datetime precision is silently truncated by web2py about web2py

[niphlod]

From [email protected] on December 04, 2011 14:08:27

I would be ok with 1 but at the dal level (no truncation when retrieving) but I do not want to milliseconds to appear in forms. It only confuses users. Perhaps the milliseconds could be stripper by the IS_DATETIME/IS_TIME validators?

[niphlod]

From [email protected] on February 03, 2012 21:40:57

Attaching a diff against git trunk ( 8fbae09 ). Only lightly tested, but seems to do the job for me; please review and comment, as I test (and possibly fix problems), I'll fix it so that it can be integrated with trunk. I haven't tested it enough, so I don't claim it is ready for integration - but I would be happy for comments.

Notes: as is, requires Python>=2.6; I can incorporate e.g. http://stackoverflow.com/questions/698223/how-can-i-parse-a-time-string-containing-milliseconds-in-it-with-python for 2.5, or disable fractional second mode for 2.5

I can't test with informix / oracle which have special cases.

Added tests to the test_dal.py

Attachment: dal.patch test_dal.patch

[niphlod]

From [email protected] on February 05, 2012 16:20:36

took a look at the dal.patch, overall it looks good by visual inspection. some comments from my perspective:

• i would utilize this on GAE where i still have apps using python2.5 (yes i know i need to upgrade, but GAE still considers python2.7 support as beta). i guess you could argue that i need to upgrade, but i think since web2py still support 2.5 this should as well
• in parse_time and parse_datetime you don't account for your WARNING setting, and you slightly change the parsing behavior i think.

otherwise looks OK to me.

[niphlod]

From [email protected] on February 05, 2012 16:46:48

Thanks.

For python2.5, that also needs to include the subsecond time parser from stack overflow above (or something equivalent). mxDateTime has a good parser, but it's a shame to pull it in as a dependency just for this.

About warning setting in prase: thanks, I'll fix that. I'm not actually using warning or parse_* myself, so I missed that (I configure sqlite to return them as datetime objects)

you slightly change the parsing behavior i think.

Yes, but I think the change is only to reject wrong things that would have previously been accepted (silently and wrongly), such as: "1:2:3:4:". But when in truncate (compatibility) mode, it should remain completely the same as before, I think?

[cassiobotaro]

@niphlod @mdipierro The proposed fix is already closed, can we close this issue too? Related to https://github.com/web2py/web2py/pull/542/files