[tap-l] TAP Datetime discussion

[Michael G Schwern]

Eric Wilhelm wrote:
>> And the proposal is here:
>> http://testanything.org/wiki/index.php/TAP_datetime
> 
> It's not clear from that what the intended usage of this datetime is.  
> It sounds like:
> 
> 1.  TAP-level metadata (e.g. stream and test start/end times.)
> 2.  TAP diagnostics.

Yep, you've got it.


> With #1, it's specified in the grammar.  Thus, the producer and consumer 
> have agreed in advance that a given key (e.g. "harness_start_time") 
> will be in that format.
> 
> But #2 is arbitrary user data.  This means the datetime format has no 
> meaning unless it is identified as such.  The example demonstrates this 
> weakness:
> 
>    not ok 1 - date of the par-tay
>      ---
>      found:  2007-09-01
>      wanted: 2007-09-02
>      ...
> 
> That's fine if the recommendation pertains to #2 only in terms of making 
> dates and times consistently readable for the user.  But, as far as 
> parsing is concerned, a datetime stored in an arbitrary key is only a 
> string unless it is identified.  If the #2 case intends to be more 
> ambitious

It doesn't, not at this time.  It's just establishing a suggestion for a
particularly common, contentious formatting issue.  Since we need to do it for
the metadata information anyway, might as well throw it out there.


> some sort of typing is needed.
> 
>   !TAP::datetime 2007-09-01
> 
> I doubt ^-- that is it.  Maybe just a string "date(2007-09-01)" or 
> whatever (or not.)

If we wanted to be ambitious we'd write an actual YAML type.


-- 
Stabbing you in the face so you don't have to.