static grn_bool
chunked_tokenize_utf8_chunk(grn_ctx *ctx,
grn_mecab_tokenizer *tokenizer,
const char *chunk,
unsigned int chunk_bytes)
{
const char *tokenized_chunk;
size_t tokenized_chunk_length;
tokenized_chunk = mecab_sparse_tostr2(tokenizer->mecab, chunk, chunk_bytes);
if (!tokenized_chunk) {
GRN_PLUGIN_ERROR(ctx, GRN_TOKENIZER_ERROR,
"[tokenizer][mecab][chunk] "
"mecab_sparse_tostr2() failed len=%d err=%s",
chunk_bytes,
mecab_strerror(tokenizer->mecab));
return GRN_FALSE;
}
if (GRN_TEXT_LEN(&(tokenizer->buf)) > 0) {
GRN_TEXT_PUTS(ctx, &(tokenizer->buf), " ");
}
tokenized_chunk_length = strlen(tokenized_chunk);
if (tokenized_chunk_length >= 1 &&
isspace(tokenized_chunk[tokenized_chunk_length - 1])) {
GRN_TEXT_PUT(ctx, &(tokenizer->buf),
tokenized_chunk, tokenized_chunk_length - 1);
} else {
GRN_TEXT_PUT(ctx, &(tokenizer->buf),
tokenized_chunk, tokenized_chunk_length);
}
return GRN_TRUE;
}
static void
check_mecab_dictionary_encoding(grn_ctx *ctx)
{
#ifdef HAVE_MECAB_DICTIONARY_INFO_T
mecab_t *mecab;
mecab = mecab_new2("-Owakati");
if (mecab) {
grn_encoding encoding;
int have_same_encoding_dictionary = 0;
encoding = GRN_CTX_GET_ENCODING(ctx);
have_same_encoding_dictionary = encoding == get_mecab_encoding(mecab);
mecab_destroy(mecab);
if (!have_same_encoding_dictionary) {
GRN_PLUGIN_ERROR(ctx, GRN_TOKENIZER_ERROR,
"[tokenizer][mecab] "
"MeCab has no dictionary that uses the context encoding"
": <%s>",
grn_enctostr(encoding));
}
} else {
GRN_PLUGIN_ERROR(ctx, GRN_TOKENIZER_ERROR,
"[tokenizer][mecab] "
"mecab_new2 failed in check_mecab_dictionary_encoding: %s",
mecab_strerror(NULL));
}
#endif
}
static const char *
mecab_global_error_message(void)
{
double version;
version = atof(mecab_version());
/* MeCab <= 0.993 doesn't support mecab_strerror(NULL). */
if (version <= 0.993) {
return "Unknown";
}
return mecab_strerror(NULL);
}
static grn_rc
mecab_init(grn_ctx *ctx, grn_obj *table, grn_proc_data *user_data)
{
grn_obj *str;
int nflags = 0;
char *buf, *s, *p;
char mecab_err[256];
grn_obj_flags table_flags;
grn_mecab_tokenizer *token;
unsigned int bufsize, maxtrial = 10, len;
if (!(str = grn_ctx_pop(ctx))) { return GRN_INVALID_ARGUMENT; }
SOLE_MECAB_CONFIRM;
if (!sole_mecab) {
GRN_LOG(ctx, GRN_LOG_ALERT, "mecab_new failed on grn_mecab_init");
return GRN_TOKENIZER_ERROR;
}
if (!(token = GRN_MALLOC(sizeof(grn_mecab_tokenizer)))) { return ctx->rc; }
user_data->ptr = token;
token->mecab = sole_mecab;
// if (!(token->mecab = mecab_new3())) {
grn_table_get_info(ctx, table, &table_flags, &token->encoding, NULL);
nflags |= (table_flags & GRN_OBJ_KEY_NORMALIZE);
if (!(token->nstr = grn_str_open_(ctx, GRN_TEXT_VALUE(str), GRN_TEXT_LEN(str),
nflags, token->encoding))) {
GRN_LOG(ctx, GRN_LOG_ALERT, "grn_str_open failed at grn_token_open");
return GRN_TOKENIZER_ERROR;
}
len = token->nstr->norm_blen;
mecab_err[sizeof(mecab_err) - 1] = '\0';
for (bufsize = len * 2 + 1; maxtrial; bufsize *= 2, maxtrial--) {
if(!(buf = GRN_MALLOC(bufsize + 1))) {
GRN_LOG(ctx, GRN_LOG_ALERT, "buffer allocation on mecab_init failed !");
GRN_FREE(token);
return ctx->rc;
}
MUTEX_LOCK(sole_mecab_lock);
s = mecab_sparse_tostr3(token->mecab, token->nstr->norm, len, buf, bufsize);
if (!s) {
strncpy(mecab_err, mecab_strerror(token->mecab), sizeof(mecab_err) - 1);
}
MUTEX_UNLOCK(sole_mecab_lock);
if (s) { break; }
GRN_FREE(buf);
if (strstr(mecab_err, "output buffer overflow") == NULL) { break; }
}
if (!s) {
GRN_LOG(ctx, GRN_LOG_ALERT, "mecab_sparse_tostr failed len=%d bufsize=%d err=%s",
len, bufsize, mecab_err);
GRN_FREE(token);
return GRN_TOKENIZER_ERROR;
}
// certain version of mecab returns trailing lf or spaces.
