Quote:
Originally Posted by 1meandog4u
I know, I know, I'm wasting your time and broadband width beating this dead horse, but it's frustrating to run a report to see how I'm doing and it means nothing... "it's a pretty graph and stats viewer" but pretty means nothing if it's not accurate.
Dog
I hadn't really payed much attention to this previously, as it appeared to be 'mostly ok'
- The Stars import always adjusts from ET to GMT (whether or not the hand history is in ET or not). Different importers have different levels of support for timezone conversions.
- The calendar module (thing thing used in both Stats and Grapher for the date)
-- Takes the dates provided (t1=2010-09-04, t2=2010-09-05)
-- Converts using mktime, which adjusts for timezone - i'm +8 GMT(adj_t1 = 2010-09-03 16:00:00) for t1, and adds 23:59:59 to t2 (adj_t2=2010-09-05 15:59:59)
-- Add/Subtracts the amount of time in <general day_start="" from the config, set to 5 by default. Setting this value to 8 gives: adj_t1=2010-09-04 00:00:00 adj_t2=2010-09-05 23:59:59 - ie. back to start. (NOTE: I wasn't aware of this parameter until about 20 minutes ago)
These values are then used in the SQL query:
Code:
AND h.startTime > '2010-09-04 00:00:00'
AND h.startTime < '2010-09-05 23:59:59'
Which may or may not be compensating for current local time at the SQL level.
Places this can mess up:
- The importer might be adjusting the date for ET, when the date in the HH is actually in CET or some other value
- Config - day_start value
- GTK/python config. May not use local time setting - may use BIOS date, or something weirder. This will be dependent on the mktime() implementation for python.
- Database insert: Any of the databases might assume that a time value is either already GMT, or adjust it based on current time (being a flaky interpretation - see GTK/python line) and adjust it for us (using a different mechanism). Also - it might be different on each different database, and each platform the database is implemented on.
Each potential pitfall has multiple verification steps, and is very irritating to track down.
I recommend adjusting the date_start value around until it hits what you are looking for.
Sorrow.