int DpsSEAMake(DPS_AGENT *Indexer, DPS_DOCUMENT *Doc, DPS_DSTR *excerpt,
const char *content_lang, size_t *indexed_size, size_t *indexed_limit,
size_t max_word_len, size_t min_word_len, int crossec, int seasec
#ifdef HAVE_ASPELL
, int have_speller, AspellSpeller *speller
#endif
) {
DPS_SENTENCELIST List;
DPS_MAPSTAT MapStat;
DPS_TEXTITEM Item;
DPS_VAR *Sec;
dpsunicode_t *sentence, *lt, savec;
double *links, *lang_cs, w;
double delta, pdiv, cur_div;
size_t l, sent_len, order;
size_t min_len = 10000000, min_pos = 0;
int it;
register size_t i, j;
#ifdef DEBUG
char lcstr[4096];
#endif
TRACE_IN(Indexer, "DpsSEAMake");
if((Sec = DpsVarListFind(&Doc->Sections, "sea"))) { /* set SEA section to NULL */
DPS_FREE(Sec->val);
DPS_FREE(Sec->txt_val);
Sec->curlen = 0;
}
bzero(&List, sizeof(List));
order = 0;
sentence = DpsUniStrTok_SEA((dpsunicode_t*)excerpt->data, <);
while(sentence) {
if (lt != NULL) { savec = *lt; *lt = 0; }
#ifdef DEBUG
DpsConv(&Indexer->uni_lc, lcstr, sizeof(lcstr), (char*)sentence, sizeof(dpsunicode_t) * (DpsUniLen(sentence) + 1));
fprintf(stderr, "Sentence.%d: %s\n", List.nitems, lcstr);
#endif
if ((sent_len = DpsUniLen(sentence)) >= Indexer->Flags.SEASentenceMinLength) {
j = 1;
for (i = 0; i < List.nitems; i++) {
if (DpsUniStrCmp(sentence, List.Sent[i].sentence) == 0) {
j = 0; break;
}
}
if (j) {
if ( List.nitems < Indexer->Flags.SEASentences ) {
if (List.nitems == List.mitems) {
List.mitems += 16;
List.Sent = (DPS_SENTENCE*)DpsRealloc(List.Sent, List.mitems * sizeof(DPS_SENTENCE));
if (List.Sent == NULL) { TRACE_OUT(Indexer); return DPS_ERROR;}
}
List.Sent[List.nitems].sentence = DpsUniDup(sentence);
List.Sent[List.nitems].len = sent_len;
List.Sent[List.nitems].order = order++;
sentence = DpsUniDup(sentence);
DpsUniStrToLower(sentence);
bzero(&List.Sent[List.nitems].LangMap, sizeof(DPS_LANGMAP));
DpsBuildLangMap(&List.Sent[List.nitems].LangMap, (char*)sentence, sent_len * sizeof(dpsunicode_t), 0, 0);
if (sent_len < min_len) { min_len = sent_len; min_pos = List.nitems; }
List.nitems++;
DPS_FREE(sentence);
} else if (sent_len > min_len) {
DPS_FREE(List.Sent[min_pos].sentence);
List.Sent[min_pos].sentence = DpsUniDup(sentence);
List.Sent[min_pos].len = sent_len;
List.Sent[min_pos].order = order++;
sentence = DpsUniDup(sentence);
DpsUniStrToLower(sentence);
bzero(&List.Sent[min_pos].LangMap, sizeof(DPS_LANGMAP));
DpsBuildLangMap(&List.Sent[min_pos].LangMap, (char*)sentence, sent_len * sizeof(dpsunicode_t), 0, 0);
DPS_FREE(sentence);
min_len = List.Sent[0].len; min_pos = 0;
for(i = 1; i < List.nitems; i++) if (List.Sent[i].len < min_len) { min_len = List.Sent[i].len; min_pos = i; }
}
}
}
#ifdef DEBUG
fprintf(stderr, "Sent. len.:%d, Min.allowed: %d\n", sent_len, Indexer->Flags.SEASentenceMinLength);
#endif
if (lt != NULL) *lt = savec;
sentence = DpsUniStrTok_SEA(NULL, <);
}
DpsLog(Indexer, DPS_LOG_DEBUG, "SEA sentences: %d", List.nitems);
