Entrypoints: extending/changing zest.releaser

A zest.releaser entrypoint gets passed a data dictionary and that’s about it. You can do tasks like generating documentation. Or downloading external files you don’t want to store in your repository but that you do want to have included in your egg.

Every release step (prerelease, release and postrelease) has three points where you can hook in an entry point:

before
Only the workingdir and name are available in the data dictionary, nothing has happened yet.
middle
All data dictionary items are available and some questions (like new version number) have been asked. No filesystem changes have been made yet.
after
The action has happened, everything has been written to disk or uploaded to pypi or whatever.

For the release step it made sense to create extra entry points:

after_checkout
The middle entry point has been handled, the tag has been made, a checkout of that tag has been made and we are now in that checkout directory. Of course, when the user chooses not to do a checkout, this entry point never triggers.
before_upload
The source distribution and maybe the wheel have been made. We are about to upload to PyPI with python setup.py or twine upload dist/*. You may want to use this hook to to sign a distribution before twine uploads it.

Note that an entry point can be specific for one package (usually the package that you are now releasing) or generic for all packages. An example of a generic one is gocept.zestreleaser.customupload, which offers to upload the generated distribution to a chosen destination (like a server for internal company use). If your entry point is specific for the current package only, you should add an extra check to make sure it is not run while releasing other packages; something like this should do the trick:

def my_entry_point(data):
    if data['name'] != 'my.package':
        return
    ...

Entry point specification

An entry point is configured like this in your setup.py:

entry_points={
    #'console_scripts': [
    #    'myscript = my.package.scripts:main'],
    'zest.releaser.prereleaser.middle': [
        'dosomething = my.package.some:some_entrypoint',
        ]},

Entry-points can also be specified in the setup.cfg file like this (The options will be split by white-space and processed in the given order.):

[zest.releaser]
prereleaser.middle =
   my.package.some.some_entrypoint
   our.package.other_module.other_function

In zest.releaser.prereleaser.middle resp. prereleaser.middle replace prereleaser with the respective command name (prerelease, release, postrelease, etc.) and middle with the respective hook name (before, middle, after, after_checkout, etc.) as needed.

See the setup.py of zest.releaser itself for some real world examples.

You’ll have to make sure that the zest.releaser scripts know about your entry points, for instance by placing your egg (with entry point) in the same zc.recipe.egg section in your buildout as where you placed zest.releaser. Or, if you installed zest.releaser globally, your egg-with-entrypoint has to be globally installed, too.

Notes:

  • Entry-points given in setup.cfg will be processed before entry-point defined via installed packages.
  • The order in which entry-point defined via installed packages are processed is undefined.

Comments about data dict items

Your entry point gets a data dictionary: the items you get in that dictionary are documented below. Some comments about them:

  • Not all items are available. If no history/changelog file is found, there won’t be any data['history_lines'] either.
  • Items that are templates are normal python string templates. They use dictionary replacement: they’re actually passed the same data dict. For instance, prerelease’s data['commit_message'] is by default Preparing release %(new_version)s. A “middle” entry point could modify this template to get a different commit message.
  • Entry-points can change the the data dictionary, thus one hook can prepare data another one is using. But be aware that changing entries in the dict may lead to malfunction.

Common data dict items

These items are shared among all commands.

commit_msg
Message template used when committing
headings
Extracted headings from the history file
history_encoding
The detected encoding of the history file
history_file
Filename of history/changelog file (when found)
history_header
Header template used for 1st history header
history_insert_line_here
Line number where an extra changelog entry can be inserted.
history_last_release
Full text of all history entries of the current release
history_lines
List with all history file lines (when found)
name
Name of the project being released
new_version
New version to write, possibly with development marker
nothing_changed_yet
First line in new changelog section, warn when this is still in there before releasing
original_version
Original package version before any changes
reporoot
Root of the version control repository
required_changelog_text
Text that must be present in the changelog. Can be a string or a list, for example [“New:”, “Fixes:”]. For a list, only one of them needs to be present.
workingdir
Original working directory

prerelease data dict items

today
Date string used in history header

release data dict items

tag
Tag we’re releasing
tag-message
Commit message for the tag
tag-signing
Sign tag using gpg or pgp
tag_already_exists
Internal detail, don’t touch this :-)
tagdir
Directory where the tag checkout is placed (if a tag checkout has been made)
tagworkingdir
Working directory inside the tag checkout. This is the same, except when you make a release from within a sub directory. We then make sure you end up in the same relative directory after a checkout is done.
version
Version we’re releasing

postrelease data dict items

dev_version
New version with development marker (so 1.1.dev0)
dev_version_template
Template for development version number
development_marker
String to be appended to version after postrelease
new_version
New version, without development marker (so 1.1)

addchangelogentry data dict items

commit_msg
Message template used when committing. Default: same as the message passed on the command line.
message
The message we want to add

bumpversion data dict items

breaking
True if we handle a breaking (major) change
clean_new_version
Clean new version (say 1.1)
feature
True if we handle a feature (minor) change
release
Type of release: breaking, feature, normal