[htoolkit-users] bug in parsing timestamp from PostgreSQL

[Jacob N. <ja...@jf...>]

I've been using HSQL with a PostgreSQL database on Linux, and things have 
been going quite swimmingly. Then I started getting errors like this:

start_time is 2005-11-19 16:20:47-08
end_time is 2005-11-19 16:21:33.863-08
*** Exception: HSQL.hsc:323:13-54: Non-exhaustive patterns in case

The first two lines are generated when I pull timestamp fields out as 
Strings. The exception on the third line occurs when I try to read the 
start_time field as a ClockTime. The end_time field works fine. The 
difference appears to be that PostgreSQL doesn't print the millisecond 
field on start_time, since it's all zeros.

Armed with this observation, I jumped into the HSQL code. I'm fairly new 
to Haskell and certainly new to parser combinators, but I found a
suspicious comment character lurking in the getDateTimeTZ instance of
fromSqlValue, on line 379 of HSQL.hsc:

    fromSqlValue SqlDateTimeTZ s = f_read getDateTimeTZ s
       where
          getDateTimeTZ :: ReadP ClockTime
          getDateTimeTZ = do
             (year, month, day, hour, minutes, seconds) <- readDateTime
             char '.' >> readDecP -- ) `mplus` (return 0)
             tz    <- parseTZ
             return (mkClockTime year month day hour minutes seconds (tz*3600))

When I got rid of it and rebuilt, I no longer got the exception.  If I
understand correctly, the mplus clause allows the parser to continue
on if the field has no milliseconds part (although we wouldn't use the
result even if it were there for us to read). Is that correct?

Is this the right change to make?

Thanks for your help,
jacob