Building changelog information in a form that is human digestible but still containing as much useful information is difficult. Many attempts over the years have resulted in a variety of methods that projects use to solve this problem:
All of these methods have their pros and cons. Some have issues with accuracy. Some have issues with lack of details. None of these methods seem to cover all of the needs of many projects and are full of potential pitfalls.
In order to solve these problems, releasedocmaker was written to automatically generate a changelog and release notes by querying Apache’s JIRA instance.
dateutil may be installed via pip:
pip install python-dateutil
Minimally, the name of the JIRA project and a version registered in JIRA must be provided:
$ releasedocmaker --project (project) --version (version)
This will query Apache JIRA, generating two files in a directory named after the given version in an extended markdown format which can be processed by both mvn site and GitHub.
This is similar to the JIRA
Release Notes button but is in tabular format and includes the priority, component, reporter, and contributor fields. It also highlights Incompatible Changes so that readers know what to look out for when upgrading. The top of the file also includes the date that the version was marked as released in JIRA.
If your JIRA project supports the release note field, this will contain any JIRA mentioned in the CHANGELOG that is either an incompatible change or has a release note associated with it. If your JIRA project does not support the release notes field, this will be the description field.
For example, to build the release documentation for HBase v1.2.0:
$ releasedocmaker --project HBASE --version 1.2.0
By default, release notes are expected to be in plain text. However, you can write them in markdown if you include a header at the top of your release note:
<!-- markdown --> remaining text
By default, it will use a header that matches the project name. But that is kind of ugly and the case may be wrong. Luckily, the title can be changed:
$ releasedocmaker --project HBASE --version 1.2.0 --projecttitle "Apache HBase"
Now instead of
HBASE, it will use
Apache HBase for some titles and headers.
It is sometimes useful to create the CHANGELOG and RELEASENOTES with versions attached.
releasedocmaker supports both independently.
$ releasedocmaker --project HBASE --version 1.2.0 --fileversions
This command line will now create CHANGELOG.1.2.0.md and RELEASENOTES.1.2.0.md files.
$ releasedocmaker --project HBASE --version 1.2.0 --dirversions
This command line will now create a directory called 1.2.0 and inside will be the CHANGELOG.md and RELEASENOTES.md files.
Using both at the same time…
$ releasedocmaker --project HBASE --version 1.2.0 --fileversions --dirversions
… results in 1.2.0/CHANGELOG.1.2.0.md and 1.2.0/RELEASENOTES.1.2.0.md files.
--fileversions or both simultaneously,
releasedocmaker can also generate multiple versions at once
$ releasedocmaker --project HBASE --version 1.0.0 --version 1.2.0 --dirversions
This will create the files for versions 1.0.0 and versions 1.2.0 in their own directories.
But what if the version numbers are not known? releasedocmaker can also generate version data based upon ranges:
$ releasedocmaker --project HBASE --version 1.0.0 --version 1.2.0 --range --fileversions
In this form, releasedocmaker will query JIRA, discover all versions that alphabetically appear to be between 1.0.0 and 1.2.0, inclusive, and generate all of the relative release documents. This is especially useful when bootstrapping an existing project.
For released versions, releasedocmaker will pull the date of the release from JIRA. However, for unreleased versions it marks the release as
Unreleased. This can be inconvenient when actually building a release and wanting to include it inside the source package.
The –usetoday option can be used to signify that instead of using Unreleased, releasedocmaker should use today’s date.
$ releasedocmaker --project HBASE --version 1.0.0 --usetoday
After using this option and release, don’t forget to change JIRA’s release date to match!
Different projects may find one type of sort better than another, depending upon their needs. releasedocmaker supports two types of sorts and each provides two different options in the direction for that sort.
By default, releasedocmaker will sort the output based upon the resolution date of the issue starting with older resolutions. This is the same as giving these options:
$ releasedocmaker --project falcon --version 0.6 --sorttype resolutiondate --sortorder older
The order can be reversed so that newer issues appear on top by providing the ‘newer’ flag:
$ releasedocmaker --project falcon --version 0.6 --sorttype resolutiondate --sortorder newer
In the case of multiple projects given on the command line, the projects will be interspersed.
An alternative to the date-based sort is to sort based upon the issue id. This may be accomplished via:
$ releasedocmaker --project falcon --version 0.6 --sorttype issueid --sortorder asc
This will now sort by the issue id, listing them in lowest to highest (or ascending) order.
The order may be reversed to list them in highest to lowest (or descending) order by providing the appropriate flag:
$ releasedocmaker --project falcon --version 0.6 --sorttype issueid --sortorder desc
In the case of multiple projects given on the command line, the projects will be grouped and then sorted by issue id.
To check if an issue is backward-incompatible the releasedocmaker script first checks the
Hadoop Flags field in the
issue. If this field is found to be blank then it searches for the ‘backward-incompatible’ label. You can override the
default value for this label by using –incompatiblelabel option e.g.
$ releasedocmaker --project falcon --version 0.6 --incompatiblelabel not-compatible
or equivalently using the shorter -X option
$ releasedocmaker --project falcon --version 0.6 -X not-compatible
In order to ensure proper formatting while using mvn site, releasedocmaker puts in periods (.) for fields that are empty or unassigned. This can be unsightly and not proper for any given project. There are also other things, such as missing release notes for incompatible changes, that are less than desirable.
In order to help release managers from having to scan through potentially large documents, releasedocmaker features a lint mode, triggered via –lint:
$ releasedocmaker --project HBASE --version 1.0.0 --lint
This will do the normal JIRA querying, looking for items it considers problematic. It will print the information to the screen and then exit with either success or failure, depending upon if any issues were discovered.
There is basic support for an autoindexer. It will create two files that contain links to all directories that have a major.minor*-style version numbering system. For example directories with names like 0.6, 1.2.2, 1.2alpha etc. will all be linked.
You can find the version of the releasedocmaker that you are using by giving the -V option. This may be helpful in finding documentation for the version you are using.
$ releasedocmaker -V