); \
assignmentValue(); \
EMIT_CONSTANT_OP(opset, arg); \
} else { \
EMIT_CONSTANT_OP(opget, arg); \
} } while (0)
2020-12-26 10:53:15 +03:00
static void namedVariable(KrkToken name, int canAssign) {
ssize_t arg = resolveLocal(current, &name);
if (arg != -1) {
DO_VARIABLE(OP_SET_LOCAL, OP_GET_LOCAL);
} else if ((arg = resolveUpvalue(current, &name)) != -1) {
DO_VARIABLE(OP_SET_UPVALUE, OP_GET_UPVALUE);
2020-12-26 10:53:15 +03:00
} else {
arg = identifierConstant(&name);
DO_VARIABLE(OP_SET_GLOBAL, OP_GET_GLOBAL);
2020-12-26 10:53:15 +03:00
}
}
#undef DO_VARIABLE
2020-12-26 10:53:15 +03:00
static void variable(int canAssign) {
namedVariable(parser.previous, canAssign);
}
static void self(int canAssign) {
if (currentClass == NULL) {
error("Invalid reference to `self` outside of a class method.");
return;
}
variable(0);
}
static void super_(int canAssign) {
if (currentClass == NULL) {
error("Invalid reference to `super` outside of a class.");
} else if (!currentClass->hasSuperClass) {
error("Invalid reference to `super` from a base class.");
}
consume(TOKEN_LEFT_PAREN, "Expected `super` to be called.");
consume(TOKEN_RIGHT_PAREN, "`super` can not take arguments.");
consume(TOKEN_DOT, "Expected a field of `super()` to be referenced.");
consume(TOKEN_IDENTIFIER, "Expected a field name.");
size_t ind = identifierConstant(&parser.previous);
namedVariable(syntheticToken("self"), 0);
namedVariable(syntheticToken("super"), 0);
EMIT_CONSTANT_OP(OP_GET_SUPER, ind);
}
2020-12-30 03:01:52 +03:00
static void list(int canAssign) {
size_t chunkBefore = currentChunk()->count;
2020-12-30 03:01:52 +03:00
KrkToken listOf = syntheticToken("listOf");
size_t ind = identifierConstant(&listOf);
EMIT_CONSTANT_OP(OP_GET_GLOBAL, ind);
2020-12-30 03:01:52 +03:00
if (!check(TOKEN_RIGHT_SQUARE)) {
KrkScanner scannerBefore = krk_tellScanner();
Parser parserBefore = parser;
expression();
/* This is a bit complicated and the Pratt parser does not handle it
* well; if we read an expression and then saw a `for`, we need to back
* up and start over, as we'll need to define a variable _after_ it
* gets used in this expression; so we record the parser state before
* reading the first expression of a list constant. If it _is_ a real
* list constant, we'll see a comma next and we can begin the normal
* loop of counting arguments. */
if (match(TOKEN_FOR)) {
/* Roll back the earlier compiler */
currentChunk()->count = chunkBefore;
/* Compile list comprehension as a function */
Compiler subcompiler;
initCompiler(&subcompiler, current->type == TYPE_METHOD ? TYPE_METHOD : TYPE_FUNCTION);
subcompiler.function->chunk.filename = subcompiler.enclosing->function->chunk.filename;
beginScope();
/* for i=0, */
emitConstant(INTEGER_VAL(0));
size_t indLoopCounter = current->localCount;
addLocal(syntheticToken("__loop_count"));
defineVariable(indLoopCounter);
/* x in... */
ssize_t loopInd = current->localCount;
varDeclaration();
defineVariable(loopInd);
consume(TOKEN_IN, "Only iterator loops (for ... in ...) are allowed in list comprehensions.");
beginScope();
2020-12-30 03:01:52 +03:00
expression();
endScope();
/* iterable... */
size_t indLoopIter = current->localCount;
addLocal(syntheticToken("__loop_iter"));
defineVariable(indLoopIter);
/* Now try to call .__iter__ on the result to produce our iterator */
KrkToken _iter = syntheticToken("__iter__");
ssize_t ind = identifierConstant(&_iter);
EMIT_CONSTANT_OP(OP_GET_PROPERTY, ind);
emitBytes(OP_CALL, 0);
/* Assign the resulting iterator to indLoopIter */
