Hash spec (for recent IRC-spawned commit)

[alexch]

I figure the best way to understand a new project is to dive in and start writing tests (and fixing bugs). When I was on IRC I noticed this commit ee0caf1 come through with no tests, so...

I hope you don't mind if I submit a stream of patches while I get my bearings. (Don't expect me to be your test monkey forever, though.)

[meh]

Could you amend the commits with a more meaningful message please?

One would have to check the specific commit to know what's going on.

[alexch]

How's that?

[meh]

I'd remove the unit test for X entirely, just Add tests for Array#select passing params to the block and Add tests for Hash.new with native object containing null.

Sorry for the bikeshedding.

[alexch]

Well, as long as we're bikeshedding...

Don't you think it's useful to have a link back to the original commit,
which has more information and context, for a pure unit test commit?

What if I put (see xyz123) in the commit body?

[meh]

I personally dislike binding commits to other commits, or tests to commits. The feature is there, and the unit tests are there to ensure there are no future regressions about the feature, no matter in what commit they were introduced or changed.

[alexch]

Okey doke.

[adambeynon]

I'm not sure Hash.new() with a native object is a good idea. We are currently converting null to nil, but we should really be converting all inline objects to hashes, as well as converting each array item to a ruby object if needed.

This is already done by JSON, so this partial implementation of that on hash seems firstly a duplication, and secondly not a full implementation of something useful.

[meh]

@adambeynon JSON doesn't just convert the object to a Hash, it also calls the json_class hooks.

But I think we should cleanup Native and move it to the stdlib.