Re: Efficient Boolean Storage

It is unintuitive to me that the low byte be padded.  Shouldn't it be
the high byte?

JLL

Joe Conway wrote:
>
> Csaba Nagy wrote:
> > BTW, is or is not using Postgres a string to store the bit data types ? I
> > have seen a few people asking this on the list, but no straight answer as of
> > yet... the docs really don't mention this aspect.
>
> Use the source ;-)
>
>  From varbit.c
>
>   /*----------
>    *      attypmod -- contains the length of the bit string in bits, or for
>    *                         varying bits the maximum length.
>    *
>    *      The data structure contains the following elements:
>    *        header  -- length of the whole data structure (incl header)
>    *                   in bytes. (as with all varying length datatypes)
>    *        data section -- private data section for the bits data structures
>    *              bitlength -- length of the bit string in bits
>    *              bitdata   -- bit string, most significant byte first
>    *
>    *      The length of the bitdata vector should always be exactly as many
>    *      bytes as are needed for the given bitlength.  If the bitlength is
>    *      not a multiple of 8, the extra low-order padding bits of the last
>    *      byte must be zeroes.
>    *----------
>    */
>
> Joe
>
> ---------------------------(end of broadcast)---------------------------
> TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org