Using Schema-less in Django with Postgres | Hacker News Reader

Using Schema-less in Django with Postgres

craigkerstiens.com

If you use SQLite, you can use Goatfish[1]. It's not production-ready or as robust as Postgres's hstore, but it's embedded (since it works with SQLite) and also allows you to index various properties.

[1] https://github.com/stochastic-technologies/goatfish

You should not use SQLite in production.

This is true, I should have said "it's not mature".

I downvoted you. Here's why:

I've deployed SQLite to a large number of production software, from embedded systems to web servers handling thousands of daily vistors. In fact, SQLite is my tool of choice when it comes to a small/medium relational-db-based web project.

Sqlite is not a toy db. It is an impressive piece of software engineering.

So, unless you have a very narrow definition of the word production, you should do read up on SQLite (the official website is excellent) and seriously consider it for your next project.

SQLite should absolutely be used in production, where it makes sense. The one area where sqlite has a big disadvantage is concurrent writes. If you have read-only data, cached data, etc. then sqlite can happily be used in a high traffic production system, with millions of rows and many GB of data. It's also a good store for configuration data, or something like an admin backend. The portability of sqlite database files is a really nice feature.

I'd been meaning to publish this post on doing a similar thing with Backbone and Node.js. Now seems as good a time as any...

http://pirt.co.uk/post/25095198148/hybrid-sql-nosql

Underreported but very useful are the omnibus GIN/GiST indexes for hstores, which let you index all the fields at once, not unlike the Postgres Full Text Search feature:

http://www.postgresql.org/docs/9.1/static/hstore.html#AEN133...

It's more expensive to update than a more selective index, of course, but for the ultimate in "do what I mean" optimization, this is it.

Unfortunately, I don't know if ORMs generate the right kind of SQL to use those indexes, yet, but I don't see why they couldn't.

For those interested there's also a Rails equivalent post at - http://schneems.com/post/19298469372/you-got-nosql-in-my-pos...

And my railsconf talk: http://confreaks.com/videos/891-railsconf2012-schemaless-sql...

So what exactly is the benefit of this? Is it superfast like HandlerSocket?

Reply on news.ycombinator.com