Thread: Anyone working on the TOAST items on the TODO list?
If not, I'm going to make a little personal project of them for myself (targeting 9.6). -- Bill Moran
On Sat, Jul 25, 2015 at 11:05 PM, Bill Moran <wmoran@potentialtech.com> wrote: > If not, I'm going to make a little personal project of them for myself > (targeting 9.6). If you mean those three ones, not that I know of: https://wiki.postgresql.org/wiki/Todo#TOAST -- Michael
On 07/25/2015 07:05 AM, Bill Moran wrote: > > If not, I'm going to make a little personal project of them for myself > (targeting 9.6). > Nope. In fact, even the one which was 90% complete (replacing zlib with lz4) completely dropped off the radar. -- Josh Berkus PostgreSQL Experts Inc. http://pgexperts.com
On Sat, 25 Jul 2015 11:39:15 -0700 Josh Berkus <josh@agliodbs.com> wrote: > On 07/25/2015 07:05 AM, Bill Moran wrote: > > > > If not, I'm going to make a little personal project of them for myself > > (targeting 9.6). > > > > Nope. In fact, even the one which was 90% complete (replacing zlib with > lz4) completely dropped off the radar. Interesting ... I wasn't looking at doing that, rather adjusting the decision logic on when to compress, as well as making the trigger values configurable so as to open the TOAST logic up to easy user configuration at the DB as well as the table level. I figure this will also need enough testing to feel like we're shipping with reasonable default values. I got the impression that this is something that's generally desired (it's something _I'd_ like to have). If I'm not misunderstanding, then I'll get started on a proposed patch. -- Bill Moran
Bill Moran <wmoran@potentialtech.com> writes: > On Sat, 25 Jul 2015 11:39:15 -0700 > Josh Berkus <josh@agliodbs.com> wrote: >> Nope. In fact, even the one which was 90% complete (replacing zlib with >> lz4) completely dropped off the radar. > Interesting ... I wasn't looking at doing that, rather adjusting the > decision logic on when to compress, as well as making the trigger values > configurable so as to open the TOAST logic up to easy user configuration > at the DB as well as the table level. I figure this will also need enough > testing to feel like we're shipping with reasonable default values. I have a vague feeling that we've accreted some code that has dependencies on the current TOAST behavior. Unfortunately it's no more than a vague feeling so I can't tell you where to look; but it'd be a good idea to look around and/or test, rather than just assume we can let users frob these knobs to whatever random settings they feel like. I don't want to discourage you from the idea of making these things accessible, just suggesting that there may be more work to do than simply making them accessible. There may be a need for fairly strict sanity checking on the values. regards, tom lane
On Sat, 25 Jul 2015 16:06:28 -0400 Tom Lane <tgl@sss.pgh.pa.us> wrote: > Bill Moran <wmoran@potentialtech.com> writes: > > On Sat, 25 Jul 2015 11:39:15 -0700 > > Josh Berkus <josh@agliodbs.com> wrote: > >> Nope. In fact, even the one which was 90% complete (replacing zlib with > >> lz4) completely dropped off the radar. > > > Interesting ... I wasn't looking at doing that, rather adjusting the > > decision logic on when to compress, as well as making the trigger values > > configurable so as to open the TOAST logic up to easy user configuration > > at the DB as well as the table level. I figure this will also need enough > > testing to feel like we're shipping with reasonable default values. > > I have a vague feeling that we've accreted some code that has dependencies > on the current TOAST behavior. Unfortunately it's no more than a vague > feeling so I can't tell you where to look; but it'd be a good idea to look > around and/or test, rather than just assume we can let users frob these > knobs to whatever random settings they feel like. > > I don't want to discourage you from the idea of making these things > accessible, just suggesting that there may be more work to do than simply > making them accessible. There may be a need for fairly strict sanity > checking on the values. Good to know. I guess it's a good thing that I'm giving myself a full year to work on this ... -- Bill Moran
On Sat, Jul 25, 2015 at 04:06:28PM -0400, Tom Lane wrote: > Bill Moran <wmoran@potentialtech.com> writes: > > On Sat, 25 Jul 2015 11:39:15 -0700 > > Josh Berkus <josh@agliodbs.com> wrote: > >> Nope. In fact, even the one which was 90% complete (replacing zlib with > >> lz4) completely dropped off the radar. > > > Interesting ... I wasn't looking at doing that, rather adjusting the > > decision logic on when to compress, as well as making the trigger values > > configurable so as to open the TOAST logic up to easy user configuration > > at the DB as well as the table level. I figure this will also need enough > > testing to feel like we're shipping with reasonable default values. > > I have a vague feeling that we've accreted some code that has dependencies > on the current TOAST behavior. Unfortunately it's no more than a vague > feeling so I can't tell you where to look; but it'd be a good idea to look > around and/or test, rather than just assume we can let users frob these > knobs to whatever random settings they feel like. You might be thinking of pg_upgrade, which _used_ to assume that the old and new cluster tables had the same TOAST usage status. -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + Everyone has their own god. +