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

fdr_streaming_compile: modernise loops, etc

Browse Source
This commit is contained in:
Justin Viiret 2016-03-02 10:15:48 +11:00 committed by Matthew Barr
parent 57cd2331f5
commit 466fc940e5
1 changed files with 28 additions and 47 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

75
src/fdr/fdr_streaming_compile.cpp
View File

@ -94,14 +94,13 @@ static
bool setupLongLits(const vector<hwlmLiteral> &lits, bool setupLongLits(const vector<hwlmLiteral> &lits,
vector<hwlmLiteral> &long_lits, size_t max_len) { vector<hwlmLiteral> &long_lits, size_t max_len) {
long_lits.reserve(lits.size()); long_lits.reserve(lits.size());
for (vector<hwlmLiteral>::const_iterator it = lits.begin(); for (const auto &lit : lits) {
it != lits.end(); ++it) { if (lit.s.length() > max_len) {
if (it->s.length() > max_len) { hwlmLiteral tmp = lit; // copy
hwlmLiteral tmp = *it; // copy tmp.s.pop_back();
tmp.s.erase(tmp.s.size() - 1, 1); // erase last char
tmp.id = 0; // recalc later tmp.id = 0; // recalc later
tmp.groups = 0; // filled in later by hash bucket(s) tmp.groups = 0; // filled in later by hash bucket(s)
long_lits.push_back(tmp); long_lits.push_back(move(tmp));
} }
} }
@ -112,15 +111,12 @@ bool setupLongLits(const vector<hwlmLiteral> &lits,
// sort long_literals by caseful/caseless and in lexicographical order, // sort long_literals by caseful/caseless and in lexicographical order,
// remove duplicates // remove duplicates
stable_sort(long_lits.begin(), long_lits.end(), LongLitOrder()); stable_sort(long_lits.begin(), long_lits.end(), LongLitOrder());
vector<hwlmLiteral>::iterator new_end = auto new_end = unique(long_lits.begin(), long_lits.end(), hwlmLitEqual);
unique(long_lits.begin(), long_lits.end(), hwlmLitEqual);
long_lits.erase(new_end, long_lits.end()); long_lits.erase(new_end, long_lits.end());
// fill in ids; not currently used // fill in ids; not currently used
for (vector<hwlmLiteral>::iterator i = long_lits.begin(), for (auto i = long_lits.begin(), e = long_lits.end(); i != e; ++i) {
e = long_lits.end(); i->id = distance(long_lits.begin(), i);
i != e; ++i) {
i->id = i - long_lits.begin();
} }
return true; return true;
} }
@ -143,23 +139,19 @@ void analyzeLits(const vector<hwlmLiteral> &long_lits, size_t max_len,
hashedPositions[m] = 0; hashedPositions[m] = 0;
} }
for (vector<hwlmLiteral>::const_iterator i = long_lits.begin(), for (auto i = long_lits.begin(), e = long_lits.end(); i != e; ++i) {
e = long_lits.end();
i != e; ++i) {
if (i->nocase) { if (i->nocase) {
boundaries[CASEFUL] = verify_u32(i - long_lits.begin()); boundaries[CASEFUL] = verify_u32(distance(long_lits.begin(), i));
break; break;
} }
} }
for (vector<hwlmLiteral>::const_iterator i = long_lits.begin(), for (const auto &lit : long_lits) {
e = long_lits.end(); MODES m = lit.nocase ? CASELESS : CASEFUL;
i != e; ++i) { for (u32 j = 1; j < lit.s.size() - max_len + 1; j++) {
MODES m = i->nocase ? CASELESS : CASEFUL;
for (u32 j = 1; j < i->s.size() - max_len + 1; j++) {
hashedPositions[m]++; hashedPositions[m]++;
} }
positions[m] += i->s.size(); positions[m] += lit.s.size();
} }
for (u32 m = CASEFUL; m < MAX_MODES; m++) { for (u32 m = CASEFUL; m < MAX_MODES; m++) {
@ -209,18 +201,15 @@ void fillHashes(const vector<hwlmLiteral> &long_lits, size_t max_len,
map<u32, deque<pair<u32, u32> > > bucketToLitOffPairs; map<u32, deque<pair<u32, u32> > > bucketToLitOffPairs;
map<u32, u64a> bucketToBitfield; map<u32, u64a> bucketToBitfield;
