*/
#include "gmqcc.h"
+/*
+ * A forward allcator for the corrector. This corrector requires a lot
+ * of allocations. This forward allocator combats all those allocations
+ * and speeds us up a little. It also saves us space in a way since each
+ * allocation isn't wasting a little header space for when NOTRACK isn't
+ * defined.
+ */
+#define CORRECT_POOLSIZE (128*1024*1024)
+
+static unsigned char **correct_pool_data = NULL;
+static unsigned char *correct_pool_this = NULL;
+static size_t correct_pool_addr = 0;
+
+static GMQCC_INLINE void correct_pool_new(void) {
+ correct_pool_addr = 0;
+ correct_pool_this = (unsigned char *)mem_a(CORRECT_POOLSIZE);
+
+ vec_push(correct_pool_data, correct_pool_this);
+}
+
+static GMQCC_INLINE void *correct_pool_alloc(size_t bytes) {
+ void *data;
+ if (correct_pool_addr + bytes >= CORRECT_POOLSIZE)
+ correct_pool_new();
+
+ data = correct_pool_this;
+ correct_pool_this += bytes;
+ correct_pool_addr += bytes;
+
+ return data;
+}
+
+static GMQCC_INLINE void correct_pool_delete(void) {
+ size_t i;
+ for (i = 0; i < vec_size(correct_pool_data); ++i)
+ mem_d(correct_pool_data[i]);
+
+ correct_pool_data = NULL;
+ correct_pool_this = NULL;
+ correct_pool_addr = 0;
+}
+
+
+static GMQCC_INLINE char *correct_outstr(const char *s) {
+ char *o = util_strdup(s);
+ correct_pool_delete();
+ return o;
+}
+
+correct_trie_t* correct_trie_new()
+{
+ correct_trie_t *t = (correct_trie_t*)mem_a(sizeof(correct_trie_t));
+ t->value = NULL;
+ t->entries = NULL;
+ return t;
+}
+
+void correct_trie_del_sub(correct_trie_t *t)
+{
+ size_t i;
+ for (i = 0; i < vec_size(t->entries); ++i)
+ correct_trie_del_sub(&t->entries[i]);
+ vec_free(t->entries);
+}
+
+void correct_trie_del(correct_trie_t *t)
+{
+ size_t i;
+ for (i = 0; i < vec_size(t->entries); ++i)
+ correct_trie_del_sub(&t->entries[i]);
+ vec_free(t->entries);
+ mem_d(t);
+}
+
+void* correct_trie_get(const correct_trie_t *t, const char *key)
+{
+ const unsigned char *data = (const unsigned char*)key;
+ while (*data) {
+ unsigned char ch = *data;
+ const size_t vs = vec_size(t->entries);
+ size_t i;
+ const correct_trie_t *entries = t->entries;
+ for (i = 0; i < vs; ++i) {
+ if (entries[i].ch == ch) {
+ t = &entries[i];
+ ++data;
+ break;
+ }
+ }
+ if (i == vs)
+ return NULL;
+ }
+ return t->value;
+}
+
+void correct_trie_set(correct_trie_t *t, const char *key, void * const value)
+{
+ const unsigned char *data = (const unsigned char*)key;
+ while (*data) {
+ unsigned char ch = *data;
+ correct_trie_t *entries = t->entries;
+ const size_t vs = vec_size(t->entries);
+ size_t i;
+ for (i = 0; i < vs; ++i) {
+ if (entries[i].ch == ch) {
+ t = &entries[i];
+ break;
+ }
+ }
+ if (i == vs) {
+ correct_trie_t *elem = (correct_trie_t*)vec_add(t->entries, 1);
+ elem->ch = ch;
+ elem->value = NULL;
+ elem->entries = NULL;
+ t = elem;
+ }
+ ++data;
+ }
+ t->value = value;
+}
+
/*
* This is a very clever method for correcting mistakes in QuakeC code
* most notably when invalid identifiers are used or inproper assignments;
*/
/* some hashtable management for dictonaries */
-static size_t *correct_find(ht table, const char *word) {
- return (size_t*)util_htget(table, word);
+static size_t *correct_find(correct_trie_t *table, const char *word) {
+ return (size_t*)correct_trie_get(table, word);
}
-static int correct_update(ht *table, const char *word) {
+static int correct_update(correct_trie_t* *table, const char *word) {
size_t *data = correct_find(*table, word);
if (!data)
return 0;
return 1;
}
+void correct_add(correct_trie_t* 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);
+ correct_trie_set(table, add, data);
+ }
+}
+
+void correct_del(correct_trie_t* dictonary, size_t **data) {
+ size_t i;
+ const size_t vs = vec_size(data);
+ for (i = 0; i < vs; i++)
+ mem_d(data[i]);
+
+ vec_free(data);
+ correct_trie_del(dictonary);
+}
+#if 1
+#undef mem_a
+#undef mem_r
+#undef mem_d
+#define mem_a(x) correct_alloc((x))
+#define mem_r(a,b) correct_realloc((a),(b))
+/* doing this in order to avoid 'unused variable' warnings */
+#define mem_d(x) ((void)(0 && (x)))
+#endif
+
/*
* _ is valid in identifiers. I've yet to implement numerics however
*/
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 (!str1) {
- str1 = mem_a(1);
- *str1 = '\0';
- }
-
- 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
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
- );
+ char *a = (char*)correct_pool_alloc(len+1);
+ memcpy(a, ident, itr);
+ memcpy(a + itr, ident + itr + 1, len - itr);
+ array[index + itr] = a;
