Hacker News Reader

Comment by leetrout | Hacker News Reader

I’m always surprised when people haven’t heard of Xapian which is also in this space.

I tell people Xapian is the SQLite of search.

I know it for being the backbone of the equally austere notmuch MUA (https://notmuchmail.org/), where it does a great job of churning instantly through dozens of thousands of mails.

Looks like its GPL not LGPL. Seems not usable for most people

I used this pretty extensively at a startup in 2012 or so. It was very easy to use and modify, but definitely lacked the polish of ElasticSearch at the time. There were SWIG bindings that let us use it with Python.

I'm not surprised. Low-tech website with zero marketing and no big players using it or throwing their weight behind it.

I've perused their docs and the immediate question for me is, why would I use this over another more popular search library or database that already has a wealth of stackoverflow and blog posts?

I'm sure it's a fine tool, but I don't see a reason why an enterprise should hedge their bets on this unknown.

Notmuch[1] is using it. It has a good reputation.

[1]: <https://notmuchmail.org/>

Thanks for the example, I'm aware of Notmuch but had no idea it relied on Xapian. However, back to my original point, there's simply not enough people or large projects/companies that rely on and use Xapian. I'm 99% positive that there is less than a handful of people at any of the last two Fortune 500 companies I've worked for that even knows what Notmuch is. But everyone knows Elastic, MongoDB, etc..

Again, I don't think Xapian is a bad tool or that it's underperforming by any stretch, but OP should not be surprised at all that most people have not heard of it.

I tried, and was never able to get the accuracy of results I expected when searching a SaaS address book. I had better results doing the search in SQL FWIW (though that was slower)

Yeah, I did a POC with Xapian ages ago. You would have a hard time convincing anyone to add a C++ indexer to their stack. I know I know, bindings, and all, but still.

Over one in Java? I’m not a Xapian proponent or anything, but my understanding was that ElasticSearch was slightly annoying to maintain because you need to worry about heap memory usage.

If my app was already running on the JVM, it's a pretty easy decision to run it embedded.

Does it do precise group-by queries? (facets) Can you do custom ranking by providing a class / plugin? (even a c++ plugin would be fine) thanks..

Yes to both https://xapian.org/features

Reply on news.ycombinator.com