if (List.nitems < 4) {
for (i = 0; i < List.nitems; i++) DPS_FREE(List.Sent[i].sentence);
DPS_FREE(List.Sent);
TRACE_OUT(Indexer);
return DPS_OK;
}
links = (double*)DpsMalloc(sizeof(double) * List.nitems * List.nitems);
lang_cs = (double*)DpsMalloc(sizeof(double) * List.nitems);
/*
k ot
links[i * List.nitems + j]
*/
if (links != NULL && lang_cs != NULL) {
for (i = 0; i < List.nitems; i++) {
DpsPrepareLangMap(&List.Sent[i].LangMap);
}
for (i = 0; i < List.nitems; i++) {
List.Sent[i].Oi = List.Sent[i].di = 0.5;
if (Doc->lang_cs_map == NULL) {
links[i * List.nitems + i] = 0.0;
} else {
MapStat.map = &List.Sent[i].LangMap;
DpsCheckLangMap6(Doc->lang_cs_map, &List.Sent[i].LangMap, &MapStat, DPS_LM_TOPCNT * DPS_LM_TOPCNT, 2 * DPS_LM_TOPCNT);
links[i * List.nitems + i] = (double)MapStat.hits / (2.0 * DPS_LM_TOPCNT) / (List.nitems + 1);
}
#ifdef DEBUG
DpsLog(Indexer, DPS_LOG_INFO, "Link %u->%u: %f [hits:%d miss:%d]", i, i, links[i * List.nitems + i], MapStat.hits, MapStat.miss);
#endif
for (j = 0; j < List.nitems; j++) {
if (j == i) continue;
MapStat.map = &List.Sent[j].LangMap;
DpsCheckLangMap6(&List.Sent[j].LangMap, &List.Sent[i].LangMap, &MapStat, DPS_LM_TOPCNT * DPS_LM_TOPCNT, 2 * DPS_LM_TOPCNT);
links[i * List.nitems + j] = (double)MapStat.hits / (2.0 * DPS_LM_TOPCNT) / (List.nitems + 1);
#ifdef DEBUG
DpsLog(Indexer, DPS_LOG_INFO, "Link %u->%u: %f [hits:%d miss:%d]", i, j, links[i * List.nitems + j], MapStat.hits, MapStat.miss);
#endif
}
}
for (l = 0; l < List.nitems; l++) {
w = 0.0;
for (i = 0; i < List.nitems; i++) {
w += links[l * List.nitems + i] * List.Sent[i].Oi;
}
w = f(w);
if (w < LOW_BORDER_EPS2) w = LOW_BORDER_EPS2;
else if (w > HI_BORDER_EPS2) w = HI_BORDER_EPS2;
List.Sent[l].di = w;
}
DpsSort(List.Sent, List.nitems, sizeof(DPS_SENTENCE), (qsort_cmp)SentCmp);
#ifdef DEBUG
DpsConv(&Indexer->uni_lc, lcstr, sizeof(lcstr), (char*)List.Sent[0].sentence, sizeof(dpsunicode_t) * (DpsUniLen(List.Sent[0].sentence) + 1));
fprintf(stderr, "Sent.0: %f %f -- %s\n", List.Sent[0].di, List.Sent[0].Oi, lcstr);
DpsConv(&Indexer->uni_lc, lcstr, sizeof(lcstr), (char*)List.Sent[1].sentence, sizeof(dpsunicode_t) * (DpsUniLen(List.Sent[1].sentence) + 1));
fprintf(stderr, "Sent.1: %f %f -- %s\n", List.Sent[1].di, List.Sent[1].Oi, lcstr);
DpsConv(&Indexer->uni_lc, lcstr, sizeof(lcstr), (char*)List.Sent[2].sentence, sizeof(dpsunicode_t) * (DpsUniLen(List.Sent[2].sentence) + 1));
fprintf(stderr, "Sent.2: %f %f -- %s\n", List.Sent[2].di, List.Sent[2].Oi, lcstr);
DpsConv(&Indexer->uni_lc, lcstr, sizeof(lcstr), (char*)List.Sent[3].sentence, sizeof(dpsunicode_t) * (DpsUniLen(List.Sent[3].sentence) + 1));
fprintf(stderr, "Sent.3: %f %f -- %s\n", List.Sent[3].di, List.Sent[3].Oi, lcstr);
DpsConv(&Indexer->uni_lc, lcstr, sizeof(lcstr), (char*)List.Sent[4].sentence, sizeof(dpsunicode_t) * (DpsUniLen(List.Sent[4].sentence) + 1));