for (vector<hwlmLiteral>::const_iterator i = long_lits.begin(), for (const auto &lit : long_lits) {
e = long_lits.end(); if ((m == CASELESS) != lit.nocase) {
i != e; ++i) {
const hwlmLiteral &l = *i;
if ((m == CASELESS) != i->nocase) {
continue; continue;
} }
for (u32 j = 1; j < i->s.size() - max_len + 1; j++) { for (u32 j = 1; j < lit.s.size() - max_len + 1; j++) {
u32 h = hashLit(l, j, max_len, m); u32 h = hashLit(lit, j, max_len, m);
u32 h_ent = h & ((1U << nbits) - 1); u32 h_ent = h & ((1U << nbits) - 1);
u32 h_low = (h >> nbits) & 63; u32 h_low = (h >> nbits) & 63;
bucketToLitOffPairs[h_ent].push_back(make_pair(i->id, j)); bucketToLitOffPairs[h_ent].emplace_back(lit.id, j);
bucketToBitfield[h_ent] |= (1ULL << h_low); bucketToBitfield[h_ent] |= (1ULL << h_low);
} }
} }
@ -231,11 +220,9 @@ void fillHashes(const vector<hwlmLiteral> &long_lits, size_t max_len,
// sweep out bitfield entries and save the results swapped accordingly // sweep out bitfield entries and save the results swapped accordingly
// also, anything with bitfield entries is put in filledBuckets // also, anything with bitfield entries is put in filledBuckets
for (map<u32, u64a>::const_iterator i = bucketToBitfield.begin(), for (const auto &m : bucketToBitfield) {
e = bucketToBitfield.end(); const u32 &bucket = m.first;
i != e; ++i) { const u64a &contents = m.second;
u32 bucket = i->first;
u64a contents = i->second;
tab[bucket].bitfield = contents; tab[bucket].bitfield = contents;
filledBuckets.set(bucket); filledBuckets.set(bucket);
} }
@ -243,12 +230,9 @@ void fillHashes(const vector<hwlmLiteral> &long_lits, size_t max_len,
// store out all our chains based on free values in our hash table. // store out all our chains based on free values in our hash table.
// find nearest free locations that are empty (there will always be more // find nearest free locations that are empty (there will always be more
// entries than strings, at present) // entries than strings, at present)
for (map<u32, deque<pair<u32, u32> > >::iterator for (auto &m : bucketToLitOffPairs) {
i = bucketToLitOffPairs.begin(), u32 bucket = m.first;
e = bucketToLitOffPairs.end(); deque<pair<u32, u32>> &d = m.second;
i != e; ++i) {
u32 bucket = i->first;
deque<pair<u32, u32> > &d = i->second;
// sort d by distance of the residual string (len minus our depth into // sort d by distance of the residual string (len minus our depth into
// the string). We need to put the 'furthest back' string first... // the string). We need to put the 'furthest back' string first...
@ -299,9 +283,8 @@ void fillHashes(const vector<hwlmLiteral> &long_lits, size_t max_len,
static static
size_t maxMaskLen(const vector<hwlmLiteral> &lits) { size_t maxMaskLen(const vector<hwlmLiteral> &lits) {
size_t rv = 0; size_t rv = 0;
vector<hwlmLiteral>::const_iterator it, ite; for (const auto &lit : lits) {
for (it = lits.begin(), ite = lits.end(); it != ite; ++it) { rv = max(rv, lit.msk.size());
rv = max(rv, it->msk.size());
} }
return rv; return rv;
} }
@ -407,9 +390,7 @@ fdrBuildTableStreaming(const vector<hwlmLiteral> &lits,
ptr += litTabSize; ptr += litTabSize;
map<u32, u32> litToOffsetVal; map<u32, u32> litToOffsetVal;
for (vector<hwlmLiteral>::const_iterator i = long_lits.begin(), for (auto i = long_lits.begin(), e = long_lits.end(); i != e; ++i) {
e = long_lits.end();
i != e; ++i) {
u32 entry = verify_u32(i - long_lits.begin()); u32 entry = verify_u32(i - long_lits.begin());
u32 offset = verify_u32(ptr - secondaryTable.get()); u32 offset = verify_u32(ptr - secondaryTable.get());