for (p = buf + strlen(buf) - 1;
buf <= p && (*p == '\n' || isspace(*(unsigned char *)p));
p--) { *p = '\0'; }
//grn_log("sparsed='%s'", s);
token->buf = (unsigned char *)buf;
token->next = (unsigned char *)buf;
token->end = (unsigned char *)buf + strlen(buf);
GRN_TEXT_INIT(&token->curr_, GRN_OBJ_DO_SHALLOW_COPY);
GRN_UINT32_INIT(&token->stat_, 0);
return GRN_SUCCESS;
}
/*
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)
{
grn_mecab_tokenizer *tokenizer;
unsigned int normalizer_flags = 0;
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, normalizer_flags);
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_global_error_message());
} 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_encoding_to_string(sole_mecab_encoding),
grn_encoding_to_string(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);
GRN_TEXT_INIT(&(tokenizer->buf), 0);
if (query->have_tokenized_delimiter) {
tokenizer->next = normalized_string;
tokenizer->end = tokenizer->next + normalized_string_length;
} else if (normalized_string_length == 0) {
tokenizer->next = "";
tokenizer->end = tokenizer->next;
} else {
grn_bool succeeded;
grn_plugin_mutex_lock(ctx, sole_mecab_mutex);
if (grn_mecab_chunked_tokenize_enabled &&
ctx->encoding == GRN_ENC_UTF8) {
succeeded = chunked_tokenize_utf8(ctx,
tokenizer,
normalized_string,
normalized_string_length);
} else {
const char *s;
s = mecab_sparse_tostr2(tokenizer->mecab,
normalized_string,
normalized_string_length);
if (!s) {
succeeded = GRN_FALSE;
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 {
succeeded = GRN_TRUE;
GRN_TEXT_PUTS(ctx, &(tokenizer->buf), s);
}
}
grn_plugin_mutex_unlock(ctx, sole_mecab_mutex);
if (!succeeded) {
grn_tokenizer_query_close(ctx, tokenizer->query);
GRN_PLUGIN_FREE(ctx, tokenizer);
return NULL;
}
{
char *buf, *p;
unsigned int bufsize;
buf = GRN_TEXT_VALUE(&(tokenizer->buf));
bufsize = GRN_TEXT_LEN(&(tokenizer->buf));
/* 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->next = buf;
tokenizer->end = p + 1;
}
}
user_data->ptr = tokenizer;
grn_tokenizer_token_init(ctx, &(tokenizer->token));
return NULL;
}
/*
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;
}
/*
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)
{
grn_obj *str;
int nflags = 0;
char *buf, *p;
const char *s;
grn_obj *table = args[0];
grn_obj_flags table_flags;
grn_encoding table_encoding;
grn_mecab_tokenizer *token;
unsigned int bufsize, len;
if (!(str = grn_ctx_pop(ctx))) {
ERR(GRN_INVALID_ARGUMENT, "missing argument");
return NULL;
}
if (!sole_mecab) {
CRITICAL_SECTION_ENTER(sole_mecab_lock);
if (!sole_mecab) {
sole_mecab = mecab_new2("-Owakati");
if (!sole_mecab) {
ERR(GRN_TOKENIZER_ERROR, "mecab_new2 failed on grn_mecab_init: %s",
mecab_strerror(NULL));
} else {
sole_mecab_encoding = get_mecab_encoding(sole_mecab);
}
}
CRITICAL_SECTION_LEAVE(sole_mecab_lock);
}
if (!sole_mecab) {
return NULL;
}
grn_table_get_info(ctx, table, &table_flags, &table_encoding, NULL);
if (table_encoding != sole_mecab_encoding) {
ERR(GRN_TOKENIZER_ERROR,
"MeCab dictionary charset (%s) does not match the context encoding: <%s>",
grn_enctostr(sole_mecab_encoding), grn_enctostr(table_encoding));
return NULL;
}
if (!(token = GRN_MALLOC(sizeof(grn_mecab_tokenizer)))) { return NULL; }
token->mecab = sole_mecab;
token->encoding = table_encoding;
nflags |= (table_flags & GRN_OBJ_KEY_NORMALIZE);
if (!(token->nstr = grn_str_open_(ctx, GRN_TEXT_VALUE(str), GRN_TEXT_LEN(str),
nflags, token->encoding))) {
GRN_FREE(token);
ERR(GRN_TOKENIZER_ERROR, "grn_str_open failed at grn_token_open");
return NULL;
}
len = token->nstr->norm_blen;
CRITICAL_SECTION_ENTER(sole_mecab_lock);
s = mecab_sparse_tostr2(token->mecab, token->nstr->norm, len);
if (!s) {
ERR(GRN_TOKENIZER_ERROR, "mecab_sparse_tostr failed len=%d err=%s",
len, mecab_strerror(token->mecab));
} else {
bufsize = strlen(s) + 1;
if (!(buf = GRN_MALLOC(bufsize))) {
GRN_LOG(ctx, GRN_LOG_ALERT, "buffer allocation on mecab_init failed !");
} else {
memcpy(buf, s, bufsize);
}
}
CRITICAL_SECTION_LEAVE(sole_mecab_lock);
if (!s || !buf) {
grn_str_close(ctx, token->nstr);
GRN_FREE(token);
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'; }
user_data->ptr = token;
token->buf = buf;
token->next = buf;
token->end = p + 1;
GRN_TEXT_INIT(&token->curr_, GRN_OBJ_DO_SHALLOW_COPY);
GRN_UINT32_INIT(&token->stat_, 0);
return NULL;
}