static void
grn_tokenizer_query_ensure_have_tokenized_delimiter(grn_ctx *ctx,
grn_tokenizer_query *query)
{
grn_tokenizer_query_ensure_normalized(ctx, query);
if (!query->need_delimiter_check) {
return;
}
query->need_delimiter_check = GRN_FALSE;
if (query->flags & GRN_TOKEN_CURSOR_ENABLE_TOKENIZED_DELIMITER) {
const char *normalized_string;
unsigned int normalized_string_length;
grn_string_get_normalized(ctx,
query->normalized_query,
&normalized_string,
&normalized_string_length,
NULL);
query->have_tokenized_delimiter =
grn_tokenizer_have_tokenized_delimiter(ctx,
normalized_string,
normalized_string_length,
query->encoding);
} else {
query->have_tokenized_delimiter = GRN_FALSE;
}
}
static grn_obj *
delimited_init(grn_ctx *ctx, int nargs, grn_obj **args, grn_user_data *user_data,
const uint8_t *delimiter, uint32_t delimiter_len)
{
grn_tokenizer_query *query;
unsigned int normalize_flags = 0;
const char *normalized;
unsigned int normalized_length_in_bytes;
grn_delimited_tokenizer *tokenizer;
query = grn_tokenizer_query_open(ctx, nargs, args, normalize_flags);
if (!query) {
return NULL;
}
if (!(tokenizer = GRN_MALLOC(sizeof(grn_delimited_tokenizer)))) {
ERR(GRN_NO_MEMORY_AVAILABLE,
"[tokenizer][delimit] "
"memory allocation to grn_delimited_tokenizer failed");
grn_tokenizer_query_close(ctx, query);
return NULL;
}
user_data->ptr = tokenizer;
tokenizer->query = query;
tokenizer->have_tokenized_delimiter =
grn_tokenizer_have_tokenized_delimiter(ctx,
tokenizer->query->ptr,
tokenizer->query->length,
tokenizer->query->encoding);
tokenizer->delimiter = delimiter;
tokenizer->delimiter_len = delimiter_len;
grn_string_get_normalized(ctx, tokenizer->query->normalized_query,
&normalized, &normalized_length_in_bytes,
NULL);
tokenizer->next = (const unsigned char *)normalized;
tokenizer->end = tokenizer->next + normalized_length_in_bytes;
grn_tokenizer_token_init(ctx, &(tokenizer->token));
return NULL;
}
grn_tokenizer_query *
grn_tokenizer_query_open(grn_ctx *ctx, int num_args, grn_obj **args,
unsigned int normalize_flags)
{
grn_obj *flags = grn_ctx_pop(ctx);
grn_obj *query_str = grn_ctx_pop(ctx);
grn_obj *tokenize_mode = grn_ctx_pop(ctx);
if (query_str == NULL) {
GRN_PLUGIN_ERROR(ctx, GRN_INVALID_ARGUMENT, "missing argument");
return NULL;
}
if ((num_args < 1) || (args == NULL) || (args[0] == NULL)) {
GRN_PLUGIN_ERROR(ctx, GRN_INVALID_ARGUMENT, "invalid NULL pointer");
return NULL;
}
{
grn_tokenizer_query * const query =
GRN_PLUGIN_MALLOC(ctx, sizeof(grn_tokenizer_query));
if (query == NULL) {
return NULL;
}
query->normalized_query = NULL;
query->query_buf = NULL;
if (flags) {
query->flags = GRN_UINT32_VALUE(flags);
} else {
query->flags = 0;
}
if (tokenize_mode) {
query->tokenize_mode = GRN_UINT32_VALUE(tokenize_mode);
} else {
query->tokenize_mode = GRN_TOKENIZE_ADD;
}
query->token_mode = query->tokenize_mode;
{
grn_obj * const table = args[0];
grn_obj_flags table_flags;
grn_encoding table_encoding;
unsigned int query_length = GRN_TEXT_LEN(query_str);
char *query_buf = (char *)GRN_PLUGIN_MALLOC(ctx, query_length + 1);
grn_obj *normalizer = NULL;
if (query_buf == NULL) {
GRN_PLUGIN_FREE(ctx, query);
GRN_PLUGIN_ERROR(ctx, GRN_TOKENIZER_ERROR,
"[tokenizer] failed to duplicate query");
return NULL;
}
grn_table_get_info(ctx, table, &table_flags, &table_encoding, NULL,
&normalizer, NULL);
{
grn_obj *normalized_query;
if (table_flags & GRN_OBJ_KEY_NORMALIZE) {
normalizer = GRN_NORMALIZER_AUTO;
}
normalized_query = grn_string_open_(ctx,
GRN_TEXT_VALUE(query_str),
GRN_TEXT_LEN(query_str),
normalizer,
normalize_flags,
table_encoding);
if (!normalized_query) {
GRN_PLUGIN_FREE(ctx, query_buf);
GRN_PLUGIN_FREE(ctx, query);
GRN_PLUGIN_ERROR(ctx, GRN_TOKENIZER_ERROR,
"[tokenizer] failed to open normalized string");
return NULL;
}
query->normalized_query = normalized_query;
grn_memcpy(query_buf, GRN_TEXT_VALUE(query_str), query_length);
query_buf[query_length] = '\0';
query->query_buf = query_buf;
query->ptr = query_buf;
query->length = query_length;
}
query->encoding = table_encoding;
if (query->flags & GRN_TOKEN_CURSOR_ENABLE_TOKENIZED_DELIMITER) {
const char *normalized_string;
unsigned int normalized_string_length;
grn_string_get_normalized(ctx,
query->normalized_query,
&normalized_string,
&normalized_string_length,
NULL);
query->have_tokenized_delimiter =
grn_tokenizer_have_tokenized_delimiter(ctx,
normalized_string,
normalized_string_length,
query->encoding);
} else {
query->have_tokenized_delimiter = GRN_FALSE;
}
}
return query;
}
}
/*
This function is called for a full text search query or a document to be
indexed. This means that both short/long strings are given.
