Usemin 2.0

[sleeper]

Usemin needs a big overhaul to be simpler to understand and use, and be more configurable.
Namely we need:

• configurable pipe for useminPrepare (e.g. users may want to disable uglifyjs, concat, ..)
• simpler directories (input, output) configuration:
  • differentiate location dir (i.e. where the looked at file is), search path (where to look for assets), staging dirs and output dir
• flexible parsers for usemin (e.g. add for example create a new parser for AngularJS)
• comprehensive docs about how it works and how it should be used.

[passy]

👍

[sindresorhus]

👍

I would also like to see anything that could be generic made into a separate node module.

[sleeper]

@sindresorhus Any precise idea ?

[sindresorhus]

The cssprocessor, htmlprocessor and revfinder might be useful outside of usemin if refactored, though haven't looked at them in-depth.

[sleeper]

Work in progress in branch v2.0 of sleeper/grunt-usemin

[passy]

The new flow system looks really awesome. This will make usemin a lot easier and more transparent to use.

[sleeper]

Yes, I hope so.
As well I'm working on clarifying the directories system ... Good progress on that ;)

[sindresorhus]

I also think a goal of 2.0 should be to make it faster. Currently it takes some time to run.

Some ideas:

• Cache. No need to rev an unchanged file multiple times
• Memoization on hot methods
• Could we use async.parallel on https://github.com/yeoman/grunt-usemin/blob/master/lib/htmlprocessor.js#L216-L245 ?

[sleeper]

Ok for the perfs.
Not looked at it deeply but not sure async.parallel can be used for the various regexp replacement: we can launch it and have several replacement in parallel, but the difficult part is to merge the output of each of these replacements...

[sindresorhus]

@sleeper good point, that would be difficult. maybe run multiple files in parallell then.

[sleeper]

Yes, I was more on this side ;)

Frederick Ros
Sent with Sparrow (http://www.sparrowmailapp.com/?sig)

Le jeudi 4 avril 2013 à 13:56, Sindre Sorhus a écrit :

@sleeper (https://github.com/sleeper) good point, that would be difficult. maybe run multiple files in parallell then.

—
Reply to this email directly or view it on GitHub (#85 (comment)).

[frederikb]

What's the status on 2.0?

Being able to configure the pipeline for individual blocks would be a feature very much appreciated. In my case I would like to disable uglifyjs when concatenating third party code. On top of that, if it were possible to switch out those referenced JS files with their preminified equivalents (provided that they exist, which they often do in case of Bower components) before the concatenation step, that would be even better.

I am aware of many different grunt tasks for combining Bower components, but in the end they often fail due to the fact that hardly any Bower components are correctly configured (main file definition(s), dependencies).

[necolas]

Please can we get this in a branch on this repo. Doesn't need to be in a state that's suitable for merging into master any time soon.

[sleeper]

Will do tonight.

[sleeper]

Oki doki. Branch v2.0 created and populated.

Note that a lot needs to be done yet:

• Review README
• Allow user to define and pass his own steps for the flow
• Ensure that we're capturing all the config produced by the steps of the flow (especially when they are created/furnished by the user)
• Add an option to use a map file produced by grunt-rev (the code is written to work with, only the option stuff and passing it to the task needs to be done)
• deprecate usage of require.js in blocks (see How to handle require.js in v2.0 ? #112)

Additionally, in order to be more efficient I submitted a PR to grunt-rev to have it produce a map of the transformed files.

Test are passing, but I still need to test it on real stuff ;)

[sindresorhus]

• make sure changes in master since the 2.0 rewrite begun is reintegrated

[sleeper]

Agreed. Also:

• document how to move from usemin v0.11 to v2.0

[sleeper]

Added option to look for file in a map object instead of on file

[sindresorhus]

@sleeper what's the status of the rewrite? :)

[sleeper]

What is missing so far:

• remove support of requirejs in blocks
• update/beautify documentation
• test it really (I mean ask people to help by testing it on real cases)
• ensure all the stuff that has been merged on master is present somehow on this branch
• Explain how to migrate the config from previous version to this one

We also need to push to have the PR for grunt-rev merged in master

[necolas]

Would it be worth breaking any of the more generic tasks into separate packages, like grunt-rev was.

