Inconsistent naming with defaults

[pelson]

@msarahan & @kalefranz - please be aware that conda-forge packaged ruamel.yaml several months ago and it appears that you guys have recently added it to defaults, but with an inconsistent name ruamel_yaml. In the interests of maintaining consistent naming schemes (currently a requirement to ensure we don't blow different versions over the top of one another) would you be willing to consider renaming your package? If it means anything, it is worth noting that the name that you have chosen in defaults is inconsistent with the experience on pip also.

Thanks!

[jakirkham]

Friendly nudge @msarahan and @kalefranz.

[msarahan]

Sadly, they are different packages. Here's a bit of history: https://bitbucket.org/ruamel/yaml/issues/28/consider-making-ruamelyaml-available-as-a

ruamel_yaml is a version that Continuum modified to not be a namespace package. Its source is at https://bitbucket.org/kalefranz/yaml - the recipe is at https://github.com/ContinuumIO/anaconda-recipes/blob/master/ruamel_yaml/meta.yaml

[jakirkham]

So what you are saying is you Continuum forked ruamel.yaml and renamed it to ruamel_yaml? Is that correct?

[msarahan]

"you" being Continuum, not me personally, yes.

[jakirkham]

Yes, that's what I meant.

[jakirkham]

Morale of the story is this can be closed as they are different packages and so should not interfere, right?

[msarahan]

Yes, I think that they are sufficiently different to not collide. Import names are different and all that. I'm OK closing this if you all are. Otherwise, you probably need to discuss this with @kalefranz

[pelson]

A surprising decision - I have previously looked at the ruamel.yaml codebase, and I wouldn't have been comfortable taking on that maintenance burden (it gets deep quick)...

In anycase, thanks for the info.

[kalefranz]

History at
https://bitbucket.org/ruamel/yaml/issues/28/consider-making-ruamelyaml-available-as-a
and
https://bitbucket.org/ruamel/yaml/pull-requests/7/ruamel_yaml

Recipe that's in the Anaconda Distribution at
https://github.com/ContinuumIO/anaconda-recipes/tree/master/ruamel_yaml

I thought we had gone back to the original repo as the source, but I guess that's not the case in the current recipe. I think we can repoint without issue though.

The major differences are

• https://github.com/ContinuumIO/anaconda-recipes/blob/master/ruamel_yaml/prepare.bash
• https://github.com/ContinuumIO/anaconda-recipes/blob/master/ruamel_yaml/__init__.py
• https://github.com/ContinuumIO/anaconda-recipes/blob/master/ruamel_yaml/setup.py

[scw]

Would bubbling up the two packages in the README of the respective feedstocks be useful? I was working on rebuilding ruamel_yaml, but accidentally instead got into ruamel.yaml, and it wasn't immediately obvious to me why it lacked the internals I was expecting.

[jakirkham]

Sorry what do you mean by bubbling up?

[scw]

@jakirkham Sorry, that wasn't clear -- just some text clarifying the two packages, and their different provenance. One practical difference (besides the namespace) is that the ruamel_yaml package pulls in the related yaml.dll and builds a PYD to referencing it, the ruamel.yaml build is pure Python.

[eferreira]

Seems like this issue was deleted so the link is dead now: https://bitbucket.org/ruamel/yaml/issues/28/consider-making-ruamelyaml-available-as-a

@kalefranz or anyone else, do you remember the general gist of what the discussion over there was about? I'm struggling with the confusion between ruamel.yaml and ruamel_yaml now (earlier I was installing it via pip, but I'm trying to install it via conda instead). This is on python 2.7 btw.

I can see that when ruamel.yaml installs via pip, it puts itself into site-packages/ruamel/yaml/ and puts a .pth file into site-packages/. What was the fundamental reason why this same file structure didn't work as a conda package?

Are people just supposed to change their downstream code to import ruamel_yaml instead of ruamel.yaml? I'd like my downstream code to be agnostic to conda installation vs pip installation so I guess I'd have to do some conditional imports? Or am I misunderstanding something here?