diff --git a/Makefile b/Makefile index 2ac7d4f..9e367e0 100644 --- a/Makefile +++ b/Makefile @@ -37,7 +37,7 @@ ifeq ($(track), no) CFLAGS += -DNOTRACK endif -OBJ_D = util.o code.o ast.o ir.o conout.o ftepp.o opts.o file.o utf8.o +OBJ_D = util.o code.o ast.o ir.o conout.o ftepp.o opts.o file.o utf8.o correct.o OBJ_T = test.o util.o conout.o file.o OBJ_C = main.o lexer.o parser.o file.o OBJ_X = exec-standalone.o util.o conout.o file.o diff --git a/correct.c b/correct.c new file mode 100644 index 0000000..e32678e --- /dev/null +++ b/correct.c @@ -0,0 +1,389 @@ +/* + * Copyright (C) 2012, 2013 + * Dale Weiler + * + * Permission is hereby granted, free of charge, to any person obtaining a copy of + * this software and associated documentation files (the "Software"), to deal in + * the Software without restriction, including without limitation the rights to + * use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies + * of the Software, and to permit persons to whom the Software is furnished to do + * so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in all + * copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ +#include "gmqcc.h" + +/* + * This is a very clever method for correcting mistakes in QuakeC code + * most notably when invalid identifiers are used or inproper assignments; + * we can proprly lookup in multiple dictonaries (depening on the rules + * of what the task is trying to acomplish) to find the best possible + * match. + * + * + * A little about how it works, and probability theory: + * + * When given an identifier (which we will denote I), we're essentially + * just trying to choose the most likely correction for that identifier. + * (the actual "correction" can very well be the identifier itself). + * There is actually no way to know for sure that certian identifers + * such as "lates", need to be corrected to "late" or "latest" or any + * other permutations that look lexically the same. This is why we + * must advocate the usage of probabilities. This implies that we're + * trying to find the correction for C, out of all possible corrections + * that maximizes the probability of C for the original identifer I. + * + * Bayes' Therom suggests something of the following: + * AC P(I|C) P(C) / P(I) + * Since P(I) is the same for every possibly I, we can ignore it giving + * AC P(I|C) P(C) + * + * This greatly helps visualize how the parts of the expression are performed + * there is essentially three, from right to left we perform the following: + * + * 1: P(C), the probability that a proposed correction C will stand on its + * own. This is called the language model. + * + * 2: P(I|C), the probability that I would be used, when the programmer + * really meant C. This is the error model. + * + * 3: AC, the control mechanisim, which implies the enumeration of all + * feasible values of C, and then determine the one that gives the + * greatest probability score. Selecting it as the "correction" + * + * + * The requirement for complex expression involving two models: + * + * In reality the requirement for a more complex expression involving + * two seperate models is considerably a waste. But one must recognize + * that P(C|I) is already conflating two factors. It's just much simpler + * to seperate the two models and deal with them explicitaly. To properly + * estimate P(C|I) you have to consider both the probability of C and + * probability of the transposition from C to I. It's simply much more + * cleaner, and direct to seperate the two factors. + */ + +/* some hashtable management for dictonaries */ +static size_t *correct_find(ht table, const char *word) { + return (size_t*)util_htget(table, word); +} + +static