/**
* This function analyses the inputs of all the tags of the given .fst2 in
* order to determine their kind. 'tokens' contains all the text tokens.
* After the execution of the function, 'number_of_patterns' will contain
* the number of patterns found in the grammar, and 'is_DIC'/'is_CDIC'/'is_SDIC'/'is_TDIC'
* will contain 1 if the tag 'DIC'/'CDIC'/'SDIC'/'TDIC' has been found. See
* the comment below about the special case for '<!DIC>'.
*/
void process_tags(int *number_of_patterns,
struct string_hash* semantic_codes,
int *is_DIC,int *is_CDIC,
int *is_SDIC,struct locate_parameters* parameters,
Abstract_allocator prv_alloc) {
(*number_of_patterns)=0;
(*is_DIC)=0;
(*is_CDIC)=0;
(*is_SDIC)=0;
Fst2* fst2=parameters->fst2;
struct string_hash* tokens=parameters->tokens;
Fst2Tag* tag=fst2->tags;
/* We get the number of the SPACE token */
unichar t[2];
t[0]=' ';
t[1]='\0';
parameters->SPACE=get_value_index(t,tokens,DONT_INSERT);
/* Then, we test all the tags */
for (int i=0;i<fst2->number_of_tags;i++) {
if (tag[i]->type!=UNDEFINED_TAG) {
/* We don't need to process again things like variables and contexts
* that have already been processed at the time of loading the fst2 */
continue;
}
int length=u_strlen(tag[i]->input);
if (!u_strcmp(tag[i]->input,"#")) {
/* If we have a #, we must check if it is the meta one that
* forbids space or the "#" token */
if (is_bit_mask_set(tag[i]->control,RESPECT_CASE_TAG_BIT_MASK)) {
/* If the case respect bit is set to 1, then we have the "#" token */
tag[i]->type=PATTERN_TAG;
tag[i]->pattern=build_token_pattern(tag[i]->input,prv_alloc);
}
else {
/* If we have the meta # */
tag[i]->type=META_TAG;
tag[i]->meta=META_SHARP;
}
}
else if (!u_strcmp(tag[i]->input,"<E>")) {
/* If we have a transition tagged by the empty word epsilon */
tag[i]->type=META_TAG;
tag[i]->meta=META_EPSILON;
}
else {
int token_number=get_value_index(tag[i]->input,tokens,DONT_INSERT);
if (token_number!=-1) {
/* If the input is an existing token */
if (token_number==parameters->SPACE) {
/* If it is a space */
tag[i]->type=META_TAG;
tag[i]->meta=META_SPACE;
} else {
/* If it is a normal token */
tag[i]->type=PATTERN_TAG;
tag[i]->pattern=build_token_pattern(tag[i]->input,prv_alloc);
}
}
else {
/* This input is not an existing token. Two cases can happen:
* 1) metas like <!MOT> or patterns like <V:K>
* 2) a word that is not in the text tokens */
if (tag[i]->input[0]!='<' || tag[i]->input[length-1]!='>') {
/* If we are in case 2, it may not be an error. For instance,
* if the tag contains "foo" and if it is a tag that allows
* case variations, we could match "FOO" if this token is in the
* text. */
tag[i]->type=PATTERN_TAG;
tag[i]->pattern=build_token_pattern(tag[i]->input,prv_alloc);
} else {
/* If we have something of the form <...>, we must test first if it is
* or not a negative tag like <!XXX> */
int negative_tag=(tag[i]->input[1]=='!')?1:0;
if (negative_tag) {
set_bit_mask(&(tag[i]->control),NEGATION_TAG_BIT_MASK);
}
/* Then, we must test if we have or not a meta. To do that, we
* extract the content without < > and ! if any.*/
unichar* content=u_strdup(&(tag[i]->input[1+negative_tag]),length-2-negative_tag,prv_alloc);
/* And we test all the possible metas */
if (!u_strcmp(content,"MOT")) {
tag[i]->type=META_TAG;
tag[i]->meta=META_MOT;
}
else if (!u_strcmp(content,"DIC")) {
tag[i]->type=META_TAG;
tag[i]->meta=META_DIC;
