BMO currently supports five—count ‘em, five—ways to search for bugs. Whenever you have five different ways to perform a similar function, you can be pretty sure the core problem is not well understood. Search has been rated, for good reason, one of the least compelling features of Bugzilla, so the BMO team want to dig in there and make some serious improvements.
At our Portland get-together a couple weeks ago, we talked about putting together a vision for BMO. It’s a tough problem, since BMO is used for so many different things. We did, however, manage to get some clarity around search. Gerv, who has been involved in the Bugzilla project for quite some time, neatly summarized the use cases. People search Bugzilla for only two reasons:
- to find a set of bugs, or
- to find a specific bug.
That’s it. The fact that BMO has five different searches, though, means either we didn’t know that, or we just couldn’t find a good way to do one, or the other, or both.
We’ve got the functionality of the first use case down pretty well, via Advanced Search: it helps you assemble a set of criteria of almost limitless specificity that will result in a list of bugs. It can be used to determine what bugs are blocking a particular release, what bugs a particular person has assigned to them, or what bugs in a particular Product have been fixed recently. Its interface is, admittedly, not great. Quick Search was developed as a different, text-based approach to Advanced Search; it can be quicker to use but definitely isn’t any more intuitive. Regardless, Advanced Search fulfills its role fairly well.
The second use of Search is how you’d answer the question, “what was that bug I was looking at a couple weeks ago?” You have some hazy recollection of a bug. You have a good idea of a few words in the summary, although you might be slightly off, and you might know the Product or the Assignee, but probably not much else. Advanced Search will give you a huge, useless result set, but you really just want one specific bug.
This kind of search isn’t easy; it needs some intelligence, like natural-language processing, in order to give useful results. Bugzilla’s solutions are the Instant and Simple searches, which eschew the standard Bugzilla::Search module that powers Advanced and Quick searches. Instead, they do full-text searches on the Summary field (and optionally in Comments as well, which is super slow). The results still aren’t very good, so BMO developers tried outsourcing the feature by adding a Google Search option. But despite Google being a great search engine for the web, it doesn’t know enough about BMO data to be much more useful, and it doesn’t know about new nor confidential bugs at all.
Since Bugzilla’s search engines were originally written, however, there have been many advances in the field, especially in FLOSS. This is another place where we need to bring Bugzilla into the modern world; MySQL full-text searches are just not good enough. In the upcoming year, we’re going to look into new approaches to search, such as running different databases in tandem to exploit their particular abilities. We plan to start with experiments using Elasticsearch, which, as the name implies, is very good at searching. By standing up an instance beside the main MySQL db and mirroring bug data over, we can refer specific-bug searches to it; even though we’ll then have to filter based on standard bug-visibility rules, we should have a net win in search times, especially when searching comments.
In sum, Mozilla developers, we understand your tribulations with Bugzilla search, and we’re on it. After all, we all have a reputation to maintain as the Godzilla of Search Engines!
