Wikis are not just for encyclopedias and websites anymore. You can use Ikiwiki in combination with your revision control system to handle issue tracking, news feeds, and other needs of a software project. The wiki can make your bug reports as much a part of your software project as its code, with interesting results.
Ikiwiki is a wiki engine with a twist. It's best
described by the term "wiki compiler". Just as a
typical software project consists of source code
that is stored in revision control and compiled with
gcc, an ikiwiki-based wiki is stored as
human-editable source in a revision control system,
and built into HTML using ikiwiki.
Ikiwiki uses your revision control system to track changes and handle tasks such as rolling back changes and merging edits. Because it takes advantage of revision control, there are no annoying warnings about other people editing a file, or finding yourself locked out of a file because someone else started editing it and left. Instead, the other person's changes will be automatically merged with yours when you commit.
In the rare cases where automatic merging fails because of concurrent edits to the same part of a page, regular commit conflict markers are shown in the file to let you resolve the conflict, as you would for conflicting edits in source code.
Ikiwiki is a full-featured wiki that you can use for a variety of purposes, from traditional wikis to weblogs, podcasting, or even aggregating other sites' RSS feeds into a Planet page. While people are using Ikiwiki for purposes ranging from genealogy research to shoe accessory sales, one thing it's especially well suited for is collaborative software development, including announcements, documentation, managing a software project's web site, and even acting as an issue tracking system.
Building a project wiki with ikiwiki
The simplest way to use ikiwiki is to build static HTML files from source wiki files. This example builds a wiki for an imaginary software project. The wiki source files used in this example are available in the examples/softwaresite section of ikiwiki's documentation.
wiki$ ls Makefile bugs.mdwn doc/ download.mdwn news/ bugs/ contact.mdwn doc.mdwn index.mdwn news.mdwn wiki$ make ikiwiki `pwd` html --wikiname FooBar --plugin=goodstuff \ --exclude=html --exclude=Makefile wiki$ w3m -dump html/doc/faq.html FooBar/ doc/ faq FooBar frequently asked questions. 1. Is this a real program? 2. Really? _Is this a real program?_ No, it's just an example. _Really?_ Yes, really. Links: contact doc Last edited Wed Nov 22 09:58:35 2006
If all you need is a simple static set of pages that can be put up on a web site, or shipped with a software package, this is a good starting point. The examples included with ikiwiki include pages for a news feed for the project (with RSS), an issue tracker, and other pages users expect to see on a project's website. You can check the wiki-format text into revision control as part of the software project, and tie it into the build system using the Makefile.
Ikiwiki can also be tied into the post-commit hook of your revision control system, so that whenever a developer commits a change to a wiki page in revision control, the project's web site is automatically updated. The ikiwiki tutorial explains in detail how to set this up using the Subversion, Git, TLA, and Mercurial revision control systems.
The tutorial also explains how to configure ikiwiki so that users can edit pages using a web interface, with their changes committed back into revision control. After all, one of the benefits of keeping a project's docs in a wiki is to make it easy for users to improve them, so that busy software developers don't have to. And if the wiki is being used for issue tracking, this will let users post and follow up on bug reports.
Using a wiki for issue tracking?
You might be wondering exactly how a wiki can be used as an issue tracking system. Three key parts of ikiwiki come together to create an issue tracker: pages, tags, and inlining.
Each issue is described on a separate page in the
wiki. There can also be an associated Discussion page,
as well as other related subpages that can be used
to hold files used to reproduce the bug, or patches,
or other related files. Since each issue is a page,
standard wiki links can be used to link related
issues, or link issues with other pages in the wiki.
Each issue has its own unique URL. Since ikiwiki
supports subdirectories, it's usual to keep all the
bugs in a
bugs/ subdirectory. You might prefer
to separate bugs and todo items, with todo items in
their own 'todo/' subdirectory.
While directories are useful for broad hierarchical grouping, tags are better for categorizing issues as bugs, wishlist items, security issues, patches, or whatever other categories are useful. Bugs can be tagged "moreinfo", "done", "unreproducible", etc, to document different stages of their lifecycle. A developer can take ownership of a bug by tagging it with something like "owner/Joey".
To tag a wiki page, edit it and add text such as "[[!tag done]]". Note that
adding a wiki link to "[[done]]" will have the same categorisation effect
as a tag, but the link will show up in the body of the page, which is a
nice effect if used in a sentence such as "This was [[done]] in version
1.1.". Another way to close a bug is to move it out of the
subdirectory, though this would prevent it from showing up in a list of
Inlining is how ikiwiki pulls individual issue pages together into something larger, be it a page listing recently opened bugs (with a form to let a user easily post a new bug), or a page listing recently closed bugs, or an index of all bugs, or all wishlist items, or RSS feeds for any of these. A flexible syntax is used for specifying what kind of pages should be inlined into a given page. A few examples:
A typical list of all open bugs, with their full text, and a form to post new bugs.
