1048843 - [Messages] Investigate about concatenating/minifying CSS

Status: RESOLVED WONTFIX

(Firefox OS Graveyard :: Gaia::SMS, defect)

Description

[Julien Wajsberg [:julienw]]

In this bug we want to see if having several CSS files has an impact at launch time (as compared to one concatenated CSS).

If this has an impact, we'll do a build-time optimization.

Comment 1

[Steve Chung [:steveck]]

Manual testing with YUI compressor(minify + concatenate) result based on light workload/flame/master build: average time is 526ms(compressed) : 564ms(without compress). Applying the compressor only reduce ~40ms, seems not an impressive result... :-/ So minifying CSS might not be the most important issue for performance improvement.  BTW, the load time without loading any css is 464ms, maybe we should also focus on fine-tune the css selector part(like shorten the long selector) for performance improvement in the future.

Comment 2

[Julien Wajsberg [:julienw]]

Attachment: github PR

tentative patch to minify CSS

Not sure at all this works properly yet.

Comment 3

[Julien Wajsberg [:julienw]]

Comment on attachment 8469293 [details] [review]
github PR

Hey Yuren,

So, this seems to work fine on all applications. What do you think?

I don't see a very noticeable improvement in load time, using my own eyes, but I didn't try using the perf tools. Eli, maybe you can help testing this?

Comment 4

[[e-mail]]

We've done some CSS minify experiment a year ago. But after packaging, the result of loadtime improvement is not that noticeable :(

Comment 5

[Yuren [:yurenju] (AFK ~ Oct. 12) (volunteer mode)]

Comment on attachment 8469293 [details] [review]
github PR

pull request looks good but I recommand to do more performance measurement before we implement it.

we have implemented css minify on bug 855989 and don't have impressive improvement. so it would be great to use other css minify tool for some app and see if it's worth to implement.

Comment 6

[:Eli Perelman]

Comment on attachment 8469293 [details] [review]
github PR

Using:
  make reference-workload-light

Against master:
  "mozPerfDurationsAverage": 1906.1065674

Against master with GAIA_OPTIMIZE=1:
  "mozPerfDurationsAverage": 1838.8873350333333

Against master+patch with GAIA_OPTIMIZE=1:
  "mozPerfDurationsAverage": 1876.0379218

Comment 7

[Julien Wajsberg [:julienw]]

Thanks Eli,

then it's quite clear it doesn't change anything, will close this bug then.