fprintf(stderr, "Sent.4: %f %f -- %s\n", List.Sent[4].di, List.Sent[4].Oi, lcstr);
#endif
DpsSort(List.Sent, TOP_SENTENCES, sizeof(DPS_SENTENCE), (qsort_cmp)SentOrderCmp);
bzero(&Item, sizeof(Item));
Item.section = seasec;
Item.href = NULL;
Item.section_name = "sea";
for (i = 0; i < TOP_SENTENCES; i++) {
dpsunicode_t *UStr = DpsUniDup(List.Sent[i].sentence);
DpsPrepareItem(Indexer, Doc, &Item, List.Sent[i].sentence, UStr, content_lang, indexed_size, indexed_limit,
max_word_len, min_word_len, crossec
#ifdef HAVE_ASPELL
, have_speller, speller, NULL
#endif
);
DPS_FREE(UStr);
}
}
DPS_FREE(lang_cs);
DPS_FREE(links);
for (i = 0; i < List.nitems; i++) DPS_FREE(List.Sent[i].sentence);
DPS_FREE(List.Sent);
TRACE_OUT(Indexer);
return DPS_OK;
}
__C_LINK int __DPSCALL DpsImportDictionary(DPS_ENV * Conf, const char *lang, const char *charset,
const char *filename, int skip_noflag, const char *first_letters){
struct stat sb;
char *str, *data = NULL, *cur_n = NULL;
char *lstr;
dpsunicode_t *ustr;
DPS_CHARSET *sys_int;
DPS_CHARSET *dict_charset;
DPS_CONV touni;
DPS_CONV fromuni;
int fd;
char savebyte;
if ((lstr = (char*) DpsMalloc(2048)) == NULL) {
DPS_FREE(str);
return DPS_ERROR;
}
if ((ustr = (dpsunicode_t*) DpsMalloc(8192)) == NULL) {
DPS_FREE(lstr);
return DPS_ERROR;
}
dict_charset = DpsGetCharSet(charset);
sys_int = DpsGetCharSet("sys-int");
if ((dict_charset == NULL) || (sys_int == NULL)) {
DPS_FREE(lstr);
DPS_FREE(ustr);
return DPS_ERROR;
}
DpsConvInit(&touni, dict_charset, sys_int, Conf->CharsToEscape, 0);
DpsConvInit(&fromuni, sys_int, dict_charset, Conf->CharsToEscape, 0);
if (stat(filename, &sb)) {
fprintf(stderr, "Unable to stat synonyms file '%s': %s", filename, strerror(errno));
DPS_FREE(lstr);
DPS_FREE(ustr);
return DPS_ERROR;
}
if ((fd = DpsOpen2(filename, O_RDONLY)) <= 0) {
fprintf(stderr, "Unable to open synonyms file '%s': %s", filename, strerror(errno));
return DPS_ERROR;
}
if ((data = (char*)DpsMalloc(sb.st_size + 1)) == NULL) {
fprintf(stderr, "Unable to alloc %ld bytes", (long)sb.st_size);
DpsClose(fd);
DPS_FREE(lstr);
DPS_FREE(ustr);
return DPS_ERROR;
}
if (read(fd, data, sb.st_size) != (ssize_t)sb.st_size) {
fprintf(stderr, "Unable to read synonym file '%s': %s", filename, strerror(errno));
DPS_FREE(data);
DpsClose(fd);
DPS_FREE(lstr);
DPS_FREE(ustr);
return DPS_ERROR;
}
data[sb.st_size] = '\0';
str = data;
cur_n = strchr(str, '\n');
if (cur_n != NULL) {
cur_n++;
savebyte = *cur_n;
*cur_n = '\0';
}
DpsClose(fd);
while(str != NULL) {
char *s;
const char *flag;
int res;
flag = NULL;
s = str;
while(*s){
if(*s == '\r') *s = '\0';
if(*s == '\n') *s = '\0';
s++;
}
if((s=strchr(str,'/'))){
*s=0;
s++;flag=s;
while(*s){
if(((*s>='A')&&(*s<='Z'))||((*s>='a')&&(*s<='z')))s++;
else{
*s=0;
break;
}
}
}else{
if(skip_noflag) goto loop_continue;
flag="";
}
res = DpsConv(&touni, (char*)ustr, 8192, str, 1024);
DpsUniStrToLower(ustr);
