HN Explorer

Ask HN: How to version control DB data?

8 points by cammil ↗ HN
What are the standard/recommended ways to store version-controlled, database data? I'm curious about solutions that would enable querying on tables/rows but restricted to those values at a certain point in time.

Fyi, I have posted to SO, with little satisfaction: http://stackoverflow.com/questions/8522766/what-are-the-standard-recommended-ways-to-store-version-controlled-database-dat

9 comments

[ 3.5 ms ] story [ 27.5 ms ] thread
JS_startup
I don't know of any type of version control that will work like that. You could, of course, add your databases to an SVN repository but you won't be able to branch/merge individual rows in the database.

I'd re-examine your requirements on this one. You normally wouldn't be doing intense revision control directly inside of the database

mattvot
WordPress will just create a new entry in the db with a different status, like draft. The current version would be 'published'.
jacksondeane
If I am reading this correctly, you want to maintain multiple versions of the same blog post, correct? This isn't really 'version control' (git, svn, etc...) in the traditional sense .

At a high level, you could just keep a record of every version of every blog post, then separately maintain which entry is most current, or active, or whatever metric you want to display/query. This will let you maintain all versions of the blog post and you can run diffs on different versions, revert to older versions and branch from any version.

whichdan
I'm curious - has anyone here had experience doing this with a few million rows of data? Even a few revisions each means the table would grow rapidly. You'd also (probably) need three operations for each insert:

1. Insert the new row

2. Find the previous row

3. Set previous row to `active`=0

Are there performance tricks to keep this reasonable? I'd also wonder how to modify the table itself (dropping/adding columns) when there would potentially be selects + inserts + updates running 100% of the time - but I realize that's borderline out of scope of the OP's question.

blcArmadillo
Not sure about the other questions but in regards to changing the actual table structure. If you're dropping columns you'd first make the changes to your code base so that it no longer relies on those columns. If you have long running queries you'd just have to wait till the old ones using those columns are finished and then you could change the DB. And you'd do it in just the opposite order when you wanted to add a column.
pasbesoin
I haven't looked into the details of what they do, but with respect to your concern about blog entries (on your SO post), I'm immediately put in mind of MediaWiki (the engine behind Wikipedia).
equark
This is already implemented in HBase. View your table as 3D: row, column, time.