Remove simpletree from the package

[ambv]

The default tree builder is now etree. I also added an "implementation" attribute to the TreeBuilder API so you can select the correct ElementTree if you have to access it. In related news, I started a change log, to be filled out before releasing 1.0 proper.

[gsnedders]

And merged in 96da7f5.

[jsocol]

Is there any discussion of why this change was made anywhere? I wasn't able to find anything on html5lib-discuss.

[jgraham]

The discussion was on IRC. Basically the feeling was that simpletree isn't production quality and there is not enough benefit in maintaining our own tree format just for testing.

[gsnedders]

And related to that, it has proved at several points to be quite a lot of work to maintain, when really it serves almost nobody: we support two tree APIs in the stdlib which should suffice in general.

[jsocol]

Well, I'll disagree with "when really it serves almost nobody" because it was incredibly helpful in Bleach, and getting around the removal (mozilla/bleach#94) is going to pin me and bleach users to 0.96 for a while. I understand the reasons, but it would be great to document hugely-API-breaking changes' decision process.

[gsnedders]

Well, what do you gain by using simpletree instead of e.g., minidom or ElementTree? I don't think there's any case in which simpletree is the right choice — thus "it serves almost nobody".

[jezdez]

This also breaks django_compressor in a non-trivial way: django-compressor/django-compressor#405

[ambv]

django-compressor/django-compressor#405 is fixed, switched to using the default etree builder and walker.

[jezdez]

Thanks, @ambv!