Increase delay between blocks of typesetting to improve responsiveness

[dpvc]

It might be worth trying out a longer delay between spurts of typesetting to see if that improves the browser's responsiveness during the typesetting phase. This may make the typesetting take longer, since there will be longer delays during the process, but currently the responsiveness is pretty bad. Not sure if it will really help, because when the typesetting os a single equation gets to be long enough, there isn't much we can do to be responsive while that is going on.

Davide

[shogun70]

Is testing this simply a matter of skipping typesetting, running pre-processing, then finding each equation and manually processing them according to some programmable schedule?

[dpvc]

No, it is easier than that. In MathJax.js, lines 1386 and 1387 are

 if (new Date().getTime() - start > this.processUpdateTime && i < scripts.length)
   {start = 0; this.RestartAfter(MathJax.Callback.Delay(1))}

which are what cause MathJax to pause briefly to allow the screen to update (to show newly typeset equations or respond to user interaction). This happens when the typesetting time has exceeded a threshold given by processUpdateTime (half a second by default). This batches together screen updates when the equations are able to processed quickly (which helped with the IE8 problems when I added a similar feature to jsMath), but still allows an incremental update of the mathematics in the page.

We should look into changing Delay(1) to be a longer time, and need to test what that should really be (e.g., Delay(10), Delay(100) or whatever) consistent with allowing some responsiveness while not making the typesetting process excessively longer.

Davide

[dpvc]

Well, I did some tests, and it looks like changing both the delay and the processUpdateTime may be the solution to responsiveness. The delay actually seemed to have little effect, but cutting processUpdateTime by half gave a noticable improvement in response to user interaction during typesetting. I'm not sure if there is any advantage to increasing the delay or not.

[dpvc]

OOPS, forgot I had started this issue already, so I started another. This is handled in the code submitted in issue #20, so I'm closing this issue.