Related to #436 - hasMany with >5 tables #554

Closed
hans-d opened this Issue Jun 23, 2012 · 3 comments

Comments

Projects
None yet
2 participants
Contributor

hans-d commented Jun 23, 2012

Performance can be gained on MySQL by using DISTINCT over GROUP BY in case of hasMany pre-query.
In my case, with 6 related tables this is from 1 min 25 sec -> 0.005 sec

Some background

When having a model with hasManys, and using the with function an initial query is fired. This looks like

SELECT Table1.key 
FROM `table1` AS `Table1`  
LEFT JOIN `relatedtableN` AS `aliasforN` ON [...] -- repeat for every relationship
WHERE [whatever]
GROUP BY Table1.key 

I have a model with 6 hasMany relationships (and need them). Using a find first this select took running on a local MySQL* this took 1min25sec. Rewritten as a SELECT DISTINCT it was done in 0.005sec (MyISAM tables, InnoDb about same timings).
Having only 5 related tables will drop the time to about 2.02 sec in the GROUP BY flavour (DISTINCT still 0.005 sec)

Directly queried on a MySQL 5.5.16, Win7x64 8GB dual core, data tables are not that big (main table 7k rows, each related table about 20k rows), all the fields in the joins are indexed. Every query executed multiple times to get good readings.

It just so happens that all related tables in this case are all the same (the various hasMany have distinctive conditions),

Contributor

jails commented Jun 23, 2012

Is this issue can be an additionnal reflexion topic for #530 ?

Contributor

hans-d commented Jun 23, 2012

nb: the memory issue was caused by my logging statements (dumping the results)

Contributor

jails commented Dec 12, 2012

This should have been solved in #705. The DISTINCT clause is now used in replacement of the GROUP BY clause for all "hasMany pre-queries".

jails closed this Dec 12, 2012

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment