/* Run a query and return a list of ranked results.
* Call this from Perl like `$index->fetch(['water', 'jar'])`. The result
* will be an arrayref, with each entry of it being an arrayref of the form
* [id, rank]. They will be sorted by their rank in descending order, so the
* most relevant document will be at the top. */
SV* fetch(SV* tokens) const
{
AV* av = reinterpret_cast<AV*>(SvRV(tokens));
std::unordered_map<std::string, int> query;
for (int i = 0; i <= av_top_index(av); ++i)
{
std::string token = string_from_sv(*av_fetch(av, i, 0));
++query[token];
}
AV* results = newAV();
for (const auto& id2rank : search(query))
{
AV* entry = newAV();
double length = lengths.find(id2rank.first)->second;
av_push(entry, newSViv(id2rank.first));
av_push(entry, newSVnv(id2rank.second / length));
av_push(results, newRV_noinc(reinterpret_cast<SV*>(entry)));
}
sortsv(AvARRAY(results), av_top_index(results) + 1, sort_ratings);
return newRV_noinc(reinterpret_cast<SV*>(results));
}
/* Add a document with the given ID and arrayref of tokens to the index.
* Each time this is called, the given ID must be greater than the previous
* one. The IDs need not be sequential, however.
* Call this from Perl like `$index->add_document(123, ['cup', 'tea'])`. */
void add_document(int id, SV* tokens)
{
std::unordered_map<std::string, int> vec;
AV* av = reinterpret_cast<AV*>(SvRV(tokens));
for (int i = 0; i <= av_top_index(av); ++i)
{
std::string token = string_from_sv(*av_fetch(av, i, 0));
++vec[token];
}
for (const auto& token2tf : vec)
{ index[token2tf.first].push_back({id, token2tf.second}); }
lengths[id] = 0;
}
static void xs_new(pTHX_ SV *cv) {
dXSARGS;
if (items < 1) croak("Usage: class, ref");
#ifndef MULTIPLICITY
AV *slots = CvXSUBANY(cv).any_ptr;
#else
MAGIC *mg = mg_findext(cv, PERL_MAGIC_ext, &ATTRS_TBL);
AV *slots = (AV *)mg->mg_obj;
#endif
SV *class = ST(0);
HV *hash = newHV();
SV *obj = sv_2mortal(newRV_noinc((SV *)hash)); // don't move to the end(leaks)
SV **args; // uniq args
int args_count = args_to_uniq(&ST(1), items - 1, &args); // skip 1(class)
int slots_count = av_top_index(slots) + 1;
for (int i = 0; i < slots_count; i++) { // NEXT_SLOT:
ECAslot *slot = sv2slot(av_fetch_or_croak(slots, i));
// iterage args, null if matched
for (int j = 0; j < args_count; j += 2) {
SV *tmp = args[j];
if (!tmp) continue; // already matched
if (!sv_cmp(tmp, slot->key)) {
if (slot->check) do_check(slot->check, args[j + 1], slot->key);
hv_he_store_or_croak(hash, slot->key, args[j + 1]);
args[j] = NULL;
goto NEXT_SLOT;
args[j] = NULL; // mark as consumed
}
}
// slot not found in passed args, decide what to do
if (slot->type == ECA_REQUIRED) {
croak("Attribute \"%s\" is required", SvPV_nolen(slot->key));
} else if (slot->type == ECA_DEFAULT) {
hv_he_store_or_croak(hash, slot->key, slot->value);
} else if (slot->type == ECA_DEFAULT_CODE) {
invoke_and_store(class, slot->value, hash, slot->key);
}
NEXT_SLOT:; // simulate continue label
}
I32 p5_av_top_index(PerlInterpreter *my_perl, AV *av) {
PERL_SET_CONTEXT(my_perl);
return av_top_index(av);
}
I32 p5_av_top_index(PerlInterpreter *my_perl, AV *av) {
return av_top_index(av);
}
