19 KiB
PyCalVer: Automatic CalVer Versioning for Python Packages
PyCalVer is a simple versioning system, which is compatible with python packaging software setuptools PEP440.
| Name | role | since | until |
|---|---|---|---|
| Manuel Barkhau (@mbarkhau) | maintainer | 2018-09 | - |
Introduction
The PyCalVer package provides the pycalver command and module
to generate version strings. These use the following format:
v{calendar_version}.{build_number}[-{release_tag}]
Some examples:
v201711.0001-alpha
v201712.0027-beta
v201801.0031
v201801.0032-post
...
v202207.18133
v202207.18134
The pycalver bump command parses your files for such strings
and rewrites them with an incremented version string.
Format
The format for PyCalVer version strings can be parsed with this regular expression:
import re
# https://regex101.com/r/fnj60p/10
pycalver_re = re.compile(r"""
\b
(?P<version>
(?P<calver>
v # "v" version prefix
(?P<year>\d{4})
(?P<month>\d{2})
)
(?P<build>
\. # "." build nr prefix
\d{4,}
)
(?P<release>
\- # "-" release prefix
(?:alpha|beta|dev|rc|post)
)?
)(?:\s|$)
""", flags=re.VERBOSE)
version_str = "v201712.0001-alpha"
version_info = pycalver_re.match(version_str).groupdict()
assert version_info == {
"version" : "v201712.0001-alpha",
"calver" : "v201712",
"year" : "2017",
"month" : "12",
"build" : ".0001",
"release" : "-alpha",
}
version_str = "v201712.0033"
version_info = pycalver_re.match(version_str).groupdict()
assert version_info == {
"version" : "v201712.0033",
"calver" : "v201712",
"year" : "2017",
"month" : "12",
"build" : ".0033",
"release" : None,
}
Versioning Behaviour
To illustrate how PyCalVer increments version strings, we can use
pycalver incr:
$ pip install pycalver
...
$ pycalver incr v201801.0033-beta
PyCalVer Version: v201809.0034-beta
PEP440 Version: 201809.34b0
This is the simple case:
- The calendar component is update to the current year and month.
- The build number is incremented by 1.
- The optional release tag is preserved as is.
Here is how to explicitly update the release tag:
$ pycalver incr v201801.0033-beta --release=alpha
PyCalVer Version: v201809.0034-alpha
PEP440 Version: 201809.34a0
$ pycalver incr v201801.0033-beta --release=final
PyCalVer Version: v201809.0034
PEP440 Version: 201809.34
The version number is padded with extra zeros, to maintain the lexical ordering of version numbers. What happens when the padding is exhausted?
$ pycalver incr v201809.0999
PyCalVer Version: v201809.11000
PEP440 Version: 201809.11000
This is because the build number is generated as a sequence of lexical ids.
Lexical Ids
The padded build number will occasionally have to be expanded. In order to preserve both lexical ordering as well numerical ordering, build numbers are incremented in a peculiar way. Examples will perhaps illustrate more clearly.
"0001"
"0002"
"0003"
...
"0999"
"11000"
"11001"
...
"19998"
"19999"
"220000"
"220001"
What is happening here is that the left-most digit is incremented early/preemptively. Whenever the left-most digit would change, the width of the id is expanded using this simple formula:
prev_id = "0999"
next_id = str(int(prev_id, 10) + 1) # "1000"
if prev_id[0] != next_id[0]: # "0" != "1"
next_id = str(int(next_id, 10) * 11) # 1000 * 11 = 11000
This behaviour ensures that the following semantic is always
preserved: old_version < new_version. This will even be the
case if the version number was incremented twice in the same
month.
Usage
Configuration
The fastest way to setup a project is to invoke
pycalver init.
$ cd my-project
~/my-project$ pycalver init
Updated setup.cfg
This will add the following to your setup.cfg:
[pycalver]
current_version = v201809.0001-dev
commit = True
tag = True
[pycalver:file:setup.cfg]
patterns =
current_version = {version}
[pycalver:file:setup.py]
patterns =
"{version}",
"{pep440_version}",
[pycalver:file:README.md]
patterns =
{version}
{pep440_version}
This may or may not cover all version numbers across your repository. Something like the following may illustrate additional changes you might need to make.
