Sort output metadata prior to writing to meta.yaml

[mandeep]

In relation to #2140:

The field values in the metadata are sorted alphabetically prior to outputting the meta.yaml file in the info/recipe directory. The build/script and test/commands values are omitted.

[msarahan]

Fingers crossed that those two fields are the only ones we need to omit. Merging for now, will revisit if issues crop up.

[mingwandroid]

@mandeep, @msarahan, sometimes the values here are lists, e.g. multiple sources and this code falls over:

meta.yaml:

package:
  name: pixman-cos6-x86_64
  version: 0.32.8

source:
  - url: http://mirror.centos.org/centos/6.9/os/x86_64/Packages/pixman-0.32.8-1.el6.x86_64.rpm
    sha256: d5e9a2ac49bae1bff6fd806367aa0f9125eef6dee1b515a6af6576b883032e91
    folder: binary
  - url: http://vault.centos.org/6.9/os/Source/SPackages/pixman-0.32.8-1.el6.src.rpm
    folder: source

outputs:
  - name: pixman-cos6-x86_64
    noarch: generic

    about:
      home: http://cgit.freedesktop.org/pixman/
      license: MIT
      license_family: MIT
      summary: "(CDT) Pixel manipulation library"
      description: |
        Pixman is a pixel manipulation library for X and cairo.

Traceback (most recent call last):
  File "/root/mcf-x-cos6_64/mc3/bin/conda-build", line 11, in <module>
    load_entry_point('conda-build==3.0.0a3+346.gdf29838d', 'console_scripts', 'conda-build')()
  File "/root/mcf-x-cos6_64/mc3/lib/python3.6/site-packages/conda_build-3.0.0a3+346.gdf29838d-py3.6.egg/conda_build/cli/main_build.py", line 340, in main
    execute(sys.argv[1:])
  File "/root/mcf-x-cos6_64/mc3/lib/python3.6/site-packages/conda_build-3.0.0a3+346.gdf29838d-py3.6.egg/conda_build/cli/main_build.py", line 331, in execute
    noverify=args.no_verify)
  File "/root/mcf-x-cos6_64/mc3/lib/python3.6/site-packages/conda_build-3.0.0a3+346.gdf29838d-py3.6.egg/conda_build/api.py", line 183, in build
    need_source_download=need_source_download, config=config, variants=variants)
  File "/root/mcf-x-cos6_64/mc3/lib/python3.6/site-packages/conda_build-3.0.0a3+346.gdf29838d-py3.6.egg/conda_build/build.py", line 1623, in build_tree
    built_packages=built_packages,
  File "/root/mcf-x-cos6_64/mc3/lib/python3.6/site-packages/conda_build-3.0.0a3+346.gdf29838d-py3.6.egg/conda_build/build.py", line 1192, in build
    built_package = bundlers[output_d.get('type', 'conda')](output_d, m, env)
  File "/root/mcf-x-cos6_64/mc3/lib/python3.6/site-packages/conda_build-3.0.0a3+346.gdf29838d-py3.6.egg/conda_build/build.py", line 725, in bundle_conda
    output['checksums'] = create_info_files(metadata, files, prefix=metadata.config.host_prefix)
  File "/root/mcf-x-cos6_64/mc3/lib/python3.6/site-packages/conda_build-3.0.0a3+346.gdf29838d-py3.6.egg/conda_build/build.py", line 499, in create_info_files
    copy_recipe(m)
  File "/root/mcf-x-cos6_64/mc3/lib/python3.6/site-packages/conda_build-3.0.0a3+346.gdf29838d-py3.6.egg/conda_build/build.py", line 217, in copy_recipe
    for key in value.keys():
AttributeError: 'list' object has no attribute 'keys'

[mandeep]

Would it be okay to move the AttributeError exception catch earlier? Maybe start the try at the beginning of the block.

[mbargull]

Maybe this behavior should be documented somewhere. That might be helpful in case of errors for some exotic/peculiar use-cases. Say someone has a weird package which uses

test:
  imports:
    - weird_module_that_installs_some_other_module_on_first_import
    - some_other_module_installed_by_the_weird_one

which wouldn't work when sorted. Of course that example is very constructed/artificial (and nothing anyone should aim for). The point is, if not documented, the sorting might be unexpected behavior from the user's stand point.
Furthermore, concerning @msarahan's comment

Fingers crossed that those two fields are the only ones we need to omit. Merging for now, will revisit if issues crop up.

: if ('build/script', 'test/commands') isn't an exhaustive blacklist and this feature causes trouble in the future, maybe a whitelist-approach will be a safer bet. BTW, there are even unordered sets (null-mappings) in standard YAML (http://yaml.org/spec/1.2/spec.html#id2761758) -- but they don't seem to used that commonly and by now it wouldn't make much sense to introduce them to conda-build, I guess.

[github-actions]

Hi there, thank you for your contribution!

This pull request has been automatically locked because it has not had recent activity after being closed.

Please open a new issue or pull request if needed.

Thanks!