Re: [HACKERS] jsonb status
It's easy to add support of other operations to hash_ops, so it will be on par with default GIN opclass, at the price of bigger size. We can add it later to contrib/jsonbext. I'm mostly worrying about changing semantics of scalar. On Sun, Mar 23, 2014 at 4:27 AM, Peter Geoghegan p...@heroku.com wrote: On Sat, Mar 22, 2014 at 5:22 PM, Bruce Momjian br...@momjian.us wrote: What did you decide about hashing values in indexes vs. putting them in literally? There are two GIN opclasses supplied. There is a default, which supports more operators (various existence operators - see the documentation). There is an alternative called jsonb_hash_ops that only supports containment, and performs considerably better than the default. Containment *is* the compelling operator to support, though - you can do rather a lot with it. This must be what you're referring to, since I recall you blogged about the response it got at pgConf.EU. Both are available. -- Peter Geoghegan -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] jsonb status
On Sun, Mar 23, 2014 at 12:32:45PM +0400, Oleg Bartunov wrote: It's easy to add support of other operations to hash_ops, so it will be on par with default GIN opclass, at the price of bigger size. We can add it later to contrib/jsonbext. I'm mostly worrying about changing semantics of scalar. On Sun, Mar 23, 2014 at 4:27 AM, Peter Geoghegan p...@heroku.com wrote: On Sat, Mar 22, 2014 at 5:22 PM, Bruce Momjian br...@momjian.us wrote: What did you decide about hashing values in indexes vs. putting them in literally? There are two GIN opclasses supplied. There is a default, which supports more operators (various existence operators - see the documentation). There is an alternative called jsonb_hash_ops that only supports containment, and performs considerably better than the default. Containment *is* the compelling operator to support, though - you can do rather a lot with it. This must be what you're referring to, since I recall you blogged about the response it got at pgConf.EU. Both are available. My question was about whether we decided to abandon the GiST support entirely as there is no method for indexing long values: http://www.postgresql.org/message-id/CAM3SWZSbsedz_YwsQm-chT6B6KX0rh-vke=5nb2gblsem0e...@mail.gmail.com In reading your reply, I now understand that GIN supports hash and non-hash indexing types, which is great! -- Bruce Momjian br...@momjian.ushttp://momjian.us EnterpriseDB http://enterprisedb.com + Everyone has their own god. + -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] jsonb status
On Sat, Mar 22, 2014 at 01:53:06PM -0700, Peter Geoghegan wrote: Attached is v12. I think I've brought this as far as I can. This is mostly just bug fixes, and some additional refactoring. I've incorporated Andres' feedback. The only points that I think worth noting are: * The documentation has been significantly expanded to discuss containment further, since it's a significant part of the feature. This could probably use some polishing and general scrutiny, which is something that Andrew may consider. I didn't have time to go over it to the extent that I'd prefer. * I altered containment semantics slightly. Now, it is not possible for a raw scalar to contain a single-element array of the same scalar, while the inverse is still possible (raw scalars still contain themselves too). This made sense to me, and is consistent with the behavior of the B-Tree operator class, where a raw scalar is not equal to a single-element array of the same scalar. Rather, array is greater than the scalar on the basis of its type alone, as at every other nesting level. The fact that an array can contain a raw scalar is explicitly presented as an exception to the general principle that containment needs to be at the same nesting level. I'm not going to go into the details of the bugs fixed, since no one reported them here, and since they're trivial in nature. For full details, you can review the log of our publicly accessible feature branch. What did you decide about hashing values in indexes vs. putting them in literally? -- Bruce Momjian br...@momjian.ushttp://momjian.us EnterpriseDB http://enterprisedb.com + Everyone has their own god. + -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] jsonb status
