Thread: TOAST : partial detoasting : only a small part of a toasted value(for pgpointcloud)
TOAST : partial detoasting : only a small part of a toasted value(for pgpointcloud)
From
Rémi Cura
Date:
Hi dear list,
I have a tricky question about TOASTED memory in Postgres related to the [pgpointcloud](https://github.com/pgpointcloud/pointcloud) extension. (using Postgres 11 if it matters)
So the pgpointcloud store sometimes very large groups of points into one row (TOASTED), something along few kB to few MB. TOAST would be in
EXTERNAL mode (no compression).From what I understand, for the moment it requires to fetch all the data from disk, then de-toast it, then select only the part of the data we are interested in.
Yet I think it is possible to detoast only a subset of the data (and thus fetch only some part of the data), considering that the [doc on toast](https://www.postgresql.org/docs/11/storage-toast.html) says
> " Use of
> (at the penalty of increased storage space) because these operations are optimized
> to fetch only the required parts of the out-of-line value when it is not compressed. "
> " Use of
EXTERNAL will make substring operations on wide text and bytea columns faster > (at the penalty of increased storage space) because these operations are optimized
> to fetch only the required parts of the out-of-line value when it is not compressed. "
So my question is how does it work, how easy would it be to implement for pgpointcloud?
Many thanks for your time and help
Remi-C
Re: TOAST : partial detoasting : only a small part of a toasted value (for pgpointcloud)
From
Tom Lane
Date:
=?UTF-8?Q?R=C3=A9mi_Cura?= <remi.cura@gmail.com> writes: > So the pgpointcloud store sometimes very large groups of points into one > row (TOASTED), something along few kB to few MB. TOAST would be in EXTERNAL > mode (no compression). > Sometimes we only want to access a part of this data (one or several blocks > within the full data). >> From what I understand, for the moment it requires to fetch all the data > from disk, then de-toast it, then select only the part of the data we are > interested in. > Yet I think it is possible to detoast only a subset of the data (and thus > fetch only some part of the data), considering that the [doc on toast]( > https://www.postgresql.org/docs/11/storage-toast.html) says >> " Use of EXTERNAL will make substring operations on wide text and bytea > columns faster >> (at the penalty of increased storage space) because these operations are > optimized >> to fetch only the required parts of the out-of-line value when it is not > compressed. " > So my question is how does it work, how easy would it be to implement for > pgpointcloud? See PG_DETOAST_DATUM_SLICE and users of that macro. regards, tom lane
Re: TOAST : partial detoasting : only a small part of a toasted value(for pgpointcloud)
From
Rémi Cura
Date:
wow,
it was right under my nose.
Thank you very much !
Cheers,
Remi-C
Le mar. 13 nov. 2018 à 19:00, Tom Lane <tgl@sss.pgh.pa.us> a écrit :
Rémi Cura <remi.cura@gmail.com> writes:
> So the pgpointcloud store sometimes very large groups of points into one
> row (TOASTED), something along few kB to few MB. TOAST would be in EXTERNAL
> mode (no compression).
> Sometimes we only want to access a part of this data (one or several blocks
> within the full data).
>> From what I understand, for the moment it requires to fetch all the data
> from disk, then de-toast it, then select only the part of the data we are
> interested in.
> Yet I think it is possible to detoast only a subset of the data (and thus
> fetch only some part of the data), considering that the [doc on toast](
> https://www.postgresql.org/docs/11/storage-toast.html) says
>> " Use of EXTERNAL will make substring operations on wide text and bytea
> columns faster
>> (at the penalty of increased storage space) because these operations are
> optimized
>> to fetch only the required parts of the out-of-line value when it is not
> compressed. "
> So my question is how does it work, how easy would it be to implement for
> pgpointcloud?
See PG_DETOAST_DATUM_SLICE and users of that macro.
regards, tom lane