Re: Are statistics gathered on function indexes?

Tom Lane wrote:
> As you can see, the estimate drops off by about a factor of 90 per
> added character.  This is probably too much, but it's not that easy
> to determine what the ratio ought to be.  The critical code involved
> in this is convert_string_to_scalar in backend/utils/adt/selfuncs.c;
> the ratio per character is essentially the same as the character range
> that it induces from the available values.  Feel free to propose a
> better implementation if you can think of one.

FYI, 7.3 will drop off quicker than <=7.2.X:

  * Change FIXED_CHAR_SEL to 0.20 from 0.04 to give better selectivity (Bruce)

--
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 853-3000
  +  If your life is a hard drive,     |  830 Blythe Avenue
  +  Christ can be your backup.        |  Drexel Hill, Pennsylvania 19026