EMIT_CONSTANT_OP(OP_SET_LOCAL, indLoopIter);
/* Mark the start of the loop */
int loopStart = currentChunk()->count;
/* Call the iterator to get a value for our list */
EMIT_CONSTANT_OP(OP_GET_LOCAL, indLoopIter);
emitBytes(OP_CALL, 0);
/* Assign the result to our loop index */
EMIT_CONSTANT_OP(OP_SET_LOCAL, loopInd);
/* Compare the iterator to the loop index;
* our iterators return themselves to say they are done;
* this allows them to return None without any issue,
* and there's no feasible way they can return themselves without
* our intended sentinel meaning, right? Surely? */
EMIT_CONSTANT_OP(OP_GET_LOCAL, indLoopIter);
emitBytes(OP_EQUAL, OP_NOT);
int exitJump = emitJump(OP_JUMP_IF_FALSE);
emitByte(OP_POP);
/* Now we can rewind the scanner to have it parse the original
* expression that uses our iterated values! */
KrkScanner scannerAfter = krk_tellScanner();
Parser parserAfter = parser;
krk_rewindScanner(scannerBefore);
parser = parserBefore;
beginScope();
expression();
endScope();
/* Then we can put the parser back to where it was at the end of
* the iterator expression and continue. */
krk_rewindScanner(scannerAfter);
parser = parserAfter;
/* We keep a counter so we can keep track of how many arguments
* are on the stack, which we need in order to find the listOf()
* method above; having run the expression and generated an
* item which is now on the stack, increment the counter */
EMIT_CONSTANT_OP(OP_INC, indLoopCounter);
/* ... and loop back to the iterator call. */
emitLoop(loopStart);
/* Finally, at this point, we've seen the iterator produce itself
* and we're done receiving objects, so mark this instruction
* offset as the exit target for the OP_JUMP_IF_FALSE above */
patchJump(exitJump);
/* Parse the ] that indicates the end of the list comprehension */
consume(TOKEN_RIGHT_SQUARE,"Expected ] at end of list expression.");
/* Pop the last loop expression result which was already stored */
emitByte(OP_POP);
/* Pull in listOf from the global namespace */
KrkToken listOf = syntheticToken("listOf");
size_t indList = identifierConstant(&listOf);
EMIT_CONSTANT_OP(OP_GET_GLOBAL, indList);
/* And move it into where we were storing the loop iterator */
EMIT_CONSTANT_OP(OP_SET_LOCAL, indLoopIter);
/* (And pop it from the top of the stack) */
emitByte(OP_POP);
/* Then get the counter for our arg count */
EMIT_CONSTANT_OP(OP_GET_LOCAL, indLoopCounter);
/* And then call the native method which should be ^ that many items down */
emitByte(OP_CALL_STACK);
/* And return the result back to the original scope */
emitByte(OP_RETURN);
/* Now because we made a function we need to fill out its upvalues
* and write the closure call for it. */
KrkFunction *subfunction = endCompiler();
size_t indFunc = krk_addConstant(currentChunk(), OBJECT_VAL(subfunction));
EMIT_CONSTANT_OP(OP_CLOSURE, indFunc);
for (size_t i = 0; i < subfunction->upvalueCount; ++i) {
emitByte(subcompiler.upvalues[i].isLocal ? 1 : 0);
if (i > 255) {
emitByte((subcompiler.upvalues[i].index >> 16) & 0xFF);
emitByte((subcompiler.upvalues[i].index >> 8) & 0xFF);
}
emitByte((subcompiler.upvalues[i].index) & 0xFF);
}
freeCompiler(&subcompiler);
/* And finally we can call the subfunction and get the result. */
emitBytes(OP_CALL, 0);
} else {
size_t argCount = 1;
while (match(TOKEN_COMMA)) {
expression();
argCount++;
}
consume(TOKEN_RIGHT_SQUARE,"Expected ] at end of list expression.");
2021-01-02 06:51:32 +03:00
EMIT_CONSTANT_OP(OP_CALL, argCount);
}
} else {
/* Empty list expression */
advance();
emitBytes(OP_CALL, 0);
