Thread

Re: [HACKERS] Arbitrary tuple size

Zeugswetter Andreas <andreas.zeugswetter@telecom.at> — 1999-07-09T16:01:15Z

>
>I knew there had to be a reason that some tests where BLCKSZ/2 and some
>BLCKSZ.
>
>Added to TODO:
>
> * Allow index on tuple greater than 1/2 block size
>
>Seems we have to allow columns over 1/2 block size for now.  Most people
>wouln't index on them.


Since an index header page has to point to at least 2 other leaf or
header pages, it stores at least 2 keys per page.

I would alter the todo to say:

* fix btree to give a useful elog when key > 1/2 (page - overhead) and not
abort

to fix the:
>
>         FATAL 1:  btree: failed to add item to the page
>

A key of more than 4k will want a more efficient index type than btree
for such data anyway.

Andreas