Summary: | Code-Assist (ctrl+space) to slow with jre-src | ||||||
---|---|---|---|---|---|---|---|
Product: | [Eclipse Project] JDT | Reporter: | Martin Möbius <m.moebius> | ||||
Component: | Core | Assignee: | David Audel <david_audel> | ||||
Status: | RESOLVED FIXED | QA Contact: | |||||
Severity: | normal | ||||||
Priority: | P1 | CC: | martinae, vladgri | ||||
Version: | 2.0 | ||||||
Target Milestone: | 2.0 M2 | ||||||
Hardware: | PC | ||||||
OS: | Windows NT | ||||||
Whiteboard: | |||||||
Attachments: |
|
Description
Martin Möbius
2001-12-12 12:48:49 EST
problem confirmed. In the JTable scenario below it takes 1s without source attachment and 6s+ with source attachment. The JavaDoc extraction kicks in after the proposals are computed and after the pop-up shows up. I don't think it is the culprit. Before perf tracing in the UI Philippe can you pls confirm that the proposal computation for a binary class with source attachment is as fast as proposal computation without source attachment. If we have source attached, then it might take longer to extract parameter names (need to parse attached sources to find them). The attached trace shows that 80% of the time are spent in the code completion engine. Moving to core The scenario is main() { JTable table; table.<code assist> Created attachment 175 [details]
optimize it trace
The claim is that no matter how many times we try, it remains slow. If the file is big (as it is), it should cost to populate the JavaModel once exactly, from there on it should be fast. Seems the parameter name computation is causing grief. Please investigate. Might also want to look at 7034 which has the same symptoms. receiver type file is read in jar file for each method of this type. fixed. |