[pycalver]
current_version = v201809.0001-beta
commit = True
tag = True
[pycalver:file:setup.cfg]
patterns =
current_version = {version}
[pycalver:file:setup.py]
patterns =
version="{pep440_version}"
[pycalver:file:src/myproject.py]
patterns =
__version__ = "{version}"
[pycalver:file:README.md]
patterns =
[PyCalVer {calver}{build}-{release}]
img.shields.io/badge/PyCalVer-{calver}{build}--{release}-blue
You can ommit patterns if the default patterns are sufficient.
These are:
patterns =
{version}
{pep440_version}
This allows for a shorter (albeit less explicit) configuration:
[pycalver]
current_version = v201809.0001-beta
commit = True
tag = True
[pycalver:file:setup.cfg]
[pycalver:file:setup.py]
[pycalver:file:src/myproject.py]
[pycalver:file:README.md]
patterns =
[PyCalVer {calver}{build}-{release}]
img.shields.io/badge/PyCalVer-{calver}{build}--{release}-blue
Pattern Search and Replacement
patterns is used both to search for version strings and to
generate the replacement strings. The following placeholders are
available for use, everything else in a pattern is treated as
literal text.
| placeholder | example |
|---|---|
pep440_version |
201809.1b0 |
version |
v201809.0001-alpha |
calver |
v201809 |
year |
2018 |
month |
09 |
build |
.0001 |
release |
-alpha |
Note that the separator/prefix characters are part of what is matched and generated for a given placeholder, and they should not be included in your patterns.
A further restriction is, that a version string cannot span multiple lines in your source file.
Now we can call pycalver bump to bump all occurrences of
version strings in these files. Normally this will change local
files, but the --dry flag will instead display a diff of the
changes that would be applied.
$ pycalver show
Current Version: v201809.0001-beta
PEP440 Version: 201809.1b0
$ pycalver bump --dry
TODO: diff output
Dont forget to do $ git push --tags
TODO: commits and tags
Rational
Other Versioning Software
This project is very similar to bumpversion, upon which it is partially based. So why another library? PyCalVer version strings can be generated automatically, usage is a bit more simple.
Some Details
- Version numbers are for public releases. For the purposes of development of the project itself, reference VCS branches and commit ids are more appropriate.
- There should be only one person or system responsible for updating the version number at the time of release, otherwise the same version number may be generated for different builds.
- Lexeographical order is
Canonical PyCalVer version strings can be parsed with this regular expression:
These are the full version strings, for public announcements and
conversations it will often be sufficient to refer simply to
v201801, by which the most recent post release build of
that month is meant.
version_str = "v201712.0027-beta"
version_dict = pycalver_re.match("v201712.0027-beta").groupdict()
import pkg_resources # from setuptools
version = pkg_resources.parse_version(version_str)
--
In [2]: version_dict
{'year': '2017', 'month': '12', 'build_nr': '0027', 'tag': 'beta'}
>>> str(version)
201712.27b0
Realities of Verion Numbers
Nobody knows what the semantics of a version number are, because nobody can guarantee that a given release adheres to whatever convention one would like to imbibe it with. Lets just keep things simple.
- Version numbers should be recognizable as such, that's what the "v" prefix does.
- A number like 201808 is recognizable to many as a number derived from a calendar.
- alpha, beta are common parlance indicating software which is still under development.
Some additional constraints are applied to conform with PEP440
Should I use PyCalVer for my Project?
If your project is 1. not useful by itself, but only when used by other software, 2. has a finite scope/a definition of "done", 3. your project has CI, a test suite with and decent code coverage, then PyCalVer is worth considering. You release at most once per month.
Marketing/Vanity
Quotes from http://sedimental.org/designing_a_version.html
Rational
PyCalVer is opinionated software. This keeps things simple, when the opintions match yours, but makes it useless for everybody else.
The less machine parsable semantics you put in your version string, the better. The ideal would be to only have a single semantic: newer == better.
Some projects depend recursively on hundreds of libraries, so compatability issues generated by your project can be a heavy burdon on thousands of users; users who learn of the existance of your library for the first time in the form of a stacktrace. PyCalVer is for projects that are comitted to and can maintain backward compatability. Newer versions are always better, updates are always safe, an update won't break things, and if it does, the maintainer's hair is on fire and they will publish a new release containing a fix ASAP.
Ideally, your user can just declare your library as a dependency, without any extra version qualifier, and never have to think about it again. If you do break something by accident, their remedy is not to change their code, but to temporarily pin an earlier version, until your bugfix release is ready.
PyCalVer is for projects which are the mundane but dependable foundations of other big shiny projects, which get to do their big and exciting 2.0 major releases.
