From: | Tzahi Fadida <Tzahi(dot)ML(at)gmail(dot)com> |
---|---|
To: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
Cc: | "PostgreSQL Development" <pgsql-hackers(at)postgresql(dot)org> |
Subject: | Re: Problems with the varlena patch in my module |
Date: | 2007-04-07 10:23:47 |
Message-ID: | 200704071323.48001.Tzahi.ML@gmail.com |
Views: | Whole Thread | Raw Message | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
On Saturday 07 April 2007 04:08:34 Tom Lane wrote:
> Tzahi Fadida <Tzahi(dot)ML(at)gmail(dot)com> writes:
> > This is an excerpt from my code:
> > newtset->tids = (bytea *) fastgetattr(tupleTSet, LABELS_ALIGNED,
> > fctx->tupleSetDesc, &isnull);
> >
> > It seems that for an empty bytea (only the size of the header), i get
> > that VARSIZE(newtset->tids)==534765440
> > instead of VARHDRSZ.
>
> That code has always been broken, you just failed to exercise the
It always worked before the patch so it was hard to know that.
You see, i knew from the start that this attribute would never reach certain
sizes so i never thought it would be toasted.
> problem before. There needs to be a detoasting call there, not
> merely a cast. DatumGetByteaP() would probably be appropriate.
Thanks. One more question though, what is the CATALOG_VERSION_NO
of the varlena patch?
The second problem i have (it is a bit long but simple so please bear with
me):
You see, before there were slots (at least stable working slots)
tuples (slot_getattr, etc...), i needed this functionality.
So, i created the following function (which broke because of the varlena patch
on the att_align function):
bool
heap_deformtuple_iterative(HeapTuple tuple,
TupleDesc tupleDesc,
Datum *values,
char *nulls,
deformTupleIterativeState * ds,
int tillAttNum, bool
finishAll)
Where tuple is tuple,
tupleDesc is of course TupleDesc.
values are to store the deformed values (the array is already allocated).
nulls, same.
ds is a pointer to a structure i made to save the state of the
heap_deformtuple in mid work.
tillAttNum is a number of attributes from the start to deform up to.
finishAll - deform all the remaining attributes.
Now, i want to move to the more standard way to do this. However, i want to
keep the structure where i have a tuple->t (the tuple), tuple->v (the values
array), tuple->n (the nulls array). I keep many tuples in memory like this
and i can't change this now. I want to keep in memory only the heaptuple
tuple->t and the deformed arrays tuple->v, tuple->n and possibly another
needed (such as ds) structure for as long as the tuple is not completely
deformed.
What do you recommend (aside from me reimplementing the heap_deformtuple
iterative again hoping it won't break again)?
--
Regards,
Tzahi.
--
Tzahi Fadida
Blog: http://tzahi.blogsite.org | Home Site: http://tzahi.webhop.info
WARNING TO SPAMMERS: see at
http://members.lycos.co.uk/my2nis/spamwarning.html
From | Date | Subject | |
---|---|---|---|
Next Message | Josh Tolley | 2007-04-07 12:01:10 | Re: Fate of pgsnmpd |
Previous Message | Simon Riggs | 2007-04-07 08:50:39 | Re: Eliminating unnecessary left joins |