This repo is a holding area for recipes destined for a conda-forge feedstock repo. To find out more about conda-forge, see https://github.com/conda-forge/conda-smithy.
- Fork this repository.
- Make a new folder in
recipes
for your package. Look at the example recipe, our documentation and the FAQ for help. - Open a pull request. Building of your package will be tested on Windows, Mac and Linux.
- When your pull request is merged a new repository, called a feedstock, will be created in the github conda-forge organization, and build/upload of your package will automatically be triggered. Once complete, the package is available on conda-forge.
For Python packages available on pypi
it is possible to use grayskull to generate the recipe. The user should review the recipe generated, specially the license and dependencies.
Installing grayskull
: conda install -c conda-forge grayskull
Generating recipe: grayskull pypi PACKAGE_NAME_HERE
Look at one of these examples in this repository and modify it as necessary.
Your final recipe should have no comments and follow the order in the example.
If there are details you are not sure about please open a pull request. The conda-forge team will be happy to answer your questions.
If your package is on PyPI, you can get the sha256 hash from your package's page on PyPI; look for the SHA256
link next to the download link for your package.
You can also generate a hash from the command line on Linux (and Mac if you install the necessary tools below). If you go this route, the sha256
hash is preferable to the md5
hash.
To generate the md5
hash: md5 your_sdist.tar.gz
To generate the sha256
hash: openssl sha256 your_sdist.tar.gz
You may need the openssl package, available on conda-forge:
conda install openssl -c conda-forge
Use the skip
key in the build
section along with a selector:
build:
skip: true # [win]
A full description of selectors is in the conda docs.
Additionally, when pushing commits for a recipe that excludes Windows, put [skip appveyor]
in the commit message to prevent CI tests
on Windows from even starting.
The build number is used when the source code for the package has not changed but you need to make a new build. For example, if one of the dependencies of the package was not properly specified the first time you build a package, then when you fix the dependency and rebuild the package you should increase the build number.
When the package version changes you should reset the build number to 0
.
No, you do not.
Short answer: yes. Long answer: In principle, as long as your dependencies are in at least one of your user's conda channels they will be able to install your package. In practice, that is difficult to manage, and we strive to get all dependencies built in conda-forge.
This should be the default install line for most Python packages. This is preferable to python setup.py
because it handles metadata in a conda
-friendlier way.
In many cases, no. Python packages almost never need it. If the build can be done with one line you can put it in the script
line of the build
section.
The maintainers "job" is to:
- keep the feedstock updated by merging eventual maintenance PRs from conda-forge's bots;
- keep the package updated by bumping the version whenever there is a new release;
- answer eventual question about the package on the feedstock issue tracker.
When a PR of recipe(s) is ready to go, it is merged into master
. This will trigger a CI build specially designed to convert the recipe(s). However, for any number of reasons the recipe(s) may not be converted right away. In the interim, the recipe(s) will remain in master
until they can be converted. There is no action required on the part of recipe contributors to resolve this. Also it should have no impact on any other PRs being proposed. If these recipe(s) pending conversion do cause issues for your submission, please ping conda-forge/core
for help.
Sometimes, some of the CI tools' builds fail due to no error within your recipe. If that happens, you can trigger a rebuild by re-creating the last commit and force pushing it to your branch:
# edit your last commit, giving it a new time stamp and hash
# (you can just leave the message as it is)
git commit --amend
# push to github, overwriting your branch
git push -f
If the problem was due to scripts in the staged-recipes
repository, you may be asked to "rebase" once these are fixed. To do so, run:
# If you didn't add a remote for conda-forge/staged-recipes yet, also run
# these lines:
# git remote add upstream https://github.com/conda-forge/staged-recipes.git
# git fetch --all
git rebase upstream/master
git push -f
12. My pull request passes all checks, but hasn't received any attention. How do I call attention to my PR? What is the customary amount of time to wait?
If your PR is passing all checks, but has not been acted on by the staged recipes maintainers, you can ping @conda-forge/staged-recipes to request action. You do not need to wait any specific amount of time once the recipe is ready to go.
If your recipe still does not recieve any attention after a few days, you may attempt to re-ping @conda-forge/staged-recipes. You may also attempt to bring the PR up in our Gitter chat room at https://gitter.im/conda-forge/conda-forge.github.io
All apologies in advance if your recipe PR does not recieve prompt attention. This is a high volume repository and issues can easily be missed. We are always looking for more staged-recipe reviewers. If you are interested in volunteering, please contact a member of @conda-forge/core. We'd love to have the help!
First, don't. Second, please don't.
Add a conda_build_config.yaml
file inside the recipe folder with the contents
channel_sources:
- conda-forge/label/cf201901,defaults # [unix]
- conda-forge,defaults # [win]
channel_targets:
- conda-forge cf201901 # [unix]
- conda-forge main # [win]
c_compiler: # [unix]
- gcc # [linux]
- clang # [osx]
cxx_compiler: # [unix]
- gxx # [linux]
- clangxx # [osx]
fortran_compiler: # [unix]
- gfortran # [unix]