2020-12-30 03:01:52 +03:00
}
}
static void dict(int canAssign) {
KrkToken dictOf = syntheticToken("dictOf");
size_t ind = identifierConstant(&dictOf);
EMIT_CONSTANT_OP(OP_GET_GLOBAL, ind);
size_t argCount = 0;
if (!check(TOKEN_RIGHT_BRACE)) {
do {
expression();
consume(TOKEN_COLON, "Expect colon after dict key.");
expression();
argCount += 2;
} while (match(TOKEN_COMMA));
}
consume(TOKEN_RIGHT_BRACE,"Expected } at end of dict expression.");
2021-01-02 06:51:32 +03:00
EMIT_CONSTANT_OP(OP_CALL, argCount);
2020-12-30 03:01:52 +03:00
}
#define RULE(token, a, b, c) [token] = {# token, a, b, c}
ParseRule rules[] = {
RULE(TOKEN_LEFT_PAREN, grouping, call, PREC_CALL),
RULE(TOKEN_RIGHT_PAREN, NULL, NULL, PREC_NONE),
2020-12-30 03:01:52 +03:00
RULE(TOKEN_LEFT_BRACE, dict, NULL, PREC_NONE),
RULE(TOKEN_RIGHT_BRACE, NULL, NULL, PREC_NONE),
2020-12-30 03:01:52 +03:00
RULE(TOKEN_LEFT_SQUARE, list, get_, PREC_CALL),
RULE(TOKEN_RIGHT_SQUARE, NULL, NULL, PREC_NONE),
RULE(TOKEN_COLON, NULL, NULL, PREC_NONE),
RULE(TOKEN_COMMA, NULL, NULL, PREC_NONE),
RULE(TOKEN_DOT, NULL, dot, PREC_CALL),
RULE(TOKEN_MINUS, unary, binary, PREC_TERM),
RULE(TOKEN_PLUS, NULL, binary, PREC_TERM),
RULE(TOKEN_SEMICOLON, NULL, NULL, PREC_NONE),
RULE(TOKEN_SOLIDUS, NULL, binary, PREC_FACTOR),
RULE(TOKEN_ASTERISK, NULL, binary, PREC_FACTOR),
RULE(TOKEN_MODULO, NULL, binary, PREC_FACTOR),
RULE(TOKEN_BANG, unary, NULL, PREC_NONE),
RULE(TOKEN_BANG_EQUAL, NULL, binary, PREC_EQUALITY),
RULE(TOKEN_EQUAL, NULL, NULL, PREC_NONE),
RULE(TOKEN_EQUAL_EQUAL, NULL, binary, PREC_EQUALITY),
RULE(TOKEN_GREATER, NULL, binary, PREC_COMPARISON),
RULE(TOKEN_GREATER_EQUAL, NULL, binary, PREC_COMPARISON),
RULE(TOKEN_LESS, NULL, binary, PREC_COMPARISON),
RULE(TOKEN_LESS_EQUAL, NULL, binary, PREC_COMPARISON),
RULE(TOKEN_IDENTIFIER, variable, NULL, PREC_NONE),
RULE(TOKEN_STRING, string, NULL, PREC_NONE),
RULE(TOKEN_BIG_STRING, string, NULL, PREC_NONE),
RULE(TOKEN_NUMBER, number, NULL, PREC_NONE),
RULE(TOKEN_AND, NULL, and_, PREC_AND),
RULE(TOKEN_CLASS, NULL, NULL, PREC_NONE),
RULE(TOKEN_ELSE, NULL, NULL, PREC_NONE),
RULE(TOKEN_FALSE, literal, NULL, PREC_NONE),
RULE(TOKEN_FOR, NULL, NULL, PREC_NONE),
RULE(TOKEN_DEF, NULL, NULL, PREC_NONE),
RULE(TOKEN_IF, NULL, NULL, PREC_NONE),
RULE(TOKEN_IN, NULL, in_, PREC_COMPARISON),
RULE(TOKEN_LET, NULL, NULL, PREC_NONE),
RULE(TOKEN_NONE, literal, NULL, PREC_NONE),
RULE(TOKEN_NOT, unary, NULL, PREC_NONE),
RULE(TOKEN_OR, NULL, or_, PREC_OR),
RULE(TOKEN_PRINT, NULL, NULL, PREC_NONE),
RULE(TOKEN_RETURN, NULL, NULL, PREC_NONE),
RULE(TOKEN_SELF, self, NULL, PREC_NONE),
RULE(TOKEN_SUPER, super_, NULL, PREC_NONE),
RULE(TOKEN_TRUE, literal, NULL, PREC_NONE),
RULE(TOKEN_WHILE, NULL, NULL, PREC_NONE),
2020-12-31 10:59:51 +03:00
RULE(TOKEN_BREAK, NULL, NULL, PREC_NONE),
RULE(TOKEN_CONTINUE, NULL, NULL, PREC_NONE),
2021-01-01 02:55:39 +03:00
RULE(TOKEN_AT, NULL, NULL, PREC_NONE),
RULE(TOKEN_TILDE, unary, NULL, PREC_NONE),
RULE(TOKEN_PIPE, NULL, binary, PREC_BITOR),
RULE(TOKEN_CARET, NULL, binary, PREC_BITXOR),
RULE(TOKEN_AMPERSAND, NULL, binary, PREC_BITAND),
RULE(TOKEN_LEFT_SHIFT, NULL, binary, PREC_SHIFT),
RULE(TOKEN_RIGHT_SHIFT, NULL, binary, PREC_SHIFT),
RULE(TOKEN_PLUS_EQUAL, NULL, NULL, PREC_NONE),
RULE(TOKEN_MINUS_EQUAL, NULL, NULL, PREC_NONE),
RULE(TOKEN_PLUS_PLUS, NULL, NULL, PREC_NONE),
RULE(TOKEN_MINUS_MINUS, NULL, NULL, PREC_NONE),
/* This is going to get interesting */
RULE(TOKEN_INDENTATION, NULL, NULL, PREC_NONE),
RULE(TOKEN_ERROR, NULL, NULL, PREC_NONE),
RULE(TOKEN_EOL, NULL, NULL, PREC_NONE),
