#1054 Quicker portable versions

[Hofman]

I have noticed that texstudio is slower and slower, since about a year, perhaps a bit more.
I try several portable version on my windows xp sp3 on a thinkpad x201, and with the current version 2.10 with a virgin profile, with around 17 documents, it takes so much time, more than 1 minute, to cut and paste about 500 lines, out of my document of 2800 lines, that it becomes unberable. Note that I have 17 documents precisely because I am forced to cut my main document into smaler ones, because texstudio is slow. Normally, I have only 7 documents open, but the slow response is still here...
With version 2.10 and virgin profile, and only my documents of 2800 lines, it takes a few dozens of second to cut the 500 lines.

I tried version also 2.7.0 and 2.8.8, each one portable, and it gets faster, with the 2.70 the fastest.
the version 2.7.0 with virgin profile again, takes not even 10 secondes, with the same 17 documents, to cut the 500 lines out od the 2800 lines.

So can there be a portable version with faster GUI?

[Tim Hoffmann]

Parsing LaTeX is quite compliciated due to the definition of the language. Over the last versions we've rewritten and improved our parser to be more flexible. Therefore, the parser can now handle more situations correctly, and also we've increased the features. This comes at the prize of slightly increased computation time. So far, performance has been good enough for most use cases (depends on the hardware and the document(s) to be processed). Therefore, our main focus is correct parsing and features. Unless we hear from more users with performance issues, we likely won't start optimizing. It would be a time-consuming task and we have to set priorities since there's always more to do than there's time.

[Benito van der Zander]

One step would be to stop using strings as hash keys (possibleCommands)

Only use ints, preferable in an array

[Tim Hoffmann]

That would be a possibility. However dictionary lookups can be quite fast. Therefore I'm not conviced that this is necessarily bottleneck. No idea if it would be worth the effort. The only way to reliably improve performance is to profile and identify the bottlenecks.