vv's (b)log

Create the conda packages of the repos you want to have available

Two ways:

A) Use conda skeleton for things that are already on PyPI

B) Use conda build from a custom made meta.yml, eg if you want to build for a custom repo that is not on PyPI, or for a version of that repo/package not yet on PyPI

Let's start with A)

Conda packages

conda build .

Had to make a package for pyaaf2 too since it wasn't on conda


package:
  name: "{{ name|lower }}"
  version: "{{ version }}"

source:
  url: "https://pypi.io/packages/source/{{ name[0] }}/{{ name }}/{{ name }}-{{ version }}.tar.gz"
  sha256: 160d3c26c7cfef7176d0bdb0e55772156570435982c3abfa415e89639f76e71b

build:
  noarch: python
  number: 0
  script: "{{ PYTHON }} -m pip install . -vv"

requirements:
  host:
    - pip
    - python >=3
  run:
    - python >=3

test:
  imports:
    - aaf2
    - aaf2.model
    - aaf2.model.ext

about:
  home: "https://github.com/markreidvfx/pyaaf2"
  license: MIT
  license_family: MIT
  license_file: LICENSE
  summary: "A python module for reading and writing advanced authoring format files"
  doc_url: "https://pyaaf.readthedocs.io/en/latest"

extra:
  recipe-maintainers:
    - vvzen

CONFIG=osx64 ./.scripts/run_docker_build.sh

#################################################################################### Resource usage summary:

Total time: 0:21:02.0 CPU usage: sys=0:00:01.5, user=0:00:03.2 Maximum memory usage observed: 60.0M Total disk usage observed (not including envs): 2.2K

touch /home/conda/staged-recipes/build_artifacts/conda-forge-build-done

then I forked the conda-recipes repo, made a branch, added the recipe, made a PR, waited for the CI to kick in, update the .yaml based on the initial feedback from the conda-linter bot, update the PR based on the human review, and then have it merged! All of that in less than 1 hour!!

noarch: if it's a pure python package.

Then, a new github repo was created:

conda-forge/pyaaf2-feedstock

And you can see that the package is now available on conda-forge: https://anaconda.org/conda-forge/pyaaf2

Now I can go on with OpenTimelineIO. In an ideal world, I could have just used conda skeleton, like this:

conda skeleton pypi opentimelineio

but since the latest PyPI version of OTIO has a tiny typo in their setup.py, I had to go the manual route once again..

conda build . --channel conda-forge

Notice that I added the conda-forge channel since that's where the pyaaf2 package lives now.

#################################################################################### Resource usage summary:

Total time: 0:04:00.6 CPU usage: sys=0:00:04.9, user=0:01:38.5 Maximum memory usage observed: 1.9G Total disk usage observed (not including envs): 256.5K


package:
  name: "{{ name|lower }}"
  version: "{{ version }}"

source:
  - url: https://github.com/PixarAnimationStudios/OpenTimelineIO/archive/refs/tags/v0.13.tar.gz
    sha256: 33a63891b4656804242512e122b33ed12e35d4038fd78610ccb82b441b9506dd
  - url: https://github.com/pybind/pybind11/archive/refs/tags/v2.6.2.tar.gz
    sha256: 8ff2fff22df038f5cd02cea8af56622bc67f5b64534f1b83b9f133b8366acff2

build:
  number: 0
  script: "{{ PYTHON }} -m pip install --no-deps --ignore-installed . -vv"

requirements:
  build:
    - "{{ compiler('cxx') }}"
    - cmake
  host:
    - pip
    - pyaaf2
    - setuptools
    - python
  run:
    - pip
    - pyaaf2
    - python

test:
  commands:
    - make test

about:
  home: https://github.com/PixarAnimationStudios/OpenTimelineIO
  license: Apache-2.0
  license_family: Apache
  license_file: LICENSE.txt
  summary: Python API for interchange of editorial cut information
  doc_url: "https://opentimelineio.readthedocs.io"

extra:
  recipe-maintainers:
    - yournamehere

Problem was that OpenTimelineIO is vendoring some dependencies, so I had to at least add pybind11.

TravisCI (OSX, IMB Power 8+) CircleCI (Linux, OSX) Azure Pipelines OSX, Linux (x86_64, native), Linux (ARMv8, emulated) and Linux (IBM Power8+, emulated)

pip install --no-deps Don't install package dependencies.

python setup.py install --single-version-externally-managed used by system package builders to create 'flat' eggs

use bash! fish is fun I was getting additional issues..

If you want to test packaging locally, you'll need docker otherwise you can try packaging just for your system

https://packaging.python.org/overview/

Contributing to conda-forge

Conda packages