[Haskell-cafe] Fwd: Data.Time

[Yitzchak Gale]

Ashley, heads up! I am CCing you on this message because
I think a problem has been found with Data.Time.Format.
Please see below.

Daniel Patterson wrote:
> I've found most of the time library to be quite useful, but
> the parsing to be worthless

The API for parsing and rendering time in Data.Time is
based on the standard API for that in C - like the libraries
in most languages. It's pretty standard stuff.

I'm sure it can be improved upon though. If you have a useful
alternative time parsing library, please release it on Hackage.

> Specifically, the formatTime function, if it ever strips out
> padding (by zeros or spaces), results in a time that is unparseable.
> The fact that formatTime and parseTime are not capable of being
> inverses of each other seems like a major flaw,

Well, it is possible to render dates and times in a way
that loses information, so you can't expect those to
be complete inverses of each other.

> when you think that this is not a parseable date:
> 2011/1/30 (because the month must be padded by zeros).

Hmm, that does seem wrong. The C API allows that to be
parsed using the format "%Y/%m/%d", since the leading zero
for %m and %d are optional when parsing.
See, for example,
http://pubs.opengroup.org/onlinepubs/009695399/functions/strptime.html

I am including Ashley as a CC on this. Perhaps he will
fix it.

It looks like the function "parseValue" is wrong for the
following format letters: "CdHIjmMSuwW".
For each of those, "decimal n" should replaced by
something that requires at least 1 digit and allows up to n.
Then the various instances of ParseTime should be
checked, but they should be OK because they all seem
to use "read".

Thanks,
Yitz