Comments on: MySQL Stored Function: parsing a JSON encoded string

By: Joe Grossberg

Joe Grossberg — Mon, 14 Apr 2008 21:00:00 +0000

Interesting proof-of-concept.

I’m curious about the motivations for this, since you said, "Although we don’t often need to query it directly, from time to time, it makes things a bit easier/faster". But then you say performance was "pretty slow", so why did you do it in SQL instead of Ruby?

Also, by limiting it to 255 chars, I’m guessing that you know the JSON keys’ names. Why not just split that JSON string and insert into multiple columns? (You say that it has no nested hashes nor arrays.)

And how do you avoid truncating that JSON when it’s too long (thus corrupting it)?

By: Warren

Warren — Mon, 14 Apr 2008 21:00:00 +0000

Long story short, we’re storing large amounts of data for analytics purposes. We pull out useful info that we know we’ll be using frequently into distinct columns so we can query it faster, but we throw the rest of the data (which has no particular structure) into a TEXT field as JSON. In terms of Ruby vs. MySQL, MySQL can process the JSON we need in seconds and discard the rows we don’t need where as just the transfer time of getting all that data into Ruby and then processing it is minutes to hours.

By: Brad Baumann

Brad Baumann — Mon, 14 Apr 2008 21:00:00 +0000

I think this will be very useful once it’s tweaked a bit.
I’m having some problems with the function detecting value types other than strings.

For example:
SELECT JSON(
‘{"fieldname":12345,"fieldname2":"test string"}’, ‘fieldname’
)

Returns: "fieldname2"

If you do this for "fieldname2", you’ll get the right return value of "test string".

I’ll try to code a quick work around… unless theres something I’m missing?