int correct_update(ht *table, const char *word) { + size_t *data = correct_find(*table, word); + if (!data) + return 0; + + (*data)++; + return 1; +} + + +/* + * _ is valid in identifiers. I've yet to implement numerics however + * because they're only valid after the first character is of a _, or + * alpha character. + */ +static const char correct_alpha[] = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ_"; + +static char *correct_strndup(const char *src, size_t n) { + char *ret; + size_t len = strlen(src); + + if (n < len) + len = n; + + if (!(ret = (char*)mem_a(len + 1))) + return NULL; + + ret[len] = '\0'; + return (char*)memcpy(ret, src, len); +} + +static char *correct_concat(char *str1, char *str2, bool next) { + char *ret = NULL; + +#if 0 + if (!str1) { + str1 = mem_a(1); + *str1 = '\0'; + } +#endif + + str1 = mem_r (str1, strlen(str1) + strlen(str2) + 1); + ret = strcat(str1, str2); + + if (str2 && next) + mem_d(str2); + + return ret; +} + +/* + * correcting logic for the following forms of transformations: + * 1) deletion + * 2) transposition + * 3) alteration + * 4) insertion + */ +static size_t correct_deletion(const char *ident, char **array, size_t index) { + size_t itr; + size_t len = strlen(ident); + + for (itr = 0; itr < len; itr++) { + array[index + itr] = correct_concat ( + correct_strndup (ident, itr), + correct_strndup (ident+itr+1, len-(itr+1)), + true + ); + } + + return itr; +} + +static size_t correct_transposition(const char *ident, char **array, size_t index) { + size_t itr; + size_t len = strlen(ident); + + for (itr = 0; itr < len - 1; itr++) { + array[index + itr] = correct_concat ( + correct_concat ( + correct_strndup(ident, itr), + correct_strndup(ident+itr+1, 1), + true + ), + correct_concat ( + correct_strndup(ident+itr, 1), + correct_strndup(ident+itr+2, len-(itr+2)), + true + ), + true + ); + } + + return itr; +} + +static size_t correct_alteration(const char *ident, char **array, size_t index) { + size_t itr; + size_t jtr; + size_t ktr; + size_t len = strlen(ident); + char cct[2] = { 0, 0 }; /* char code table, for concatenation */ + + for (itr = 0, ktr = 0; itr < len; itr++) { + for (jtr = 0; jtr < sizeof(correct_alpha); jtr++, ktr++) { + *cct = correct_alpha[jtr]; + array[index + ktr] = correct_concat ( + correct_concat ( + correct_strndup(ident, itr), + (char *) &cct, + false + ), + correct_strndup ( + ident + (itr+1), + len - (itr+1) + ), + true + ); + } + } + + return ktr; +} + +static size_t correct_insertion(const char *ident, char **array, size_t index) { + size_t itr; + size_t jtr; + size_t ktr; + size_t len = strlen(ident); + char cct[2] = { 0, 0 }; /* char code table, for concatenation */ + + for (itr = 0, ktr = 0; itr <= len; itr++) { + for (jtr = 0; jtr < sizeof(correct_alpha); jtr++, ktr++) { + *cct = correct_alpha[jtr]; + array[index + ktr] = correct_concat ( + correct_concat ( + correct_strndup (ident, itr), + (char *) &cct, + false + ), + correct_strndup ( + ident+itr, + len - itr + ), + true + ); + } + } + + return ktr; +} + +static GMQCC_INLINE size_t correct_size(const char *ident) { + /* + * deletion = len + * transposition = len - 1 + * alteration = len * sizeof(correct_alpha) + * insertion = (len + 1) * sizeof(correct_alpha) + */ + + register size_t len = strlen(ident); + return (len) + (len - 1) + (len * sizeof(correct_alpha)) + ((len + 1) * sizeof(correct_alpha)); +} + +static char **correct_edit(const char *ident) { + size_t next; + char **find = (char**)mem_a(correct_size(ident) * sizeof(char*)); + + if (!find) + return NULL; + + next = correct_deletion (ident, find, 0); + next += correct_transposition(ident, find, next); + next += correct_alteration (ident, find, next); + /*****/ correct_insertion (ident, find, next); + + return find; +} + +/* + * We could use a hashtable but the space complexity isn't worth it + * since we're only going to determine the "did you mean?" identifier + * on error. + */ +static int correct_exist(char **array, size_t rows, char *ident) { + size_t itr; + for (itr = 0; itr < rows; itr++) + if (!strcmp(array[itr], ident)) + return 1; + + return 0; +} + +static char **correct_known(ht table, char **array, size_t rows, size_t *next) { + size_t itr; + size_t jtr; + size_t len; + size_t row; + char **res = NULL; + char **end; + + for (itr = 0, len = 0; itr < rows; itr++) { + end = correct_edit(array[itr]); + row = correct_size(array[itr]); + + for (jtr = 0; jtr < row; jtr++) { + if (correct_find(table, end[jtr]) && !correct_exist(res, len, end[jtr])) { + res = mem_r(res, sizeof(char*) * (len + 1)); + res[len++] = end[jtr]; + } else { + mem_d(end[jtr]); + } + } + + mem_d(end); + } + + *next = len; + return res; +} + +static char *correct_maximum(ht table, char **array, size_t rows) { + char *str = NULL; + size_t *itm = NULL; + size_t itr; + size_t top; + + for (itr = 0, top = 0; itr < rows; itr++) { + if ((itm = correct_find(table, array[itr])) && (*itm > top)) { + top = *itm; + str = array[itr]; + } + } + + return str; +} + +static void correct_cleanup(char **array, size_t rows) { + size_t itr; + for (itr = 0; itr < rows; itr++) + mem_d(array[itr]); + + mem_d(array); +} + +/* + * This is the exposed interface: + * takes a table for the dictonary a vector of sizes (used for internal + * probability calculation, and an identifier to "correct" + * + * the add function works the same. Except the identifier is used to + * add to the dictonary. + */ +void correct_add(ht table, size_t ***size, const char *ident) { + size_t *data = NULL; + const char *add = ident; + + if (!correct_update(&table, add)) { + data = (size_t*)mem_a(sizeof(size_t)); + *data = 1; + + vec_push((*size), data); + util_htset(table, add, data); + } +} + +char *correct_str(ht table, const char *ident) { + char **e1; + char **e2; + char *e1ident; + char *e2ident; + char *found = util_strdup(ident); + + size_t e1rows = 0; + size_t e2rows = 0; + + /* needs to be allocated for free later */ + if (correct_find(table, ident)) + return found; + + if ((e1rows = correct_size(ident))) { + e1 = correct_edit(ident); + + if ((e1ident = correct_maximum(table, e1, e1rows))) { + mem_d(found); + found = util_strdup(e1ident); + correct_cleanup(e1, e1rows); + return found; + } + } + + e2 = correct_known(table, e1, e1rows, &e2rows); + if (e2rows && ((e2ident = correct_maximum(table, e2, e2rows)))) { + mem_d(found); + found = util_strdup(e2ident); + } + + correct_cleanup(e1, e1rows); + correct_cleanup(e2, e2rows); + + return found; +} + +void correct_del(ht dictonary, size_t **data) { + size_t i; + for (i = 0; i < vec_size(data); i++) + mem_d(data[i]); + + vec_free(data); + util_htdel(dictonary); +} diff --git a/gmqcc.h b/gmqcc.h index 36a67fd..036cab9 100644 --- a/gmqcc.h +++ b/gmqcc.h @@ -245,9 +245,9 @@ /*===================================================================*/ /*=========================== util.c ================================*/ /*===================================================================*/ -void *util_memory_a (size_t, unsigned int, const char *); -void util_memory_d (void *, unsigned int, const char *); -void *util_memory_r (void *, size_t, unsigned int, const char *); +void *util_memory_a (size_t, /*****/ unsigned int, const char *); +void *util_memory_r (void *, size_t, unsigned int, const char *); +void util_memory_d (void *); void util_meminfo (); bool util_filexists (const char *); @@ -275,7 +275,7 @@ int util_asprintf (char **ret, const char *fmt, ...); # define mem_r(x, n) realloc((void*)x, n) #else # define mem_a(x) util_memory_a((x), __LINE__, __FILE__) -# define mem_d(x) util_memory_d((void*)(x), __LINE__, __FILE__) +# define mem_d(x) util_memory_d((void*)(x)) # define mem_r(x, n) util_memory_r((void*)(x), (n), __LINE__, __FILE__) #endif @@ -423,6 +423,13 @@ GMQCC_INLINE FILE *file_open (const char *, const char *); /*NOINLINE*/ int file_getline(char **, size_t *, FILE *); +/*===================================================================*/ +/*=========================== correct.c =============================*/ +/*===================================================================*/ +void correct_del(ht, size_t **); +void correct_add(ht, size_t ***, const char *); +char *correct_str(ht, /********/ const char *); + /*===================================================================*/ /*=========================== code.c ================================*/ /*===================================================================*/ diff --git a/parser.c b/parser.c index 229d2ec..2d7ebe3 100644 --- a/parser.c +++ b/parser.c @@ -74,6 +74,10 @@ typedef struct { ht htglobals; ht *typedefs; + /* same as above but for the spelling corrector */ + ht *correct_variables; + size_t ***correct_variables_score; /* vector of vector of size_t* */ + /* not to be used directly, we use the hash table */ ast_expression **_locals; size_t *_blocklocals; @@ -1614,13 +1618,15 @@ static ast_expression* parse_expression_leave(parser_t *parser, bool stopatcomma } else { + size_t i; + char *correct = NULL; + /* * sometimes people use preprocessing predefs without enabling them * i've done this thousands of times already myself. Lets check for * it in the predef table. And diagnose it better :) */ if (!OPTS_FLAG(FTEPP_PREDEFS)) { - size_t i; for (i = 0; i < sizeof(ftepp_predefs)/sizeof(*ftepp_predefs); i++) { if (!strcmp(ftepp_predefs[i].name, parser_tokval(parser))) { parseerror(parser, "unexpected ident: %s (use -fftepp-predef to enable pre-defined macros)", parser_tokval(parser)); @@ -1629,7 +1635,29 @@ static ast_expression* parse_expression_leave(parser_t *parser, bool stopatcomma } } - parseerror(parser, "unexpected ident: %s", parser_tokval(parser)); + /* + * TODO: determine the best score for the identifier: be it + * a variable, a field. + * + * We should also consider adding correction tables for + * other things as well. + */ + for (i = 0; i < vec_size(parser->correct_variables); i++) { + correct = correct_str(parser->correct_variables[i], "ello"); + if (strcmp(correct, parser_tokval(parser))) { + break; + } else if (correct) { + mem_d(correct); + } + } + + if (correct) { + parseerror(parser, "unexpected ident: %s (did you mean %s?)", parser_tokval(parser), correct); + mem_d(correct); + } else { + parseerror(parser, "unexpected ident: %s", parser_tokval(parser)); + } + goto onerr; } } @@ -1968,6 +1996,10 @@ static void parser_enterblock(parser_t *parser) vec_push(parser->typedefs, util_htnew(TYPEDEF_HT_SIZE)); vec_push(parser->_blocktypedefs, vec_size(parser->_typedefs)); vec_push(parser->_block_ctx, parser_ctx(parser)); + + /* corrector */ + vec_push(parser->correct_variables, util_htnew(PARSER_HT_SIZE)); + vec_push(parser->correct_variables_score, NULL); } static bool parser_leaveblock(parser_t *parser) @@ -1981,7 +2013,11 @@ static bool parser_leaveblock(parser_t *parser) } util_htdel(vec_last(parser->variables)); + correct_del(vec_last(parser->correct_variables), vec_last(parser->correct_variables_score)); + vec_pop(parser->variables); + vec_pop(parser->correct_variables); + vec_pop(parser->correct_variables_score); if (!vec_size(parser->_blocklocals)) { parseerror(parser, "internal error: parser_leaveblock with no block (2)"); return false; @@ -2008,6 +2044,7 @@ static bool parser_leaveblock(parser_t *parser) vec_pop(parser->typedefs); vec_pop(parser->_block_ctx); + return rv; } @@ -2015,6 +2052,13 @@ static void parser_addlocal(parser_t *parser, const char *name, ast_expression * { vec_push(parser->_locals, e); util_htset(vec_last(parser->variables), name, (void*)e); + + /* corrector */ + correct_add ( + vec_last(parser->correct_variables), + &vec_last(parser->correct_variables_score), + name + ); } static ast_expression* process_condition(parser_t *parser, ast_expression *cond, bool *_ifnot) @@ -3549,6 +3593,7 @@ static bool parse_function_body(parser_t *parser, ast_value *var) vec_push(parser->globals, (ast_expression*)thinkfunc); util_htset(parser->htglobals, thinkfunc->name, thinkfunc); + nextthink = (ast_expression*)thinkfunc; } else { @@ -4790,6 +4835,14 @@ static bool parse_variable(parser_t *parser, ast_block *localblock, bool nofield /* Add it to the local scope */ util_htset(vec_last(parser->variables), var->name, (void*)var); + + /* corrector */ + correct_add ( + vec_last(parser->correct_variables), + &vec_last(parser->correct_variables_score), + var->name + ); + /* now rename the global */ ln = strlen(var->name); vec_append(defname, ln, var->name); @@ -4803,6 +4856,13 @@ static bool parse_variable(parser_t *parser, ast_block *localblock, bool nofield for (i = 0; i < 3; ++i) { util_htset(vec_last(parser->variables), me[i]->name, (void*)(me[i])); + /* corrector */ + correct_add( + vec_last(parser->correct_variables), + &vec_last(parser->correct_variables_score), + me[i]->name + ); + vec_shrinkto(defname, prefix_len); ln = strlen(me[i]->name); vec_append(defname, ln, me[i]->name); @@ -5349,6 +5409,17 @@ void parser_cleanup() vec_free(parser->_blocklocals); vec_free(parser->_locals); + /* corrector */ + for (i = 0; i < vec_size(parser->correct_variables); ++i) { + correct_del(parser->correct_variables[i], parser->correct_variables_score[i]); + } + for (i = 0; i < vec_size(parser->correct_variables_score); ++i) { + vec_free(parser->correct_variables_score[i]); + } + vec_free(parser->correct_variables); + vec_free(parser->correct_variables_score); + + for (i = 0; i < vec_size(parser->_typedefs); ++i) ast_delete(parser->_typedefs[i]); vec_free(parser->_typedefs); diff --git a/util.c b/util.c index 5d0d683..92a99f7 100644 --- a/util.c +++ b/util.c @@ -54,20 +54,18 @@ void *util_memory_a(size_t byte, unsigned int line, const char *file) { mem_start->prev = info; mem_start = info; - util_debug("MEM", "allocation: % 8u (bytes) address 0x%08X @ %s:%u\n", byte, data, file, line); mem_at++; mem_ab += info->byte; return data; } -void util_memory_d(void *ptrn, unsigned int line, const char *file) { +void util_memory_d(void *ptrn) { struct memblock_t *info = NULL; if (!ptrn) return; info = ((struct memblock_t*)ptrn - 1); - util_debug("MEM", "released: % 8u (bytes) address 0x%08X @ %s:%u\n", info->byte, ptrn, file, line); mem_db += info->byte; mem_dt++; @@ -89,18 +87,16 @@ void *util_memory_r(void *ptrn, size_t byte, unsigned int line, const char *file if (!ptrn) return util_memory_a(byte, line, file); if (!byte) { - util_memory_d(ptrn, line, file); + util_memory_d(ptrn); return NULL; } oldinfo = ((struct memblock_t*)ptrn - 1); newinfo = ((struct memblock_t*)malloc(sizeof(struct memblock_t) + byte)); - util_debug("MEM", "reallocation: % 8u -> %u (bytes) address 0x%08X -> 0x%08X @ %s:%u\n", oldinfo->byte, byte, ptrn, (void*)(newinfo+1), file, line); - /* new data */ if (!newinfo) { - util_memory_d(oldinfo+1, line, file); + util_memory_d(oldinfo+1); return NULL; }