Overview --------
conda-smithy is a tool for combining a conda recipe with configurations to build using freely hosted CI services into a single repository, also known as a feedstock.
conda-smithy is still a work-in-progress, but when complete, conda-smithy will:
+ Create a git repo with a conda recipe and the files to run conda builds via CI services. + Register the repo on github and push it. + Connect the repo to the CI services dev.azure.com
Installation ------------
The easiest way to install conda-smithy is to use conda and conda-forge:
conda install -n root -c conda-forge conda-smithyTo install conda-smithy from source, see the requirements file in requirements.txt, clone this
repo, and python -m pip install ..
Setup -----
You need a token from github to try out
conda-smithy. The commands which need this will tell you where to get these tokens and where to
place them. If you need help getting tokens please ask on the
conda-forge google group.
You should be able to test parts of conda-smithy with whatever tokens you have.
For example, you should be able to conda smithy register-github without the CI service tokens.
Re-rendering an existing feedstock is also possible without CI service tokens set.
Re-rendering an existing feedstock ----------------------------------
Periodically feedstocks need to be upgraded to include new features. To do
this we use conda-smithy to go through a process called re-rendering.
Make sure you have installed conda-smithy before proceeding.
Re-rendering an existing feedstock is possible without CI service tokens set.
cdconda smithy rerender [--commit]- Commit and push all changes
conda-smithy version 1.4.1+.
To do this just use the --commit/-c option. By default this will open an editor to make a commit.
It will provide a default commit message and show the changes to be added. If you wish to do this
automatically, please just use --commit auto/-c auto and it will use the stock commit message.
Making a new feedstock ----------------------
- Make the feedstock repo:
conda smithy init
foo, this creates a
directory called foo-feedstock, populates it with CI setup skeletons, adds the recipe under
recipe and initializes it as a git repo.
- Create a github repo:
conda smithy register-github --organization conda-forge ./foo-feedstock.
--user github_user_name. If you are interested in adding teams for your feedstocks,
you can provide the --add-teams option to create them. This can be done when creating
the feedstock or after.
- Register the feedstock with CI services:
conda smithy register-ci --organization conda-forge --feedstock_directory ./foo-feedstock.
This requires tokens for the CI services. You can give the name of a user instead
of organization with --user github_user_name. By default this command requires an Anaconda/Binstar token
to be available in ~/.conda-smithy/anaconda.token, or as BINSTAR_TOKEN in the environment. This can be opted
out of by specifying --without-anaconda-token, as such execpted package uploads will not be attempted.
* For Azure, you will have to create a service connection with the same name as your github user or org
https://dev.azure.com/YOUR_ORG/feedstock-builds/_settings/adminservices
* For Azure builds, you will have to export the environment variable AZURE_ORG_OR_USER to point to your Azure org
* If this is your first build on Azure, make sure to add Library Variable Group containing your BINSTAR_TOKEN for automated anaconda uploads.
- Specify the feedstock channel and label:
recipe/conda_build_config.yaml.
channel_sources:
- mysourcechannel1,mysourcechannel2,conda-forge,defaults
channel_targets:
- target_channel target_labelconda-forge,defaults. Default for channel targets is conda-forge main.
- Specify your branding in the README.md:
conda-forge.yml file:
github:
user_or_org: YOUR_GITHUB_USER_OR_ORG- Re-render the feedstock: `
conda smithy rerender --feedstock_directory ./foo-feedstock
- Commit the changes:
cd foo-feedstock && git commit, then pushgit push upstream master.
When everything is configured you can trigger a build with a push to the feedstock repo on github.
Developing conda-smithy -----------------------
To develop conda smithy, use your favortite conda-based environment manager and create an environment based on the environment.yml.
$ conda env createReleasing conda-smithy ----------------------
Before making a release, consult @conda-forge/core and wait some time for objections.
To release a new version of conda-smithy, you can use the
rever release managment tool.
Run rever in the root repo directory with the version number you want to release.
For example,
$ rever 0.1.2Deprecation policy ------------------
conda-smithy's API is defined as the collection of all reachable symbols whose fully qualified import
path does not feature a leading underscore in any of its components. The API covers renames
and, if callable, changes in signatures (argument and keyword argument names and types,
plus the return types). The API also covers the command-line interface. Any other symbol may change
at any time and has no guaranteed API.
All API changes must undergo a 60-day deprecation period, must be clearly indicated via
a DeprecationWarning, and must have a news entry in the Deprecated` section.
Conda-smithy in a nutshell --------------------------
