Underscore JS is at 1.4.2 instead of 1.4.4 (newer version)

[tjoyal]

1.4.4 — Jan. 30, 2013
1.4.3 — Dec. 4, 2012

https://github.com/meteor/meteor/blob/master/packages/underscore/underscore.js

[gschmidt]

Thanks. Assuming we don't find any problems, we will do this the next time we rebuild dev_bundle (Meteor's upstream dependency kit) which will likely, but not definitely, be before the next Meteor releases. If there is a reason why this is more urgent please let us know.

[glasser]

devel is now running with Underscore 1.4.4.

[n1mmy]

We had to revert this change. 1.4.4 seems like a big performance regression. Further investigation is needed to determine why and what to do about this.

[glasser]

client-side specifically, right?

[n1mmy]

At least client side. Haven't gotten a good comparison server side yet.

[fightingtheboss]

Has there been any thought to switching to lodash.js from underscore.js? In theory it's a drop-in replacement (using the underscore build option) and the reported performance gains are significant.

[gschmidt]

@fightingtheboss Yes, definitely have been thinking about lodash, and also about providing an easy mechanism to define additional _ extensions in packages. We've even done some (very preliminary and hypothetical) thinking about automatically linking in just the lodash functions that your app (or your package dependencies) actually need.

[glasser]

Specifically, we tested lodash a month or two ago and found that it broke UglifyJS, which required us to upgrade Uglify... which we found broke Fibers, which required us to work with the maintainer of Fibers to improve Fibers... and then we haven't found time to get back to testing lodash again. Definitely interested in it, though. There are a few functions we would like to use heavily (eg we would replace all uses of _.each which don't need to be polymorphic between arrays and objects with _.forOwn which means we don't break if an object has a numeric length field).

[fightingtheboss]

I totally know how deep and dark some rabbit holes get. Looking forward to the ultimate results of all your hard work.

[alexgorbatchev]

👍 for lodash

[avital]

We're at underscore 1.5.1 now.

[jdalton]

The current stable is Underscore 1.5.2. Also, have you revisited Lo-Dash, any issues I can help with?

[glasser]

Heh, "try lodash" has continually been on the "should do this sometime when I feel like I have time for lots of QA" list :)

It's also still the case that http://lodash.com/ doesn't seem to tell me what I should expect to happen if I swap in the modern or compatibility build for Underscore: ie, it lists the extra features that lodash has, but not if any existing Underscore behavior is changed by the non-Underscore build.

[jdalton]

It's also still the case that http://lodash.com/ doesn't seem to tell me what I should expect to happen if I swap in the modern or compatibility build for Underscore: ie, it lists the extra features that lodash has, but not if any existing Underscore behavior is changed by the non-Underscore build.

It's mentioned on the custom-builds page via a reference to the wiki. There's also an extensive blog post by @kitcambridge over the various builds.

[glasser]

Ooh, that is pretty clear!

[raix]

So would lodash be before or after version 1? (still got the one old confirmed bug #594 ) and would you take pr's or have the core on this?

Nice - would we get config/options in packages or even as runtime bundles? I'm thinking the custom build looks nice - got same issue with bootstrap 3 configurations. (@gschmidt did ask for feedback)