While trying to clean up ParseDateTime so it works reliably with full
timezone names, I found out about a "feature" that so far as I can tell
has never been documented except in comments in datetime.c. The
datetime input code tries to recognize what it calls "POSIX time zones",
which are timezone abbreviations followed by an additional hour/minute
offset:
/* DecodePosixTimezone()
* Interpret string as a POSIX-compatible timezone:
* PST-hh:mm
* PST+h
* PST
* - thomas 2000-03-15
However this doesn't actually work in all cases:
regression=# select '12:34:00 PDT+00:30'::timetz;
timetz
----------------
12:34:00-07:30
(1 row)
regression=# select '12:34:00 PDT-00:30'::timetz;
ERROR: invalid input syntax for type time with time zone: "12:34:00 PDT-00:30"
(The behavior varies depending on which PG release you try it with, but
I can't find any that produce the expected result for a negative
fractional-hour offset.)
This syntax is ambiguous against some full timezone names present in the
zic database, such as "GMT+0", and it's also responsible for a number of
really ugly special cases in the datetime parser. In view of the fact
that it's never entirely worked and never been documented, I'm inclined
to take it out. Comments? Is anyone actually using this?
regards, tom lane
