[wtp-dev] Question on AST creation for decoration-like jobs

[Victor Rubezhny <vrubezhny@xxxxxxxxxx>]

Chris, Nitin, and JSDT adopters!

I've got a question to JSDT Developers (more experienced than me) and to 
adopters that use JSDT in/for their projects.

What is the point of resolving bindings each time we need to create an 
AST when reconcile? (It happens for almost each character typed in by a 
user or the selection is changed in the editor).

The method 
org.eclipse.wst.jsdt.internal.ui.javaeditor.ASTProvider.createAST(IJavaScriptElement, 
IProgressMonitor) is used in many operations and actions, such as 
StructuredSelectionAction, SemanticHighlightingReconciler, 
OverrideIndicatorManager, SelectionListenerWithASTManager, 
OverrideIndicatorLabelDecoration and way many more. And each time it's 
executed it tries to resolve bindings, because it's set up to do so:

556:       final ASTParser parser = ASTParser.newParser(SHARED_AST_LEVEL);
557:       parser.setResolveBindings(true);
558: parser.setStatementsRecovery(SHARED_AST_STATEMENT_RECOVERY);
559:       parser.setBindingsRecovery(SHARED_BINDING_RECOVERY);

Do we really need the bindings to be resolved when calculating the 
semantic highlighting, for example?

When I turned off the parser's settings that are set up in lines 557-558 
(actually by commenting these lines) I'm getting the AST produced almost 
8 times faster and no more StackOverflowErrors in "Semantic Highlighting 
Job", "Requesting JavaScript AST from selection" and "Decoration 
Calculation" Jobs.
OK, to be honest, not all the StackOverflowErrors have gone, but most of 
them have gone.

I'm going to disable the bindings resolving when creating an AST in 
ASTProvider.createAST(...) method or create an additional methods for 
the Jobs like a decoration one in order to improve the JavaScript Editor 
performance. So, I'd like to know which way is better: to completely 
disable the bindings resolving in createAST(...) method, or to create a 
special methods for AST creation and use that special method in 
'decoration'-like jobs.

Can you tell me what's going to be broken if I'll go the first way (by 
just commenting out the lines 557-559)? Or what I should test with care 
to be sure that I didn't break something?

I've tested it briefly in my environment and I didn't spot any problems, 
while the performance was really boosted for some test files. Also, not 
a JUnit test forced to fail after such a change in my environment.

Thanks in advance,
Victor