Use less memory when compiling content extensions
https://bugs.webkit.org/show_bug.cgi?id=143857
Reviewed by Benjamin Poulain.
When compiling a content extension, we convert the rule list into several intermediate forms:
1) A String.
2) A JSValue from JSONParse in loadEncodedRules.
3) A Vector of ContentExtensionRules.
4) A CombinedURLFilters object representing the pieces of the regular expressions from the triggers.
5) A Vector of NFAs.
6) A DFA for each NFA.
7) A Vector of DFABytecode.
Each one of these contains all the information contained in the content extension,
so we do not need to keep them all in memory at the same time like we are doing now.
When we are done with one, we can free that memory to greatly reduce the maximum memory usage while compiling.
The next step will be to reduce the copies of the original JSON String and to generate NFAs one at a time.
* contentextensions/CombinedURLFilters.cpp:
(WebCore::ContentExtensions::CombinedURLFilters::clear):
* contentextensions/CombinedURLFilters.h:
* contentextensions/ContentExtensionCompiler.cpp:
(WebCore::ContentExtensions::compileRuleList):
Clear structures when finished using them.
git-svn-id: https://svn.webkit.org/repository/webkit/trunk@182935
268f45cc-cd09-0410-ab3c-
d52691b4dbfc