From 977a0f13ea3f242bb663bc5f9b83d08ffb2f5d28 Mon Sep 17 00:00:00 2001
From: Jan Travnicek <Jan.Travnicek@fit.cvut.cz>
Date: Wed, 27 Feb 2019 17:15:30 +0100
Subject: [PATCH] reimplement V8 unbounded regexp optimisation pattern
---
.../simplify/RegExpOptimizeUnboundedPart.hpp | 26 ++++++++++---------
1 file changed, 14 insertions(+), 12 deletions(-)
diff --git a/alib2algo/src/regexp/simplify/RegExpOptimizeUnboundedPart.hpp b/alib2algo/src/regexp/simplify/RegExpOptimizeUnboundedPart.hpp
index 5c971f0e8f..d2d3fce9c4 100644
--- a/alib2algo/src/regexp/simplify/RegExpOptimizeUnboundedPart.hpp
+++ b/alib2algo/src/regexp/simplify/RegExpOptimizeUnboundedPart.hpp
@@ -766,16 +766,16 @@ bool RegExpOptimize::V6( UnboundedRegExpAlternation < SymbolType > & /* node */
* @return bool true if optimization applied else false
*/
template < class SymbolType >
-bool RegExpOptimize::V8( UnboundedRegExpConcatenation < SymbolType > & /* node */ ) {
+bool RegExpOptimize::V8( UnboundedRegExpConcatenation < SymbolType > & node ) {
bool optimized = false;
// interpretation: if there is iteration in concatenation node, and element of iteration contains eps and is straight before this iteration, then this element can be omitted
- /*if ( node.getChildren ( ).size ( ) == 0 )
+ if ( node.getChildren ( ).size ( ) == 0 )
return false;
- for( auto it = next ( node.getChildren ( ).begin( ) ); it != node.getChildren ( ).end( ); ) {
- UnboundedRegExpIteration < SymbolType > * iter = dynamic_cast < UnboundedRegExpIteration < SymbolType > * > ( & * it );
+ for( auto it = node.getChildren ( ).begin( ); it != node.getChildren ( ).end( ); ) {
+ const UnboundedRegExpIteration < SymbolType > * iter = dynamic_cast < const UnboundedRegExpIteration < SymbolType > * > ( & * it );
if( ! iter ) {
++ it;
@@ -783,7 +783,7 @@ bool RegExpOptimize::V8( UnboundedRegExpConcatenation < SymbolType > & /* node *
}
// if element of iteration is concatenation, we need to check this specially
- UnboundedRegExpConcatenation < SymbolType > * concat = dynamic_cast < UnboundedRegExpConcatenation < SymbolType > * > ( & iter->getChild ( ) );
+ const UnboundedRegExpConcatenation < SymbolType > * concat = dynamic_cast < const UnboundedRegExpConcatenation < SymbolType > * > ( & iter->getChild ( ) );
if( concat ) {
// check if not out of bounds
@@ -800,23 +800,25 @@ bool RegExpOptimize::V8( UnboundedRegExpConcatenation < SymbolType > & /* node *
equal ( concat->getChildren ( ).begin( ), concat->getChildren ( ).end( ), it2, [ ] ( const UnboundedRegExpElement < SymbolType > & a, const UnboundedRegExpElement < SymbolType > & b ) -> bool { return a == b; } ) ) {
optimized = true;
- it = node.getChildren ( ).erase ( it2, it );
+ it = node.erase ( it2, it );
} else
++ it;
} else {
+ // check if not at the first node
+ if ( it == node.getChildren ( ).begin ( ) ) {
+ it ++;
+ continue;
+ }
+
auto prev = std::prev ( it );
if ( regexp::properties::RegExpEpsilon::languageContainsEpsilon ( iter->getElement ( ) ) && iter->getElement ( ) == * prev ) {
- it = node.getChildren ( ).erase ( prev );
+ it = node.erase ( prev );
optimized = true;
-
- // in case xxx*, we need to stay on the iter element, not to go behind it
- if( it != node.getChildren().begin( ) )
- it = std::prev( it );
} else
++ it;
}
- }*/
+ }
return optimized;
}
--
GitLab