Re: parse_time()

[Bruce Korb]

Hi Bruno,

On Sun, Nov 2, 2008 at 10:52 AM, Bruno Haible <address@hidden> wrote:
> Hi,
>
> Bruce Korb wrote:
>> I had need of a function to convert some variation of a time (duration)
>> specification into a count of seconds.  For your amusement and optional
>> inclusion...(I *think* I've got papers on file for gnulib....)
>>
>> It accepts several formats:
>>
>>    [DD d] [HH h] [MM m] [SS s]
>>    [DD d] [[HH:]MM:]SS
>>    [DD d] [HH h] [MM:]SS
>>
>> though it is unhappy if it finds nothing.
>
> Yes, I think if several programs can agree to parse durations in the same
> way, this can only be beneficial.

My motivation was that I noticed that someone implemented what I implemented
as ``timed-task'' as ``timeout'' and I used basically this code to parse the
"how long should I wait" option.  Though "timeout" only accepts it as
an operand.

> Looking at wikipedia [1], I would find it good if
>  1) the function was called 'parse_duration', not 'parse_time' (since "time"
>     often denotes a time instant within a day),

Sure.  My coin landed "heads" instead of "tails" when I was deciding.

>  2) the three duration formats described in [1] were also supported.

3?  I only see one, though with lots of optional fields.  That format also
doesn't allow for visual space.  Anyway, it's trivial enough to look for a
string starting with 'P' or 'T' and handle the ISO-8601 duration.
The little question remaining though is "how many seconds in a year"
and, more importantly, "how many seconds in a month"?  In other words,
if some one in February were expecting P1M to represent a 28 day duration
and a 31 day duration had it been March, um, well, I think maybe they should
implement it.  :)  Dealing with "leap seconds" would get extremely
over the top.

Cheers - Bruce