RULE(TOKEN_EOF, NULL, NULL, PREC_NONE),
};
static void parsePrecedence(Precedence precedence) {
advance();
ParseFn prefixRule = getRule(parser.previous.type)->prefix;
if (prefixRule == NULL) {
errorAtCurrent("Unexpected token.");
return;
}
2020-12-26 10:53:15 +03:00
int canAssign = precedence <= PREC_ASSIGNMENT;
prefixRule(canAssign);
while (precedence <= getRule(parser.current.type)->precedence) {
advance();
ParseFn infixRule = getRule(parser.previous.type)->infix;
2020-12-26 10:53:15 +03:00
infixRule(canAssign);
}
if (canAssign && matchAssignment()) {
2020-12-26 10:53:15 +03:00
error("invalid assignment target");
}
}
static ssize_t identifierConstant(KrkToken * name) {
2020-12-28 05:11:50 +03:00
return krk_addConstant(currentChunk(), OBJECT_VAL(krk_copyString(name->start, name->length)));
2020-12-26 10:53:15 +03:00
}
static ssize_t resolveLocal(Compiler * compiler, KrkToken * name) {
for (ssize_t i = compiler->localCount - 1; i >= 0; i--) {
Local * local = &compiler->locals[i];
if (identifiersEqual(name, &local->name)) {
if (local->depth == -1) {
error("Can not initialize value recursively (are you shadowing something?)");
}
return i;
}
}
return -1;
}
static void addLocal(KrkToken name) {
if (current->localCount + 1 > current->localsSpace) {
size_t old = current->localsSpace;
current->localsSpace = GROW_CAPACITY(old);
current->locals = GROW_ARRAY(Local,current->locals,old,current->localsSpace);
}
Local * local = &current->locals[current->localCount++];
local->name = name;
local->depth = -1;
local->isCaptured = 0;
}
static void declareVariable() {
if (current->scopeDepth == 0) return;
KrkToken * name = &parser.previous;
/* Detect duplicate definition */
for (ssize_t i = current->localCount - 1; i >= 0; i--) {
Local * local = &current->locals[i];
2020-12-28 05:11:50 +03:00
if (local->depth != -1 && local->depth < (ssize_t)current->scopeDepth) break;
2020-12-30 10:59:21 +03:00
if (identifiersEqual(name, &local->name)) {
error("Duplicate definition");
__asm__("int $3");
}
}
addLocal(*name);
}
static ssize_t parseVariable(const char * errorMessage) {
2020-12-26 10:53:15 +03:00
consume(TOKEN_IDENTIFIER, errorMessage);
declareVariable();
if (current->scopeDepth > 0) return 0;
2020-12-26 10:53:15 +03:00
return identifierConstant(&parser.previous);
}
static void defineVariable(size_t global) {
if (current->scopeDepth > 0) {
markInitialized();
return;
}
2020-12-26 10:53:15 +03:00
EMIT_CONSTANT_OP(OP_DEFINE_GLOBAL, global);
}
2021-01-03 11:59:50 +03:00
static void call(int canAssign) {
size_t argCount = 0, specialArgs = 0, keywordArgs = 0, seenKeywordUnpacking = 0;
if (!check(TOKEN_RIGHT_PAREN)) {
do {
2021-01-03 11:59:50 +03:00
if (match(TOKEN_ASTERISK)) {
specialArgs++;
if (match(TOKEN_ASTERISK)) {
seenKeywordUnpacking = 1;
emitBytes(OP_EXPAND_ARGS, 2); /* Outputs something special */
expression(); /* Expect dict */
continue;
} else {
if (seenKeywordUnpacking) {
error("Iterable expansion follows keyword argument unpacking.");
return;
}
emitBytes(OP_EXPAND_ARGS, 1); /* outputs something special */
expression();
continue;
}
}
if (match(TOKEN_IDENTIFIER)) {
KrkToken argName = parser.previous;
if (check(TOKEN_EQUAL)) {
/* This is a keyword argument. */
advance();
/* Output the name */
size_t ind = identifierConstant(&argName);
EMIT_CONSTANT_OP(OP_CONSTANT, ind);
expression();
keywordArgs++;
2021-01-03 11:59:50 +03:00
specialArgs++;
continue;
} else {
/*
* This is a regular argument that happened to start with an identifier,
* roll it back so we can process it that way.
