spell.c

PostgreSQL 8.1.4的源码 适用于Linux下的开源数据库系统

	Affix->data->aff = (AFFIX **) malloc(sizeof(AFFIX *) * cnt);
	MEMOUT(Affix->data->aff);
	Affix->data->naff = (uint32) cnt;

	cnt = 0;
	for (i = start; i < end; i++)
		if (Conf->Affix[i].replen == 0)
		{
			Affix->data->aff[cnt] = Conf->Affix + i;
			cnt++;
		}
}

void
NISortAffixes(IspellDict * Conf)
{
	AFFIX	   *Affix;
	size_t		i;
	CMPDAffix  *ptr;
	int			firstsuffix = -1;

	if (Conf->naffixes > 1)
		qsort((void *) Conf->Affix, Conf->naffixes, sizeof(AFFIX), cmpaffix);

	Conf->CompoundAffix = ptr = (CMPDAffix *) malloc(sizeof(CMPDAffix) * Conf->naffixes);
	MEMOUT(Conf->CompoundAffix);
	ptr->affix = NULL;

	for (i = 0; i < Conf->naffixes; i++)
	{
		Affix = &(((AFFIX *) Conf->Affix)[i]);
		if (Affix->type == FF_SUFFIX)
		{
			if (firstsuffix < 0)
				firstsuffix = i;
			if (Affix->flagflags & FF_COMPOUNDONLYAFX)
			{
				if (!ptr->affix ||
					strbncmp((const unsigned char *) (ptr - 1)->affix,
							 (const unsigned char *) Affix->repl,
							 (ptr - 1)->len))
				{
					/* leave only unique and minimals suffixes */
					ptr->affix = Affix->repl;
					ptr->len = Affix->replen;
					ptr++;
				}
			}
		}
	}
	ptr->affix = NULL;
	Conf->CompoundAffix = (CMPDAffix *) realloc(Conf->CompoundAffix, sizeof(CMPDAffix) * (ptr - Conf->CompoundAffix + 1));

	Conf->Prefix = mkANode(Conf, 0, firstsuffix, 0, FF_PREFIX);
	Conf->Suffix = mkANode(Conf, firstsuffix, Conf->naffixes, 0, FF_SUFFIX);
	mkVoidAffix(Conf, 1, firstsuffix);
	mkVoidAffix(Conf, 0, firstsuffix);
}

static AffixNodeData *
FinfAffixes(AffixNode * node, const char *word, int wrdlen, int *level, int type)
{
	AffixNodeData *StopLow,
			   *StopHigh,
			   *StopMiddle;
	uint8		symbol;

	if (node->isvoid)
	{							/* search void affixes */
		if (node->data->naff)
			return node->data;
		node = node->data->node;
	}

	while (node && *level < wrdlen)
	{
		StopLow = node->data;
		StopHigh = node->data + node->length;
		while (StopLow < StopHigh)
		{
			StopMiddle = StopLow + ((StopHigh - StopLow) >> 1);
			symbol = GETWCHAR(word, wrdlen, *level, type);
			if (StopMiddle->val == symbol)
			{
				(*level)++;
				if (StopMiddle->naff)
					return StopMiddle;
				node = StopMiddle->node;
				break;
			}
			else if (StopMiddle->val < symbol)
				StopLow = StopMiddle + 1;
			else
				StopHigh = StopMiddle;
		}
		if (StopLow >= StopHigh)
			break;
	}
	return NULL;
}

static char *
CheckAffix(const char *word, size_t len, AFFIX * Affix, char flagflags, char *newword)
{

	if (flagflags & FF_COMPOUNDONLYAFX)
	{
		if ((Affix->flagflags & FF_COMPOUNDONLYAFX) == 0)
			return NULL;
	}
	else
	{
		if (Affix->flagflags & FF_COMPOUNDONLYAFX)
			return NULL;
	}

	if (Affix->type == FF_SUFFIX)
	{
		strcpy(newword, word);
		strcpy(newword + len - Affix->replen, Affix->find);
	}
	else
	{
		strcpy(newword, Affix->find);
		strcat(newword, word + Affix->replen);
	}

	if (Affix->issimple)
		return newword;
	else if (Affix->isregis)
	{
		if (Affix->compile)
		{
			RS_compile(&(Affix->reg.regis), (Affix->type == FF_SUFFIX) ? 1 : 0, Affix->mask);
			Affix->compile = 0;
		}
		if (RS_execute(&(Affix->reg.regis), newword, -1))
			return newword;
	}
	else
	{
		regmatch_t	subs[2];	/* workaround for apache&linux */
		int			err;
		pg_wchar   *data;
		size_t		data_len;
		int			dat_len;

