rose: use program offset, not final_id, in atable

This removes the need to look up the program offset in a table when
handling an anchored literal match.
This commit is contained in:
Justin Viiret
2016-07-14 11:40:49 +10:00
committed by Matthew Barr
parent 4dbbc4eaa5
commit 8754cbbd24
4 changed files with 42 additions and 9 deletions

View File

@@ -204,6 +204,28 @@ void remapAnchoredReports(RoseBuildImpl &tbi) {
}
}
static
void remapIds(flat_set<ReportID> &reports, const vector<u32> &litPrograms) {
flat_set<ReportID> new_reports;
for (auto id : reports) {
assert(id < litPrograms.size());
new_reports.insert(litPrograms.at(id));
}
reports = move(new_reports);
}
/**
* \brief Replace the reports (which are literal final_ids) in the given
* raw_dfa with program offsets.
*/
static
void remapIdsToPrograms(raw_dfa &rdfa, const vector<u32> &litPrograms) {
for (dstate &ds : rdfa.states) {
remapIds(ds.reports, litPrograms);
remapIds(ds.reports_eod, litPrograms);
}
}
static
void populate_holder(const simple_anchored_info &sai, const set<u32> &exit_ids,
NGHolder *h_in) {
@@ -826,7 +848,7 @@ vector<raw_dfa> buildAnchoredDfas(RoseBuildImpl &build) {
aligned_unique_ptr<anchored_matcher_info>
buildAnchoredMatcher(RoseBuildImpl &build, vector<raw_dfa> &dfas,
size_t *asize) {
const vector<u32> &litPrograms, size_t *asize) {
const CompileContext &cc = build.cc;
if (dfas.empty()) {
@@ -835,6 +857,10 @@ buildAnchoredMatcher(RoseBuildImpl &build, vector<raw_dfa> &dfas,
return nullptr;
}
for (auto &rdfa : dfas) {
remapIdsToPrograms(rdfa, litPrograms);
}
vector<aligned_unique_ptr<NFA>> nfas;
vector<u32> start_offset; // start offset for each dfa (dots removed)
size_t total_size = buildNfas(dfas, &nfas, &start_offset, cc, build.rm);