Community
Participate
Working Groups
It seems that the parse with Nashorm+Esprima takes a lot of times with big file. I have tried to open the big JS file https://github.com/Microsoft/TypeScript/blob/master/lib/tsserver.js with the JSDT 2.0.0 Editor (getted source code from the git master) and it takes between 20-30 seconds (multiply with 2 because it is parsed twice) although with JSDT 1.x it takes less 2 seconds. I have tested with Windows 64 bits + jre1.8.0_77. I have copied/pasted nashorn.jar in the org.eclipse.wst.jsdt.core project (is it the problem?). If you need more info, don't hesitate to tell me.
tssserver.js is about 2.5 Mb. The resulting parsed JSON is 33Mb, so you can imagine while it takes so long to parse and then convert to the dom model. Of course it could be done asynchronously, but anyway this Java->JS->Java->JS conversion seems to be not a very good idea. And that's why using Google closure compiler looks like better solution.
Thank's Eugene for your feedack. Have you tested with Google closure compiler? Which times take it?
(In reply to Eugene Melekhov from comment #1) > tssserver.js is about 2.5 Mb. > The resulting parsed JSON is 33Mb, so you can imagine while it takes so long > to parse and then convert to the dom model. > In our current nashorn based solution Esprima model is not converted to JSON we basically traverse java representation of the esprima parser. I have done some testing with vanilla esprima and it actually takes time to parse the file. I suspect conversion to DOM AST is less of a problem in this case.
(In reply to Gorkem Ercan from comment #3) > In our current nashorn based solution Esprima model is not converted to JSON > we basically traverse java representation of the esprima parser. It's java representation of JS object with all that string property names etc and traversing it with string comparison etc really takes time. Here are results for opening and closing that file in JSDT editor 3 times First number is parsing time in seconds Second number is translation time in seconds Parsing finished 10.738 7.854 Parsing finished 18.736 8.316 Parsing finished 12.929 5.261 ------- Parsing finished 6.194 4.826 Parsing finished 6.036 3.830 ------- Parsing finished 4.562 4.982 Parsing finished 4.675 3.942 Of course first run is slower, others are better. Funny though that file is parsed 3 times at first run and twice at subsequent ones. > I have done > some testing with vanilla esprima and it actually takes time to parse the > file. I suspect conversion to DOM AST is less of a problem in this case. And as you can see conversion takes almost the same time as parsing, which seems "not right" :-)