/* Dont load words if first letter is not required */
/* It allows to optimize loading at search time */
if(*first_letters) {
DpsConv(&fromuni, lstr, 2048, ((const char*)ustr),(size_t)res);
if(!strchr(first_letters,lstr[0]))
goto loop_continue;
}
res = DpsSpellAdd(&Conf->Spells,ustr,flag,lang);
if (res != DPS_OK) {
DPS_FREE(lstr);
DPS_FREE(ustr); DPS_FREE(data);
return res;
}
if (Conf->Flags.use_accentext) {
dpsunicode_t *af_uwrd = DpsUniAccentStrip(ustr);
if (DpsUniStrCmp(af_uwrd, ustr) != 0) {
res = DpsSpellAdd(&Conf->Spells, af_uwrd, flag, lang);
if (res != DPS_OK) {
DPS_FREE(lstr);
DPS_FREE(ustr); DPS_FREE(data); DPS_FREE(af_uwrd);
return res;
}
}
DPS_FREE(af_uwrd);
if (strncasecmp(lang, "de", 2) == 0) {
dpsunicode_t *de_uwrd = DpsUniGermanReplace(ustr);
if (DpsUniStrCmp(de_uwrd, ustr) != 0) {
res = DpsSpellAdd(&Conf->Spells, de_uwrd, flag, lang);
if (res != DPS_OK) {
DPS_FREE(lstr);
DPS_FREE(ustr); DPS_FREE(data); DPS_FREE(de_uwrd);
return res;
}
}
DPS_FREE(de_uwrd);
}
}
loop_continue:
str = cur_n;
if (str != NULL) {
*str = savebyte;
cur_n = strchr(str, '\n');
if (cur_n != NULL) {
cur_n++;
savebyte = *cur_n;
*cur_n = '\0';
}
}
}
DPS_FREE(data);
DPS_FREE(lstr);
DPS_FREE(ustr);
return DPS_OK;
}
__C_LINK int __DPSCALL DpsSynonymListLoad(DPS_ENV * Env,const char * filename){
struct stat sb;
char *str, *data = NULL, *cur_n = NULL;
char lang[64]="";
DPS_CHARSET *cs=NULL;
DPS_CHARSET *sys_int=DpsGetCharSet("sys-int");
DPS_CONV file_uni;
DPS_WIDEWORD *ww = NULL;
size_t key = 1;
int flag_th = 0;
int fd;
char savebyte;
if (stat(filename, &sb)) {
fprintf(stderr, "Unable to stat synonyms file '%s': %s", filename, strerror(errno));
return DPS_ERROR;
}
if ((fd = DpsOpen2(filename, O_RDONLY)) <= 0) {
dps_snprintf(Env->errstr,sizeof(Env->errstr)-1, "Unable to open synonyms file '%s': %s", filename, strerror(errno));
return DPS_ERROR;
}
if ((data = (char*)DpsMalloc(sb.st_size + 1)) == NULL) {
dps_snprintf(Env->errstr,sizeof(Env->errstr)-1, "Unable to alloc %d bytes", sb.st_size);
DpsClose(fd);
return DPS_ERROR;
}
if (read(fd, data, sb.st_size) != (ssize_t)sb.st_size) {
dps_snprintf(Env->errstr,sizeof(Env->errstr)-1, "Unable to read synonym file '%s': %s", filename, strerror(errno));
DPS_FREE(data);
DpsClose(fd);
return DPS_ERROR;
}
data[sb.st_size] = '\0';
str = data;
cur_n = strchr(str, '\n');
if (cur_n != NULL) {
cur_n++;
savebyte = *cur_n;
*cur_n = '\0';
}
while(str != NULL) {
if(str[0]=='#'||str[0]==' '||str[0]=='\t'||str[0]=='\r'||str[0]=='\n') goto loop_continue;
if(!strncasecmp(str,"Charset:",8)){
char * lasttok;
char * charset;
if((charset = dps_strtok_r(str + 8, " \t\n\r", &lasttok))) {
cs=DpsGetCharSet(charset);
if(!cs){
dps_snprintf(Env->errstr, sizeof(Env->errstr), "Unknown charset '%s' in synonyms file '%s'",
charset, filename);
DPS_FREE(data);
DpsClose(fd);
return DPS_ERROR;
}
DpsConvInit(&file_uni, cs, sys_int, Env->CharsToEscape, 0);
}
}else
if(!strncasecmp(str,"Language:",9)){
char * lasttok;
char * l;
if((l = dps_strtok_r(str + 9, " \t\n\r", &lasttok))) {
dps_strncpy(lang, l, sizeof(lang)-1);
}
}else
if(!strncasecmp(str, "Thesaurus:", 10)) {
char * lasttok;
char *tok = dps_strtok_r(str + 10, " \t\n\r", &lasttok);
flag_th = (strncasecmp(tok, "yes", 3) == 0) ? 1 : 0;
}else{
char *av[255];
size_t ac, i, j;
dpsunicode_t *t;
if(!cs){
dps_snprintf(Env->errstr,sizeof(Env->errstr)-1,"No Charset command in synonyms file '%s'",filename);
DpsClose(fd); DPS_FREE(data);
return DPS_ERROR;
}
if(!lang[0]){
dps_snprintf(Env->errstr,sizeof(Env->errstr)-1,"No Language command in synonyms file '%s'",filename);
DpsClose(fd); DPS_FREE(data);
return DPS_ERROR;
}
ac = DpsGetArgs(str, av, 255);
if (ac < 2) goto loop_continue;
if ((ww = (DPS_WIDEWORD*)DpsRealloc(ww, ac * sizeof(DPS_WIDEWORD))) == NULL) return DPS_ERROR;
for (i = 0; i < ac; i++) {
ww[i].word = av[i];
ww[i].len = dps_strlen(av[i]);
ww[i].uword = t = (dpsunicode_t*)DpsMalloc((3 * ww[i].len + 1) * sizeof(dpsunicode_t));
if (ww[i].uword == NULL) return DPS_ERROR;
DpsConv(&file_uni, (char*)ww[i].uword, sizeof(dpsunicode_t) * (3 * ww[i].len + 1), av[i], ww[i].len + 1);
DpsUniStrToLower(ww[i].uword);
ww[i].uword = DpsUniNormalizeNFC(NULL, ww[i].uword);
DPS_FREE(t);
}
for (i = 0; i < ac - 1; i++) {
for (j = i + 1; j < ac; j++) {
if((Env->Synonyms.nsynonyms + 1) >= Env->Synonyms.msynonyms){
Env->Synonyms.msynonyms += 64;
Env->Synonyms.Synonym = (DPS_SYNONYM*)DpsRealloc(Env->Synonyms.Synonym,
sizeof(DPS_SYNONYM)*Env->Synonyms.msynonyms);
if (Env->Synonyms.Synonym == NULL) {
Env->Synonyms.msynonyms = Env->Synonyms.nsynonyms = 0;
return DPS_ERROR;
}
}
bzero((void*)&Env->Synonyms.Synonym[Env->Synonyms.nsynonyms], sizeof(DPS_SYNONYM));
/* Add direct order */
Env->Synonyms.Synonym[Env->Synonyms.nsynonyms].p.uword = DpsUniDup(ww[i].uword);
Env->Synonyms.Synonym[Env->Synonyms.nsynonyms].s.uword = DpsUniDup(ww[j].uword);
Env->Synonyms.Synonym[Env->Synonyms.nsynonyms].p.count =
Env->Synonyms.Synonym[Env->Synonyms.nsynonyms].s.count = (size_t)((flag_th) ? key : 0);
Env->Synonyms.nsynonyms++;
bzero((void*)&Env->Synonyms.Synonym[Env->Synonyms.nsynonyms], sizeof(DPS_SYNONYM));
/* Add reverse order */
Env->Synonyms.Synonym[Env->Synonyms.nsynonyms].p.uword = DpsUniDup(ww[j].uword);
Env->Synonyms.Synonym[Env->Synonyms.nsynonyms].s.uword = DpsUniDup(ww[i].uword);
Env->Synonyms.Synonym[Env->Synonyms.nsynonyms].p.count =
Env->Synonyms.Synonym[Env->Synonyms.nsynonyms].s.count = (size_t)((flag_th) ? key : 0);
Env->Synonyms.nsynonyms++;
}
}
for (i = 0; i < ac; i++) {
DPS_FREE(ww[i].uword);
}
do { key++; } while (key == 0);
}
loop_continue:
str = cur_n;
if (str != NULL) {
*str = savebyte;
cur_n = strchr(str, '\n');
if (cur_n != NULL) {
cur_n++;
savebyte = *cur_n;
*cur_n = '\0';
}
}
}
DPS_FREE(data);
DPS_FREE(ww);
DpsClose(fd);
return DPS_OK;
}
