GoTo Line has a pause after hitting "return" (due to colorization?)

[dsyme]

Repro steps

1. Take a large file
2. Use "Go to Line" (Ctrl-G) then type a number then hit return

Expected behavior

Goes to the line very fast, then colorizes

Actual behavior

Takes a while, I think it is colorizing the file. (Is it waiting for a semantic colorization? Or just slowly doing syntax colorization)

Known workarounds

Be patient and just wait

[vasily-kirichenko]

I'm afraid we don't control this.

[dsyme]

Yeah I was wondering. Maybe it is just incentive for us to make the appropriate colorizer faster, or submit a roslyn bug to delay colorization

[majocha]

I think something's not right on our side. For comparison open ProjectNode.cs in ProjectSystem.Base. Over 6k lines and go to line is instantaneous.

[vasily-kirichenko]

The only thing I can think is syntactic colorization, which is synchronous in Roslyn (for a reason I don't fully understand, something like "we made AddSyntacticClassificationsAsync async, but call in synchronously from UI thread because we expect it to be super fast, because it's just tokenizing after all). The problem is that we need to tokenize all lines above current one in order to keep proper lexer state, so, when you jump to line #10000 from line #1, 10K lines are tokenized, which is fast, but not lighting fast.

(that's how I see it, maybe I'm wrong)

[majocha]

I suspect it's the UI of that go to popup that blocks and stutters for some reason. Try making large jumps around the code by right clicking scrollbar and selecting "scroll here" and it's always snappy.

[vasily-kirichenko]

My VFT is broken, see the latest issue. Cannot test.

[dsyme]

The only thing I can think is syntactic colorization..

I think it must be that. If you go to the last line of a large file, it's slow. Then back to the top, and try again to the last line, and its fast. That strongly seems to indicate colorization.

[vasily-kirichenko]

I reproduced it on tast.fs, it took about 2-3 seconds each time I went to line #5000. dotTrace shows this:

After about a minute this stopped happening and it jumped instantly (no CPU load were for all that time).

[dsyme]

@vasily-kirichenko I can't see anything causal on the dotTrace info - can you? Thanks

[vasily-kirichenko]

That's the problem :)

[cartermp]

cc @Pilchie

[dsyme]

I tried again and the pause seems much reduced now in large files. This is an old issue so closing