[[!inline pages="bugs/* and !link(done) and !*/Discussion" actions=yes postform=yes show=0 rootpage="bugs"]]
Index of the 30 most recently fixed bugs.
[[!inline pages="bugs/* and link(done) and !*/Discussion" sort=mtime show=30 archive=yes]]
Index of the 10 most recently active bugs.
[[!inline pages="bugs/* and !link(done) and !*/Discussion" sort=mtime show=10]]
Open security issues.
[[!inline pages="bugs/* and link(security) and !link(done) and !*/Discussion"]]
Full text of bugs assigned to Joey.
[[!inline pages="bugs/* and link(owner/Joey) and !link(done) and !*/Discussion" show=0]]
It may seem strange to consider using a wiki for issue tracking when there are several dedicated bug tracking systems, like Bugzilla, that handle all aspects of it already. The weakest part of using ikiwiki for issue tracking, and certainly the place where a dedicated bug tracker like Bugzilla shines in comparison, is storing and querying structured data about bugs. Ikiwiki has little structured data except for page filenames and tags, so if you need lots of queryable data such as what versions a bug affects and what version it was fixed in, ikiwiki may not be a good fit for your issue tracking.
On the other hand, by using a wiki for issue tracking, there is one less system for users and developers to learn, and all the flexibility of a wiki to take advantage of. Ikiwiki even supports OpenID, so it's easy for users to use it for filing bugs without going through an annoying registration process.
Developers who work offline, or at the other end of a slow connection, might appreciate having a full copy of the project bug tracking system, too.
Realistically, there are plusses and minuses to letting users edit a software project's documentation in a wiki. Like any wiki, to be successful, some review is needed of the changes users make. In some cases it will be easiest to limit the pages that users are allowed to edit. Still, keeping the wiki open for user edits will probably turn up some passionate users who prove very useful at filling in holes in the documentation and cleaning up the site.
Programmers are supposed to be bad at writing documentation, and putting a project's docs into a wiki might not solve that. But it can make it a little bit easier. Consider a programmer who's just coded up a new feature. He can commit that to a development branch in revision control, and then go update the docs on the web site to document it. But the feature isn't available in a released version yet, so it's probably easier to skip updating the website. Maybe once it's released, the web site will be updated to mention the feature, but maybe (probably) not.
Now consider what happens if instead the web site is a wiki that has its source included in the project's revision control system. The programmer codes up the feature, and can easily update the docs in the wiki to match. When he commits his changes to a development branch, the docs are committed too. Later, when that change is merged to the release branch, the doc changes are also merged, and automatically go live on the web site. Updating the documentation to reflect each change made and publishing it on the website has become a standard part of the programmer's workflow.
But this still requires programmers to write documentation, so maybe it still won't work. Let's go back a step. Before the programmer wrote that feature, he probably got some requests for it, and maybe he developed those into a specification for how the feature should work. Since ikiwiki can be used as an issue tracker, the requests were made using it, and were collaboratively edited on the wiki, to develop the specification. Once the feature is implemented, that issue can be closed. What better way to close it than to move it out of the issue tracking system, and into the project's documentation? In Subversion:
svn mv wiki/bugs/new_feature.mdwn wiki/doc/
If the spec is written well enough to be useful for end user documentation, the programmer doesn't have to write a lot of docs after all; that was done when the feature was designed. By using ikiwiki for issue tracking, plus editing the spec, plus documentation, plus the website, each of these steps has built on the other and the programmer has had to do less busywork.
A different example of how ikiwiki can tie things together is how a security hole might be handled. First it's discovered, and a bug filed about it. When it's fixed, the commit that fixes the bug can include a change to the bug's page, marking it as done. Since it's a security hole, the project needs to make an announcement right away so users will know they need to upgrade. This announcement can be added to the wiki's news feed, and committed along with the fix, and the announcement can use a regular wiki link to link to the bug that describes the security hole in detail. If the security hole also affects an older version of the software, the fix, along with the wiki documentation for that fix, can be merged into the branch for the older version.
Another benefit of keeping the bug tracking system in revision control with the wiki is that it allows for disconnected development. So there's no need to be online to review the project's bug list, and there's no need to remember to close fixed bugs once you're back online.
For fans of distributed revision control, ikiwiki opens even more possibilities. With a project's website and issue tracker kept in distributed revision control with the project, these become distributed as well, rather than centralized appendixes to the project. Developers can pass around changesets that not only fix bugs, but mark them as done. If large changes are being made in someone's branch, they can choose to put up their own version of the website, use it to track bugs for that branch, and when the branch is ready, all these changes can be merged back into the mainline of the project.