Re: Vacuum explained

Tom Lane writes:

> > Unused = # of tuples awaiting being reused
>
> These are not actually tuples, but tuple slots (page item pointers)
> that are open for re-use.

Maybe this is a good place to ask this: How do you actually reuse a tuple (or
the space the tuple needed on disk) if the number of bytes needed to store a
new tuple depends on the number of non null columns, the variable length
columns etc. (If I understood the documentation right, you store the tuples
within a page packed, i.e. variable length tuples are stored one after the
other, sometimes with alignment between the tuple header and the tuple data.)

Do you try to reuse tuples with a similar storage size (or something like
that) to prevent fragmentation or don't you have fragmentation by design (and
I miss something).

Any pointers to documentation or into the source code are welcome!

Regards,

Tilo Schwarz