if (!negative_tag) {
/* We mark that the DIC tag has been found, but only
* if it is not the negative one (<!DIC>). We do this
* because things matched by <DIC> will be taken from
* the 'dlf' and 'dlc' files, whereas things matched by <!DIC>
* will be taken from the 'err' file. Such a trick is necessary
* if we don't want 'priori' to be taken as an unknown word since
* it is part of the compound word 'a priori' */
(*is_DIC)=1;
}
}
else if (!u_strcmp(content,"CDIC")) {
tag[i]->type=META_TAG;
tag[i]->meta=META_CDIC;
(*is_CDIC)=1;
}
else if (!u_strcmp(content,"SDIC")) {
tag[i]->type=META_TAG;
tag[i]->meta=META_SDIC;
(*is_SDIC)=1;
}
else if (!u_strcmp(content,"TDIC")) {
tag[i]->type=META_TAG;
tag[i]->meta=META_TDIC;
}
else if (!u_strcmp(content,"MAJ")) {
tag[i]->type=META_TAG;
tag[i]->meta=META_MAJ;
}
else if (!u_strcmp(content,"MIN")) {
tag[i]->type=META_TAG;
tag[i]->meta=META_MIN;
}
else if (!u_strcmp(content,"PRE")) {
tag[i]->type=META_TAG;
tag[i]->meta=META_PRE;
}
else if (!u_strcmp(content,"NB")) {
tag[i]->type=META_TAG;
tag[i]->meta=META_NB;
if (negative_tag) {
error("Negative mark will be ignored in <!NB>\n");
}
}
else if (!u_strcmp(content,"TOKEN")) {
tag[i]->type=META_TAG;
tag[i]->meta=META_TOKEN;
}
else {
/* If we arrive here, we have not a meta but a pattern like
* <be>, <be.V>, <V:K>, ... */
tag[i]->type=PATTERN_TAG;
tag[i]->pattern=build_pattern(content,semantic_codes,parameters->tilde_negation_operator,prv_alloc);
if (tag[i]->pattern->type==CODE_PATTERN ||
tag[i]->pattern->type==LEMMA_AND_CODE_PATTERN ||
tag[i]->pattern->type==FULL_PATTERN ||
tag[i]->pattern->type==INFLECTED_AND_LEMMA_PATTERN) {
/* If the pattern we obtain contains grammatical/semantic
* codes, then we put it in the pattern tree and we note its number. */
tag[i]->pattern_number=add_pattern(number_of_patterns,tag[i]->pattern,parameters->pattern_tree_root,prv_alloc);
if (tag[i]->pattern->type==CODE_PATTERN) {
/* If we have a code pattern, then the tag will just need to contain
* the pattern number, BUT, WE DO NOT FREE THE PATTERN,
* since this pattern still could be used in morphological mode */
tag[i]->type=PATTERN_NUMBER_TAG;
}
}
}
/* We don't forget to free the content */
free_cb(content,prv_alloc);
}
}
}
}
}
static
error_t parse_opt(int key, char *arg, struct argp_state *state)
{
struct cmdctx *cmdctx = state->input;
switch (key) {
case 'a':
cmdctx->_flags |= OPT_SEARCH_ALL;
break;
case 'c':
cmdctx->_flags |= OPT_SEARCH_CNFL;
break;
case 'l':
case 'f':
cmdctx->_flags |= OPT_SEARCH_FL;
break;
case 'g':
cmdctx->_flags |= OPT_SEARCH_GROUP;
break;
case 'o':
cmdctx->_flags |= OPT_SEARCH_OBSL;
break;
case 'p':
cmdctx->_flags |= OPT_SEARCH_CAP;
break;
case 'r':
cmdctx->_flags |= OPT_SEARCH_REQ;
break;
case 'S':
cmdctx->_flags |= OPT_SEARCH_SUGS;
break;
case 's':
cmdctx->_flags |= OPT_SEARCH_SUMM;
break;
case 'd':
cmdctx->_flags |= OPT_SEARCH_DESC;
break;
case 'L':
cmdctx->_flags |= OPT_SEARCH_CHANGELOG;
break;
case OPT_PATTERN_PCRE:
cmdctx->_flags |= OPT_PATTERN_PCRE;
break;
case ARGP_KEY_ARG:
if (arg == NULL)
break;
DBGF("search.arg (%s)\n", arg);
if (cmdctx->_data != NULL) { /* already got pattern */
poldek_ts_add_pkgmask(cmdctx->ts, arg);
} else {
struct pattern *pt;
if ((pt = build_pattern(cmdctx, arg)) == NULL) {
argp_usage(state);
return EINVAL;
}
cmdctx->_data = pt;
cmdctx->rtflags |= CMDCTX_GOTARGS;
}
break;
default:
return ARGP_ERR_UNKNOWN;
}
return 0;
}