The return value of this function is ignored. When an error occurs in this
function, `ctx->rc' is overwritten with an error code (not GRN_SUCCESS).
*/
static grn_obj *
mecab_init(grn_ctx *ctx, int nargs, grn_obj **args, grn_user_data *user_data)
{
char *buf, *p;
const char *s;
grn_mecab_tokenizer *tokenizer;
unsigned int bufsize;
grn_tokenizer_query *query;
grn_obj *normalized_query;
const char *normalized_string;
unsigned int normalized_string_length;
query = grn_tokenizer_query_open(ctx, nargs, args);
if (!query) {
return NULL;
}
if (!sole_mecab) {
grn_plugin_mutex_lock(ctx, sole_mecab_mutex);
if (!sole_mecab) {
sole_mecab = mecab_new2("-Owakati");
if (!sole_mecab) {
GRN_PLUGIN_ERROR(ctx, GRN_TOKENIZER_ERROR,
"[tokenizer][mecab] "
"mecab_new2() failed on mecab_init(): %s",
mecab_strerror(NULL));
} else {
sole_mecab_encoding = get_mecab_encoding(sole_mecab);
}
}
grn_plugin_mutex_unlock(ctx, sole_mecab_mutex);
}
if (!sole_mecab) {
grn_tokenizer_query_close(ctx, query);
return NULL;
}
if (query->encoding != sole_mecab_encoding) {
grn_tokenizer_query_close(ctx, query);
GRN_PLUGIN_ERROR(ctx, GRN_TOKENIZER_ERROR,
"[tokenizer][mecab] "
"MeCab dictionary charset (%s) does not match "
"the table encoding: <%s>",
grn_enctostr(sole_mecab_encoding),
grn_enctostr(query->encoding));
return NULL;
}
if (!(tokenizer = GRN_PLUGIN_MALLOC(ctx, sizeof(grn_mecab_tokenizer)))) {
grn_tokenizer_query_close(ctx, query);
GRN_PLUGIN_ERROR(ctx, GRN_NO_MEMORY_AVAILABLE,
"[tokenizer][mecab] "
"memory allocation to grn_mecab_tokenizer failed");
return NULL;
}
tokenizer->mecab = sole_mecab;
tokenizer->query = query;
normalized_query = query->normalized_query;
grn_string_get_normalized(ctx,
normalized_query,
&normalized_string,
&normalized_string_length,
NULL);
tokenizer->have_tokenized_delimiter =
grn_tokenizer_have_tokenized_delimiter(ctx,
normalized_string,
normalized_string_length,
query->encoding);
if (tokenizer->have_tokenized_delimiter) {
tokenizer->buf = NULL;
tokenizer->next = normalized_string;
tokenizer->end = tokenizer->next + normalized_string_length;
} else {
grn_plugin_mutex_lock(ctx, sole_mecab_mutex);
s = mecab_sparse_tostr2(tokenizer->mecab,
normalized_string,
normalized_string_length);
if (!s) {
GRN_PLUGIN_ERROR(ctx, GRN_TOKENIZER_ERROR,
"[tokenizer][mecab] "
"mecab_sparse_tostr() failed len=%d err=%s",
normalized_string_length,
mecab_strerror(tokenizer->mecab));
} else {
bufsize = strlen(s) + 1;
if (!(buf = GRN_PLUGIN_MALLOC(ctx, bufsize))) {
GRN_PLUGIN_LOG(ctx, GRN_LOG_ALERT,
"[tokenizer][mecab] "
"buffer allocation on mecab_init failed !");
} else {
memcpy(buf, s, bufsize);
}
}
grn_plugin_mutex_unlock(ctx, sole_mecab_mutex);
if (!s || !buf) {
grn_tokenizer_query_close(ctx, tokenizer->query);
GRN_PLUGIN_FREE(ctx, tokenizer);
return NULL;
}
/* A certain version of mecab returns trailing lf or spaces. */
for (p = buf + bufsize - 2;
buf <= p && isspace(*(unsigned char *)p);
p--) { *p = '\0'; }
tokenizer->buf = buf;
tokenizer->next = buf;
tokenizer->end = p + 1;
}
user_data->ptr = tokenizer;
grn_tokenizer_token_init(ctx, &(tokenizer->token));
return NULL;
}