Breaking Things is a Big Deal
Using an increment in a version string to express that a release may break client code is not tennable. A developer cannot be expected to think about how their code may or may not break as a consequence of your decision to rename some functions. As the author of any software, there is a great temptation to move fast and break things. This is great when no other software depends on yours. If something breaks, you jump up and fix it. The story is quite different even when only a few dozen people depend on your software.
The less the users of your library have to know about your project, the better. The less they have to deal with issues of compatability, the better. SemVer can be overly specifc for some kinds of projects. If you are writing a library and you have a commitment to backward compatability
PyCalVer version strings can be parsed according to PEP440 https://www.python.org/dev/peps/pep-0440/
A Word on Marketing
This setup of expectations for users can go one of two ways,
We use version numbers to communicate between the authors of software and its users. For users of libraries Particularly for libraries, it pays to keep things as simple as possible for your human users.
Commitment to Compatability
Software projects can depend on many libraries. Consider that one package introducing a breaking change is enough to mess up your day. Especially in the case of libraries, your users should be able to write code that uses it and not have that code break at any point in the future. Users cannot be asked to keep track of all the changes to every little library that they use.
PyCalVer is explicitly non semantic. A PyCalVer version number does not express anything about
- Don't ever break things. When users depend on your
software, backward compatability matters and the way to
express backward incompatible changes is not to bump a
version number, but to change the package name. A change
in the package name clearly communicates that a user must
change their code so that it will work with the changed
API. Everybody who does not have the bandwith for those
changes, doesn't even have to be aware of your new
release.
- When you do break something, that should be considered a
bug that has to be fixed as quickly as possible in a new
version. It should always be safe for a user to update
their dependencies. If something does break, users have to
temporarilly pin an older (known good) version, or update
to a newer fixed version.
- Version numbers should not require a parser (present
package excluded of course). A newer version number should
always be lexeographically greater than an older one.
TODO:
https://setuptools.readthedocs.io/en/latest/setuptools.html#specifying-your-project-s-version
The main component of the version number is based on the calendar date. This is allows you to show your commitment (or lack thereof) to the maintenance of your libarary. It also allows users to see at a glance that their dependency might be out of date. In this versioning scheme it is completely reasonable to bump the version number without any changes, simply to express to your users, that you are still actively maintaining the software and that it is in a known good state.
For a much more detailed exposition of CalVer, see http://sedimental.org/designing_a_version.html https://calver.org/
from pkg_resources import parse_version
The Life of a Library
mylib v201711.001-alpha # birth of a project (in alpha)
mylib v201711.002-alpha # new features (in alpha)
mylib v201712.003-beta # bugfix release (in beta)
mylib v201712.004-rc # release candidate
mylib v201712.005 # stable release
mylib v201712.006 # stable bugfix release
mylib2 v201712.007-beta # breaking change (new package name!)
mylib2 v201801.008-beta # new features (in beta)
mylib2 v201801.009 # stable release
mylib v201802.007 # security fix for legacy version
mylib2 v201802.010 # security fix
mylib2 v202604.9900 # threshold for four digit build numbers
mylib2 v202604.9901 # still four digits in the same month
mylib2 v202604.9911 # last build number with four digits
mylib2 v202605.09912 # build number zero padding added with date turnover
mylib2 v202605.09913 # stable release
mylib2 v203202.16051-rc # release candidate
mylib2 v203202.16052 # stable release
...
v202008.500 # 500 is the limit for four digit build numbers, but
v202008.508 # zero padding is added only after the turnover to
v202009.0509 # a new month, so that lexical ordering is preserved.
The date portion of the version, gives the user an indication of how up their dependency is, whether or not a project is still being maintained.
The build number, gives the user an idea of the maturity of the project. A project which has been around long enough to produce hundreds of builds, might be considered mature, or at least a project that is only on build number 10, is probably still in early development.
FAQ
Q: "So you're trying to tell me I need to create a whole new package every time I introduce a introduce a breaking change?!".
A: First of all, what the hell are you doing? Secondly, YES! Let's assume your little package has even just 100 users. Do you have any idea about the total effort that will be expended because you decided it would be nice to change the name of a function? It is completely reasonable introduce that the friction for the package author when the price to users is orders of magnitude larger.
1801
I have given up on the idea that version numbers express anything about changes made between versions. Trying to express such information assumes 1. that the author of a package is aware of how a given change needs to be reflected in a version number and 2. that users and packaging softare correctly parse that meaning. When I used semantic versioning, I realized that the major version number of my packages would never change, because I don't think breaking changes should ever be