← Back to team overview

oqgraph-dev team mailing list archive

Progress on VARCHAR

 

Hi Arjen, Antony

I have pushed my work so far to

http://bazaar.launchpad.net/~andymc73/maria/10.0-oqgraph3-varchar

I thought things were flying along according to plan... then I hit how to
interpret the keys (O_o) which from a mysql internals is a bit beyond my
understanding still.


To date, I have:

* changed the latch to VARCHAR
* retained backwards compatbility (with an #ifdef) allowing latch SHORTINT to
still be created, but reporting a deprecation warning
* added warnings to create() and a better message than 'table crashed' if the
OQGRAPH table created incorrectly.  So the user can go show warnings and see
which column they mucked up.
* implemented string -> int latch parsing in search(), which should work, but
I havent been able to test it as yet, because

when doing

SELECT * FROM graph2 WHERE latch = 'breadth_first' AND origid = 1 AND weight = 1;

aka the first select in basic.test, I get an assertion on key store_length
inside ha_oqgraph::records_in_range(); at this point I realised I am a little
stuck, and need some explaining as to how the keys are supposed to work in
this case...

FWIW the existing basic.test still works with the backward compat #ifdef
turned on.  (Except I cant flush at the end, and I get a server memory leak
assertion thereafter.  But I doubt that is related to my changes so far)



--Andrew