extern int fp_Read( void *handle, const char *fname, int maxngrams )
{
fp_t *h = (fp_t *)handle;
FILE *fp;
char line[1024];
int cnt = 0;
fp = fopen( fname, "r" );
if (!fp) {
#ifdef VERBOSE
fprintf( stderr, "Failed to open fingerprint file '%s'\n", fname);
#endif
return 0;
}
h->fprint = (ngram_t *)wg_malloc(maxngrams * sizeof(ngram_t));
while (cnt < maxngrams && wg_getline(line,1024,fp)) {
char *p;
wg_trim(line, line);
p = strpbrk( line, " \t" );
if ( p ) {
*p = '\0';
}
if ( strlen(line) > MAXNGRAMSIZE ) {
continue;
}
strcpy( h->fprint[cnt].str, line );
h->fprint[cnt].rank = cnt;
cnt++;
}
h->size = cnt;
/*** Sort n-grams, for easy comparison later on ***/
qsort( h->fprint, h->size, sizeof(ngram_t), ngramcmp_str );
fclose(fp);
return 1;
}
extern void *textcat_Init( const char *conffile )
{
textcat_t *h;
char line[1024];
FILE *fp;
fp = fopen( conffile, "r" );
if ( !fp ) {
#ifdef VERBOSE
fprintf( stderr, "Failed to open config file '%s'\n", conffile);
#endif
return NULL;
}
h = (textcat_t *)wg_malloc(sizeof(textcat_t));
h->size = 0;
h->maxsize = 16;
h->fprint = (void **)wg_malloc( sizeof(void*) * h->maxsize );
while ( wg_getline( line, 1024, fp ) ) {
char *p;
char *segment[4];
int res;
/*** Skip comments ***/
#ifdef HAVE_STRCHR
if (( p = strchr(line,'#') )) {
#else
if (( p = index(line,'#') )) {
#endif
*p = '\0';
}
if ((res = wg_split( segment, line, line, 4)) < 2 ) {
continue;
}
/*** Ensure enough space ***/
if ( h->size == h->maxsize ) {
h->maxsize *= 2;
h->fprint = (void *)wg_realloc( h->fprint, sizeof(void*) * h->maxsize );
}
/*** Load data ***/
if ((h->fprint[ h->size ] = fp_Init( segment[1] ))==NULL) {
goto ERROR;
}
if ( fp_Read( h->fprint[h->size], segment[0], 400 ) == 0 ) {
textcat_Done(h);
goto ERROR;
}
h->size++;
}
fclose(fp);
return h;
ERROR:
fclose(fp);
return NULL;
}
extern char *textcat_Classify( void *handle, const char *buffer, size_t size )
{
textcat_t *h = (textcat_t *)handle;
uint4 i, cnt = 0;
int minscore = MAXSCORE;
int threshold = minscore;
char *result = h->output;
#ifdef HAVE_ALLOCA
candidate_t *candidates = (candidate_t *)alloca( sizeof(candidate_t) * h->size );
#else
candidate_t *candidates = (candidate_t *)malloc( sizeof(candidate_t) * h->size );
#define SHOULD_FREE 1
#endif
void *unknown;
unknown = fp_Init(NULL);
if ( fp_Create( unknown, buffer, size, MAXNGRAMS ) == 0 ) {
/*** Too little information ***/
result = _TEXTCAT_RESULT_SHORT;
goto READY;
}
/*** Calculate the score for each category. ***/
for (i=0; i<h->size; i++) {
int score = fp_Compare( h->fprint[i], unknown, threshold );
candidates[i].score = score;
candidates[i].name = fp_Name( h->fprint[i] );
if ( score < minscore ) {
minscore = score;
threshold = (int)( (double)score * THRESHOLDVALUE );
}
}
/*** Find the best performers ***/
for (i=0; i<h->size; i++) {
if ( candidates[i].score < threshold ) {
if ( ++cnt == MAXCANDIDATES+1 ) {
break;
}
memcpy( &candidates[cnt-1], &candidates[i], sizeof(candidate_t) );
}
}
/*** The verdict ***/
if ( cnt == MAXCANDIDATES+1 ) {
result = _TEXTCAT_RESULT_UNKOWN;
}
else {
char *p = result;
char *plimit = result+MAXOUTPUTSIZE;
qsort( candidates, cnt, sizeof(candidate_t), cmpcandidates );
*p = '\0';
for (i=0; i<cnt; i++) {
p = wg_strgmov( p, "[", plimit );
p = wg_strgmov( p, candidates[i].name, plimit );
p = wg_strgmov( p, "]", plimit );
}
}
READY:
fp_Done(unknown);
#ifdef SHOULD_FREE
free(candidates);
#undef SHOULD_FREE
#endif
return result;
}