		if (Affix->compile)
		{
			int			wmasklen,
						masklen = strlen(Affix->mask);
			pg_wchar   *mask;

			mask = (pg_wchar *) palloc((masklen + 1) * sizeof(pg_wchar));
			wmasklen = pg_mb2wchar_with_len(Affix->mask, mask, masklen);

			err = pg_regcomp(&(Affix->reg.regex), mask, wmasklen, REG_EXTENDED | REG_ICASE | REG_NOSUB);
			pfree(mask);
			if (err)
			{
				char		regerrstr[ERRSTRSIZE];

				pg_regerror(err, &(Affix->reg.regex), regerrstr, ERRSTRSIZE);
				elog(ERROR, "Regex error in '%s': %s", Affix->mask, regerrstr);
			}
			Affix->compile = 0;
		}

		/* Convert data string to wide characters */
		dat_len = strlen(newword);
		data = (pg_wchar *) palloc((dat_len + 1) * sizeof(pg_wchar));
		data_len = pg_mb2wchar_with_len(newword, data, dat_len);

		if (!(err = pg_regexec(&(Affix->reg.regex), data, data_len, 0, NULL, 1, subs, 0)))
		{
			pfree(data);
			return newword;
		}
		pfree(data);
	}

	return NULL;
}


static char **
NormalizeSubWord(IspellDict * Conf, char *word, char flag)
{
	AffixNodeData *suffix = NULL,
			   *prefix = NULL;
	int			slevel = 0,
				plevel = 0;
	int			wrdlen = strlen(word),
				swrdlen;
	char	  **forms;
	char	  **cur;
	char		newword[2 * MAXNORMLEN] = "";
	char		pnewword[2 * MAXNORMLEN] = "";
	AffixNode  *snode = Conf->Suffix,
			   *pnode;
	int			i,
				j;

	if (wrdlen > MAXNORMLEN)
		return NULL;
	strlower(word);
	cur = forms = (char **) palloc(MAX_NORM * sizeof(char *));
	*cur = NULL;


	/* Check that the word itself is normal form */
	if (FindWord(Conf, word, 0, flag & FF_COMPOUNDWORD))
	{
		*cur = pstrdup(word);
		cur++;
		*cur = NULL;
	}

	/* Find all other NORMAL forms of the 'word' (check only prefix) */
	pnode = Conf->Prefix;
	plevel = 0;
	while (pnode)
	{
		prefix = FinfAffixes(pnode, word, wrdlen, &plevel, FF_PREFIX);
		if (!prefix)
			break;
		for (j = 0; j < prefix->naff; j++)
		{
			if (CheckAffix(word, wrdlen, prefix->aff[j], flag, newword))
			{
				/* prefix success */
				if (FindWord(Conf, newword, prefix->aff[j]->flag, flag & FF_COMPOUNDWORD) && (cur - forms) < (MAX_NORM - 1))
				{
					/* word search success */
					*cur = pstrdup(newword);
					cur++;
					*cur = NULL;
				}
			}
		}
		pnode = prefix->node;
	}

	/*
	 * Find all other NORMAL forms of the 'word' (check suffix and then
	 * prefix)
	 */
	while (snode)
	{
		/* find possible suffix */
		suffix = FinfAffixes(snode, word, wrdlen, &slevel, FF_SUFFIX);
		if (!suffix)
			break;
		/* foreach suffix check affix */
		for (i = 0; i < suffix->naff; i++)
		{
			if (CheckAffix(word, wrdlen, suffix->aff[i], flag, newword))
			{
				/* suffix success */
				if (FindWord(Conf, newword, suffix->aff[i]->flag, flag & FF_COMPOUNDWORD) && (cur - forms) < (MAX_NORM - 1))
				{
					/* word search success */
					*cur = pstrdup(newword);
					cur++;
					*cur = NULL;
				}
				/* now we will look changed word with prefixes */
				pnode = Conf->Prefix;
				plevel = 0;
				swrdlen = strlen(newword);
				while (pnode)
				{
					prefix = FinfAffixes(pnode, newword, swrdlen, &plevel, FF_PREFIX);
					if (!prefix)
						break;
					for (j = 0; j < prefix->naff; j++)
					{
						if (CheckAffix(newword, swrdlen, prefix->aff[j], flag, pnewword))
						{
							/* prefix success */
							int			ff = (prefix->aff[j]->flagflags & suffix->aff[i]->flagflags & FF_CROSSPRODUCT) ?
							0 : prefix->aff[j]->flag;

