github_mirrors/vectorscan
3
0
Fork 1
mirror of https://github.com/VectorCamp/vectorscan.git synced 2025-06-28 16:41:01 +03:00
Code Issues Packages Projects Releases Wiki Activity

Remove "dot" entries from leftfix lookarounds

Browse Source 
Note that we have to be careful to leave the first lookaround entry in
place, if it's a dot. This should eventually be done with a program
instruction.
This commit is contained in:
Justin Viiret 2016-01-06 11:45:31 +11:00 committed by Matthew Barr
parent e92a20e5fa
commit e051077a26
1 changed files with 25 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
src/rose/rose_build_lookaround.cpp
View File

@ -582,6 +582,30 @@ bool getTransientPrefixReach(const NGHolder &g, u32 lag,
return true;
}
static
void normaliseLeftfix(map<s32, CharReach> &look) {
// We can erase entries where the reach is "all characters", except for the
// very first one -- this might be required to establish a minimum bound on
// the literal's match offset.
// TODO: It would be cleaner to use a literal program instruction to check
// the minimum bound explicitly.
if (look.empty()) {
return;
}
const auto earliest = begin(look)->first;
vector<s32> dead;
for (const auto &m : look) {
if (m.second.all() && m.first != earliest) {
dead.push_back(m.first);
}
}
erase_all(&look, dead);
}
bool makeLeftfixLookaround(const RoseBuildImpl &build, const RoseVertex v,
vector<LookEntry> &lookaround) {
lookaround.clear();
@ -606,6 +630,7 @@ bool makeLeftfixLookaround(const RoseBuildImpl &build, const RoseVertex v,
}
trimLiterals(build, v, look);
normaliseLeftfix(look);
if (look.size() > MAX_LOOKAROUND_ENTRIES) {
DEBUG_PRINTF("lookaround too big (%zu entries)\n", look.size());