jim@reptiles.org (Jim Mercer) writes:
> [ concern about speed of converting datetime values to/from text for
> Postgres ]
FWIW, I used to be really concerned about that too, because my
applications do lots of storage and retrieval of datetimes.
Then one day I did some profiling, and found that the datetime
conversion code was down in the noise. Now I don't worry so much.
It *would* be nice though if there were some reasonably cheap documented
conversions between datetime and a standard Unix time_t displayed as a
number. Not so much because of speed, as because there are all kinds
of ways to get the conversion wrong on the client side --- messing up
the timezone and not coping with all the Postgres datestyles are two
easy ways to muff it.
BTW, I believe Thomas is threatening to replace all the datetime-like
types with what is currently called datetime (ie, a float8 measuring
seconds with epoch 1/1/2000), so relying on the internal representation
of abstime would be a bad idea...
regards, tom lane