							if (FindWord(Conf, pnewword, ff, flag & FF_COMPOUNDWORD) && (cur - forms) < (MAX_NORM - 1))
							{
								/* word search success */
								*cur = pstrdup(pnewword);
								cur++;
								*cur = NULL;
							}
						}
					}
					pnode = prefix->node;
				}
			}
		}

		snode = suffix->node;
	}

	if (cur == forms)
	{
		pfree(forms);
		return (NULL);
	}
	return (forms);
}

typedef struct SplitVar
{
	int			nstem;
	char	  **stem;
	struct SplitVar *next;
}	SplitVar;

static int
CheckCompoundAffixes(CMPDAffix ** ptr, char *word, int len)
{
	while ((*ptr)->affix)
	{
		if (len > (*ptr)->len && strncmp((*ptr)->affix, word, (*ptr)->len) == 0)
		{
			len = (*ptr)->len;
			(*ptr)++;
			return len;
		}
		(*ptr)++;
	}
	return 0;
}

static SplitVar *
CopyVar(SplitVar * s, int makedup)
{
	SplitVar   *v = (SplitVar *) palloc(sizeof(SplitVar));

	v->stem = (char **) palloc(sizeof(char *) * (MAX_NORM));
	v->next = NULL;
	if (s)
	{
		int			i;

		v->nstem = s->nstem;
		for (i = 0; i < s->nstem; i++)
			v->stem[i] = (makedup) ? pstrdup(s->stem[i]) : s->stem[i];
	}
	else
		v->nstem = 0;
	return v;
}


static SplitVar *
SplitToVariants(IspellDict * Conf, SPNode * snode, SplitVar * orig, char *word, int wordlen, int startpos, int minpos)
{
	SplitVar   *var = NULL;
	SPNodeData *StopLow,
			   *StopHigh,
			   *StopMiddle = NULL;
	SPNode	   *node = (snode) ? snode : Conf->Dictionary;
	int			level = (snode) ? minpos : startpos;	/* recursive
														 * minpos==level */
	int			lenaff;
	CMPDAffix  *caff;
	char	   *notprobed;

	notprobed = (char *) palloc(wordlen);
	memset(notprobed, 1, wordlen);
	var = CopyVar(orig, 1);

	while (node && level < wordlen)
	{
		StopLow = node->data;
		StopHigh = node->data + node->length;
		while (StopLow < StopHigh)
		{
			StopMiddle = StopLow + ((StopHigh - StopLow) >> 1);
			if (StopMiddle->val == ((uint8 *) (word))[level])
				break;
			else if (StopMiddle->val < ((uint8 *) (word))[level])
				StopLow = StopMiddle + 1;
			else
				StopHigh = StopMiddle;
		}
		if (StopLow >= StopHigh)
			break;

		/* find word with epenthetic */
		caff = Conf->CompoundAffix;
		while (level > startpos && (lenaff = CheckCompoundAffixes(&caff, word + level, wordlen - level)) > 0)
		{
			/*
			 * there is one of compound suffixes, so check word for existings
			 */
			char		buf[MAXNORMLEN];
			char	  **subres;

			lenaff = level - startpos + lenaff;

			if (!notprobed[startpos + lenaff - 1])
				continue;

			if (level + lenaff - 1 <= minpos)
				continue;

			memcpy(buf, word + startpos, lenaff);
			buf[lenaff] = '\0';

			subres = NormalizeSubWord(Conf, buf, FF_COMPOUNDWORD | FF_COMPOUNDONLYAFX);
			if (subres)
			{
				/* Yes, it was a word from dictionary */
				SplitVar   *new = CopyVar(var, 0);
				SplitVar   *ptr = var;
				char	  **sptr = subres;

				notprobed[startpos + lenaff - 1] = 0;

				while (*sptr)
				{
					new->stem[new->nstem] = *sptr;
					new->nstem++;
					sptr++;
				}
				pfree(subres);

				while (ptr->next)
					ptr = ptr->next;
				ptr->next = SplitToVariants(Conf, NULL, new, word, wordlen, startpos + lenaff, startpos + lenaff);

				pfree(new->stem);
				pfree(new);
			}
		}

		/* find infinitive */
		if (StopMiddle->isword && StopMiddle->compoundallow && notprobed[level])
		{
			/* ok, we found full compoundallowed word */
			if (level > minpos)
			{
				/* and its length more than minimal */
				if (wordlen == level + 1)
				{
					/* well, it was last word */
					var->stem[var->nstem] = strnduplicate(word + startpos, wordlen - startpos);
					var->nstem++;
					pfree(notprobed);
					return var;
				}
				else
				{
					/* then we will search more big word at the same point */
					SplitVar   *ptr = var;

					while (ptr->next)
						ptr = ptr->next;
					ptr->next = SplitToVariants(Conf, node, var, word, wordlen, startpos, level);
					/* we can find next word */
					level++;
					var->stem[var->nstem] = strnduplicate(word + startpos, level - startpos);
					var->nstem++;
					node = Conf->Dictionary;
					startpos = level;
					continue;
				}
			}
		}
		level++;
		node = StopMiddle->node;
	}

	var->stem[var->nstem] = strnduplicate(word + startpos, wordlen - startpos);
	var->nstem++;
	pfree(notprobed);
	return var;
}