}
return 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
- );
+ char tmp;
+ char *a = (char*)correct_pool_alloc(len+1);
+ memcpy(a, ident, len+1);
+ tmp = a[itr];
+ a[itr ] = a[itr+1];
+ a[itr+1] = tmp;
+ array[index + itr] = a;
}
return 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
- );
+ for (jtr = 0; jtr < sizeof(correct_alpha)-1; jtr++, ktr++) {
+ char *a = (char*)correct_pool_alloc(len+1);
+ memcpy(a, ident, len+1);
+ a[itr] = correct_alpha[jtr];
+ array[index + ktr] = a;
}
}
size_t itr;
size_t jtr;
size_t ktr;
- size_t len = strlen(ident);
- char cct[2] = { 0, 0 }; /* char code table, for concatenation */
+ const size_t len = strlen(ident);
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
- );
+ for (jtr = 0; jtr < sizeof(correct_alpha)-1; jtr++, ktr++) {
+ char *a = (char*)correct_pool_alloc(len+2);
+ memcpy(a, ident, itr);
+ memcpy(a + itr + 1, ident + itr, len - itr + 1);
+ a[itr] = correct_alpha[jtr];
+ array[index + ktr] = a;
}
}
*/
register size_t len = strlen(ident);
- return (len) + (len - 1) + (len * sizeof(correct_alpha)) + ((len + 1) * sizeof(correct_alpha));
+ return (len) + (len - 1) + (len * (sizeof(correct_alpha)-1)) + ((len + 1) * (sizeof(correct_alpha)-1));
}
static char **correct_edit(const char *ident) {
size_t next;
- char **find = (char**)mem_a(correct_size(ident) * sizeof(char*));
+ char **find = (char**)correct_pool_alloc(correct_size(ident) * sizeof(char*));
if (!find)
return NULL;
return 0;
}
-static char **correct_known(ht table, char **array, size_t rows, size_t *next) {
+static GMQCC_INLINE char **correct_known_resize(char **res, size_t *allocated, size_t size) {
+ size_t oldallocated = *allocated;
+ char **out;
+ if (size+1 < *allocated)
+ return res;
+
+ *allocated += 32;
+ out = correct_pool_alloc(sizeof(*res) * *allocated);
+ memcpy(out, res, sizeof(*res) * oldallocated);
+ return out;
+}
+
+static char **correct_known(correct_trie_t* 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;
+ size_t nxt = 8;
+ char **res = correct_pool_alloc(sizeof(char *) * nxt);
+ char **end = NULL;
for (itr = 0, len = 0; itr < rows; itr++) {
end = correct_edit(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 = correct_known_resize(res, &nxt, len+1);
res[len++] = end[jtr];
}
}
-
- mem_d(end);
}
*next = len;
return res;
}
-static char *correct_maximum(ht table, char **array, size_t rows) {
+static char *correct_maximum(correct_trie_t* table, char **array, size_t rows) {
char *str = NULL;
size_t *itm = NULL;
size_t 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]);
-}
-
/*
* This is the exposed interface:
* takes a table for the dictonary a vector of sizes (used for internal
* 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_correct(ht table, const char *ident) {
+char *correct_str(correct_trie_t* table, const char *ident) {
char **e1;
char **e2;
char *e1ident;
size_t e1rows = 0;
size_t e2rows = 0;
+ correct_pool_new();
+
/* needs to be allocated for free later */
if (correct_find(table, ident))
- return found;
+ return correct_outstr(found);
- mem_d(found);
if ((e1rows = correct_size(ident))) {
e1 = correct_edit(ident);
if ((e1ident = correct_maximum(table, e1, e1rows))) {
found = util_strdup(e1ident);
- correct_cleanup(e1, e1rows);
- mem_d(e1);
- return found;
+ return correct_outstr(found);
}
}
e2 = correct_known(table, e1, e1rows, &e2rows);
- if (e2rows && ((e2ident = correct_maximum(table, e2, e2rows))))
+ if (e2rows && ((e2ident = correct_maximum(table, e2, e2rows)))) {
found = util_strdup(e2ident);
-
- correct_cleanup(e1, e1rows);
- correct_cleanup(e2, e2rows);
-
- mem_d(e1);
- mem_d(e2);
-
- 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);
-}
-
-int main() {
- opts.debug = true;
- opts.memchk = true;
- con_init();
-
- ht t = util_htnew(1024);
- size_t **d = NULL;
-
- correct_add(t, &d, "hellobain");
- correct_add(t, &d, "ellor");
- correct_add(t, &d, "world");
-
- printf("found identifiers: (2)\n");
- printf(" 1: hellobain\n");
- printf(" 2: ellor\n");
- printf(" 3: world\n");
-
- char *b = correct_correct(t, "rld");
- char *a = correct_correct(t, "ello");
- char *c = correct_correct(t, "helbain");
-
- printf("invalid identifier: `%s` (did you mean: `%s`?)\n", "ello", a);
- printf("invalid identifier: `%s` (did you mean: `%s`?)\n", "rld", b);
- printf("invalid identifier: `%s` (did you mean: `%s`?)\n", "helbain", c);
-
- correct_del(t, d);
- mem_d(b);
- mem_d(a);
- mem_d(c);
-
- /*util_meminfo();*/
+ }
+ return correct_outstr(found);
}
projects / xonotic / gmqcc.git / blobdiff