*/
krk_ungetToken(parser.current);
parser.current = argName;
}
2021-01-03 11:59:50 +03:00
} else if (seenKeywordUnpacking) {
error("positional argument follows keyword argument unpacking");
return;
} else if (keywordArgs) {
2021-01-03 11:59:50 +03:00
error("Positional argument follows keyword argument");
return;
} else if (specialArgs) {
emitBytes(OP_EXPAND_ARGS, 0);
expression();
specialArgs++;
continue;
}
expression();
argCount++;
} while (match(TOKEN_COMMA));
}
consume(TOKEN_RIGHT_PAREN, "Expected ')' after arguments.");
2021-01-03 11:59:50 +03:00
if (specialArgs) {
/*
* Creates a sentinel at the top of the stack to tell the CALL instruction
* how many keyword arguments are at the top of the stack. This value
* triggers special handling in the CALL that processes the keyword arguments,
* which is relatively slow, so only use keyword arguments if you have to!
*/
2021-01-03 11:59:50 +03:00
EMIT_CONSTANT_OP(OP_KWARGS, specialArgs);
/*
* We added two elements - name and value - for each keyword arg,
* plus the sentinel object that will show up at the end after the
* OP_KWARGS instruction complets, so make sure we have the
* right depth into the stack when we execute CALL
*/
2021-01-03 11:59:50 +03:00
argCount += 1 /* for the sentinel */ + 2 * specialArgs;
}
2021-01-03 11:59:50 +03:00
EMIT_CONSTANT_OP(OP_CALL, argCount);
}
2020-12-26 14:39:47 +03:00
static void and_(int canAssign) {
int endJump = emitJump(OP_JUMP_IF_FALSE);
emitByte(OP_POP);
parsePrecedence(PREC_AND);
patchJump(endJump);
}
static void or_(int canAssign) {
int endJump = emitJump(OP_JUMP_IF_TRUE);
emitByte(OP_POP);
parsePrecedence(PREC_OR);
patchJump(endJump);
}
static ParseRule * getRule(KrkTokenType type) {
return &rules[type];
}
2020-12-28 04:54:25 +03:00
KrkFunction * krk_compile(const char * src, int newScope, char * fileName) {
krk_initScanner(src);
Compiler compiler;
initCompiler(&compiler, TYPE_MODULE);
2020-12-28 05:11:50 +03:00
compiler.function->chunk.filename = krk_copyString(fileName, strlen(fileName));
if (newScope) beginScope();
parser.hadError = 0;
parser.panicMode = 0;
advance();
2020-12-26 10:53:15 +03:00
while (!match(TOKEN_EOF)) {
declaration();
if (check(TOKEN_EOL) || check(TOKEN_INDENTATION) || check(TOKEN_EOF)) {
/* There's probably already and error... */
advance();
}
2020-12-26 10:53:15 +03:00
}
KrkFunction * function = endCompiler();
freeCompiler(&compiler);
return parser.hadError ? NULL : function;
}
void krk_markCompilerRoots() {
Compiler * compiler = current;
while (compiler != NULL) {
krk_markObject((KrkObj*)compiler->function);
compiler = compiler->enclosing;
}
}