TSLexeme *
NINormalizeWord(IspellDict * Conf, char *word)
{
	char	  **res = NormalizeSubWord(Conf, word, 0);
	TSLexeme   *lcur = NULL,
			   *lres = NULL;
	uint16		NVariant = 1;

	if (res)
	{
		char	  **ptr = res;

		lcur = lres = (TSLexeme *) palloc(MAX_NORM * sizeof(TSLexeme));
		while (*ptr)
		{
			lcur->lexeme = *ptr;
			lcur->flags = 0;
			lcur->nvariant = NVariant++;
			lcur++;
			ptr++;
		}
		lcur->lexeme = NULL;
		pfree(res);
	}

	if (Conf->compoundcontrol != '\t')
	{
		int			wordlen = strlen(word);
		SplitVar   *ptr,
				   *var = SplitToVariants(Conf, NULL, NULL, word, wordlen, 0, -1);
		int			i;

		while (var)
		{
			if (var->nstem > 1)
			{
				char	  **subres = NormalizeSubWord(Conf, var->stem[var->nstem - 1], FF_COMPOUNDWORD);

				if (subres)
				{
					char	  **subptr = subres;

					if (!lcur)
						lcur = lres = (TSLexeme *) palloc(MAX_NORM * sizeof(TSLexeme));

					while (*subptr)
					{
						for (i = 0; i < var->nstem - 1; i++)
						{
							lcur->lexeme = (subptr == subres) ? var->stem[i] : pstrdup(var->stem[i]);
							lcur->flags = 0;
							lcur->nvariant = NVariant;
							lcur++;
						}

						lcur->lexeme = *subptr;
						lcur->flags = 0;
						lcur->nvariant = NVariant;
						lcur++;
						subptr++;
						NVariant++;
					}

					lcur->lexeme = NULL;
					pfree(subres);
					var->stem[0] = NULL;
					pfree(var->stem[var->nstem - 1]);
				}
			}

			for (i = 0; i < var->nstem && var->stem[i]; i++)
				pfree(var->stem[i]);
			ptr = var->next;
			pfree(var->stem);
			pfree(var);
			var = ptr;
		}
	}
	return lres;
}


static void
freeSPNode(SPNode * node)
{
	SPNodeData *data;

	if (!node)
		return;
	data = node->data;
	while (node->length)
	{
		freeSPNode(data->node);
		data++;
		node->length--;
	}
	free(node);
}

static void
freeANode(AffixNode * node)
{
	AffixNodeData *data;

	if (!node)
		return;
	data = node->data;
	while (node->length)
	{
		freeANode(data->node);
		if (data->naff)
			free(data->aff);
		data++;
		node->length--;
	}
	free(node);
}


void
NIFree(IspellDict * Conf)
{
	int			i;
	AFFIX	   *Affix = (AFFIX *) Conf->Affix;
	char	  **aff = Conf->AffixData;

	if (aff)
	{
		while (*aff)
		{
			free(*aff);
			aff++;
		}
		free(Conf->AffixData);
	}


	for (i = 0; i < Conf->naffixes; i++)
	{
		if (Affix[i].compile == 0)
		{
			if (Affix[i].isregis)
				RS_free(&(Affix[i].reg.regis));
			else
				pg_regfree(&(Affix[i].reg.regex));
		}
		free(Affix[i].mask);
		free(Affix[i].find);
		free(Affix[i].repl);
	}
	if (Conf->Spell)
	{
		for (i = 0; i < Conf->nspell; i++)
			free(Conf->Spell[i].word);
		free(Conf->Spell);
	}

	if (Conf->Affix)
		free(Conf->Affix);
	if (Conf->CompoundAffix)
		free(Conf->CompoundAffix);
	freeSPNode(Conf->Dictionary);
	freeANode(Conf->Suffix);
	freeANode(Conf->Prefix);
	memset((void *) Conf, 0, sizeof(IspellDict));
	return;
}