[sleeper]

Yes, I was thinking about this as well... Let's say that in a first phase, I would like to be sure it really works, and then split what can be split ... Not sure I have a lot of generic stuff there though ;)

[sleeper]

OK, so last changes:

• support for requireJS in blocks has been removed
• usemin will now first look if a map object is available in grunt.filerev.summary otherwise it will look on the disk. You can still override the map object by using the revmap option)

[robwierzbowski]

@sleeper: Will usemin blocks in 2.0 have the ability to set alternate search paths like the ones described in yeoman/yeoman#959?

[sleeper]

Humm .. this is not exactly what I have ... the alternate path are used as alternate search path, not alternate output path ...
What is the use case behind this ?

[robwierzbowski]

I think those describe alternate search paths. So

<!-- build:js(app,.tmp) scripts/tmp.js -->
...
<!-- endbuild -->

would look in .tmp then app for a script or stylesheet to add to the concatenation array. It addresses a couple of existing issues and would allow the build process and the server process to follow the same pattern in generators with preprocessors.

[sleeper]

Oh, my fault I didn't read properly the issue you were mentioning.
AFAIR alternate path are already working in standard usemin and should in usemin v2.0 ;)

[robwierzbowski]

From what I see it doesn't cascade (app then .tmp), which is useful if you need to concatenate preprocessed css/js with vanilla css/js that lives in app.

Would it be better for me to open a separate issue with use case, references, etc?

[sleeper]

yep, it would be simple ;)
I'm starting to give a "real life" run to usemin v2.0 (e.g. real world testing ;)), so I should wipe some issues

[robwierzbowski]

Thanks! New issue posted.

[sindresorhus]

@sleeper hey man. How far along is the rewrite at this point?

[sleeper]

I would say a couple of things to review/fix, plus rebasing on master to get the patch that have been incorporated on master since the initial branching ... which is not really cool, considering some people have probably checked out this branch .. but considering the huge amount of changes this is the only suitable way to do it ...

[sindresorhus]

That's great to hear. Would be nice to get it into generator-webapp soon to test it out ;)

[sleeper]

OK, so I invoked the wrath of Git Gods by squashing and rebasing v2.0 on master.
As some files have disappeared some PR are not "merged" anymore (#106, #137, #141). I'll need to re-implement this in v2.0.

[guifromrio]

Many anxious users ready to rock and roll version 2!

On Fri, Jul 19, 2013 at 8:42 PM, Frederick Ros notifications@github.comwrote:

OK, so I invoked the wrath of Git Gods by squashing and rebasing v2.0 on
master.
As some files have disappeared some PR are not "merged" anymore (#106#106,
#137 #137, #141#141).
I'll need to re-implement this in v2.0.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/85#issuecomment-21283738
.

[sleeper]

OK, support for #106 and #141 are now part of v2.0

[sindresorhus]

@sleeper how do you feel about doing the 2.0 release this/next week?

Usemin 1.0 has a lot of problems and would be a great help to people if they got 2.0 ;)

[guifromrio]

There's sure lots of folk dying in anticipation. I'm one of them!

[sleeper]

Sorry for the delay in answering, I have/had lot of work plus additional things that kept me far from my duties ;)

From my point of view v2.0 is pretty stable... Most of the last fixes on the master branch have been back ported to v2.0.
Some additional stuff are not yet done (like the conditional stuff for IE).

In my view, we should ask for a wider test by integrating it with generator-webapp for example.

[sindresorhus]

Thanks for the update @sleeper and no worries :)

Anyone up for doing a PR on generator-webapp?

[sindresorhus]

// @kevva @stephenplusplus @passy @addyosmani

[kevva]

What's to be updated? Seems like it should work out of the box. Maybe we need to change dirs to assetsDirs in the usemin task but other than that there shouldn't be any super big changes.

[sindresorhus]

@kevva i haven't looked at it closely, but you're probably right. Most important thing is testing all edge cases and reporting back if something isn't working.

[kevva]

Implemented in yeoman/generator-webapp@40a2595. Could we release this soon @sleeper?

[guifromrio]

Christmas morning feelings.

[sleeper]

Dudes, sorry for being so absent these days.
I'm going to try merging the 2.0 branch in master tomorrow morning (Paris time ;)).

[sleeper]

Version 0.1.13 published: last head version before merge ;)

[sleeper]

Holly crap dudes: v2.0 has been merged in master and has been published in npm.

Enjoy !!

[kevva]

👏

[sindresorhus]

Awesome! Thanks for doing that @sleeper :D

We should get out a new release of generator-webapp