Re: Query Builder internals

On tor, 2006-02-23 at 14:57 +0200, Jukka Zitting wrote:
> Hi,
> 
> The query_builder.c source file has seen almost a hundred changes
> since it was created in April last year. Despite a number of
> discussions there hasn't been a very clear roadmap or shared
> understanding of what the QB should look like, and as a result the QB
> internals are already showing signs of software rot (see for example
> the _mqb_get_new_property function).
> 
> I'd like to start an effort to clean up and document the QB internals
> in order to build a stronger base for Midgard 2.0 and to solve the
> current performance issues that are more caused by the implementation
> than any inherent design defect.
> 
> To start with I'm planning to set up some unit tests that would be
> able to check the SQL queries generated by the QB for a number of
> common cases. For example the ideal queries for a non-multilang schema
> "foo" would look something like this:
> 
> $qb = new midgard_query_builder("foo");
> ...
> $qb->execute():
> 
>     SELECT ... FROM foo WHERE ...
> 
> $qb->count():
> 
>     SELECT count(*) FROM foo WHERE ...
> 
> $qb->distinct("bar"):
> 
>     SELECT DISTINCT bar FROM foo WHERE ...
> 
> I'm also hoping to extend the PHP unit test suite created by Tarjei to
> more completely cover of the various QB corner cases. Once this is
> done we can feel a lot safer about refactoring the QB internals.
+1!

I allready got some issues in the querybuilder/limit.php testfile, but
that might be due to wrong asumptions about querybuilder.

You'll find me on irc for any discussions needed.
tarjei 

> BR,
> 
> Jukka Zitting
> 
> --
> Yukatan - http://yukatan.fi/ - [email protected]
> Software craftsmanship, JCR consulting, and Java development
-- 
Tarjei Huse <tarjei-y/[email protected]>