small touchups to the aspell suggestions code

This commit is contained in:
Jean-Francois Dockes 2012-10-01 17:28:30 +02:00
parent 8f1b706b81
commit 4f0ccb69b7
2 changed files with 22 additions and 26 deletions

View file

@ -258,8 +258,11 @@ public:
{}
void newData() {
while (m_db.termWalkNext(m_tit, *m_input)) {
if (!Rcl::Db::isSpellingCandidate(*m_input))
LOGDEB2(("Aspell::buildDict: term: [%s]\n", m_input->c_str()));
if (!Rcl::Db::isSpellingCandidate(*m_input)) {
LOGDEB2(("Aspell::buildDict: SKIP\n"));
continue;
}
#ifndef RCL_INDEX_STRIPCHARS
if (!o_index_stripchars) {
string lower;
@ -270,7 +273,7 @@ public:
#endif
// Got a non-empty sort-of appropriate term, let's send it to
// aspell
LOGDEB2(("ASpExecPv: [%s]\n", m_input->c_str()));
LOGDEB2(("Apell::buildDict: SEND\n"));
m_input->append("\n");
return;
}
@ -411,14 +414,13 @@ bool Aspell::suggest(Rcl::Db &db, const string &_term,
AspellStringEnumeration *els = aapi.aspell_word_list_elements(wl);
const char *word;
while ((word = aapi.aspell_string_enumeration_next(els)) != 0) {
// stemDiffers checks that the word exists (we don't want
// aspell computed stuff, only exact terms from the dictionary),
// and that it stems differently to the base word (else it's not
// useful to expand the search). Or is it ?
// ******** This should depend if
// stemming is turned on or not for querying *******
string sw(word);
if (db.termExists(sw) && db.stemDiffers("english", sw, mterm))
// Check that the word exists in the index (we don't want
// aspell computed stuff, only exact terms from the
// dictionary). We used to also check that it stems
// differently from the base word but this is complicated
// (stemming on/off + language), so we now leave this to the
// caller.
if (db.termExists(word))
suggestions.push_back(word);
}
aapi.delete_aspell_string_enumeration(els);

View file

@ -190,32 +190,26 @@ void QtGuiResListPager::suggest(const vector<string>uterms,
LOGERR(("QtGuiResListPager:: aspell not initialized\n"));
return;
}
for (vector<string>::const_iterator uit = uterms.begin();
uit != uterms.end(); uit++) {
list<string> asuggs;
string reason;
// If the term is in the index, we don't suggest alternatives.
// Actually, we may want to check the frequencies and propose something
// anyway if a possible variation is much more common (as google does)
#warning need to take case and diacs sensibility into account somehow
// Maybe use the xapian index instead ? How to retrieve the
// sensitivity flags ?
// We used to call aspell->check() here and continue if it
// succeeded. but this does not work if we are in
// case-sensitive mode and the term was not found because of a
// case difference (our aspell is all lowercase).
// if (aspell->check(*uit, reason))
// continue;
// else if (!reason.empty())
// return;
// If the term is in the dictionary, Aspell::suggest won't
// list alternatives. In fact we may want to check the
// frequencies and propose something anyway if a possible
// variation is much more common (as google does) ?
if (!aspell->suggest(*rcldb, *uit, asuggs, reason)) {
LOGERR(("QtGuiResListPager::suggest: aspell failed: %s\n",
reason.c_str()));
continue;
}
// We should check that the term stems differently from the
// base word (else it's not useful to expand the search). Or
// is it ? This should depend if stemming is turned on or not
if (!asuggs.empty()) {
sugg[*uit] = vector<string>(asuggs.begin(), asuggs.end());
if (sugg[*uit].size() > 5)