On Sat, Mar 22, 2014 at 5:22 PM, Bruce Momjian br...@momjian.us wrote: What did you decide about hashing values in indexes vs. putting them in literally? There are two GIN opclasses supplied. There is a default, which supports more operators (various existence operators - see the documentation). There is an alternative called jsonb_hash_ops that only supports containment, and performs considerably better than the default. Containment *is* the compelling operator to support, though - you can do rather a lot with it. This must be what you're referring to, since I recall you blogged about the response it got at pgConf.EU. Both are available. -- Peter Geoghegan -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] jsonb status
Hi, On 2014-03-13 17:00:33 -0400, Andrew Dunstan wrote: Peter Geoghegan has been doing a lot of great cleanup of the jsonb code, after moving in the bits we wanted from nested hstore. You can see the current state of the code at https://github.com/feodor/postgres/tree/jsonb_and_hstore I've pushed one commit with minor fixes, and one with several FIXMEs to http://git.postgresql.org/gitweb/?p=users/andresfreund/postgres.git;a=shortlog;h=refs/heads/jsonb_and_hstore The most imortant things are: * Jsonb vs JsonbValue is just bad, the latter needs to be renamed, and there needs to be a very clear explanation about why two forms exist and what each is used for. * The whole datastructure doesn't have any sensible highlevel documentation. * I don't really like the JsonbSuperHeader name/abstraction. What does the Super mean? The only interpetation I see is that it's the toplevel header (why not Top then?), but JEntry has the comment: /* May be accessed as superheader */... Greetings, Andres Freund -- Andres Freund http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Training Services -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] jsonb status
On 03/19/2014 09:28 AM, Andres Freund wrote: * The whole datastructure doesn't have any sensible highlevel documentation. Explain ... ? I'm planning on improving the docs through the beta period for this, so can you explain what kind of docs we're missing here? -- Josh Berkus PostgreSQL Experts Inc. http://pgexperts.com -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] jsonb status
On 2014-03-19 09:55:03 -0700, Josh Berkus wrote: On 03/19/2014 09:28 AM, Andres Freund wrote: * The whole datastructure doesn't have any sensible highlevel documentation. Explain ... ? I'm planning on improving the docs through the beta period for this, so can you explain what kind of docs we're missing here? Ah, sorry, that was easy to misunderstand. I was talking about the on-disk datastructure on the code level, not the user/SQL exposed part. I've added more verbose FIXMEs to the relevant places where I think such documentation should be added. Greetings, Andres Freund -- Andres Freund http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Training Services -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] jsonb status
On 03/19/2014 03:58 PM, Peter Geoghegan wrote: On Wed, Mar 19, 2014 at 9:28 AM, Andres Freund and...@2ndquadrant.com wrote: * Jsonb vs JsonbValue is just bad, the latter needs to be renamed, and there needs to be a very clear explanation about why two forms exist and what each is used for. Agreed. We should probably emphasize the distinction. Perhaps Oleg and Teodor might like to chime in on this. * The whole datastructure doesn't have any sensible highlevel documentation. And this. cheers andrew -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] jsonb status
On Wed, Mar 19, 2014 at 9:28 AM, Andres Freund and...@2ndquadrant.com wrote: I've pushed one commit with minor fixes, and one with several FIXMEs to http://git.postgresql.org/gitweb/?p=users/andresfreund/postgres.git;a=shortlog;h=refs/heads/jsonb_and_hstore Cool. * Jsonb vs JsonbValue is just bad, the latter needs to be renamed, and there needs to be a very clear explanation about why two forms exist and what each is used for. Agreed. We should probably emphasize the distinction. * The whole datastructure doesn't have any sensible highlevel documentation. I should probably go make some ascii art, per your comment. * I don't really like the JsonbSuperHeader name/abstraction. What does the Super mean? The only interpetation I see is that it's the toplevel header (why not Top then?), but JEntry has the comment: /* May be accessed as superheader */... What it means is that you can pass a pointer to just past the varlena Jsonb pointer in some contexts (a VARDATA() pointer), and a pointer into a jbvBinary buffer in others. The only reason that functions like findJsonbValueFromSuperHeader() take a super header rather than a Jsonb pointer is that if they did take a Jsonb*, you'd be making a representation that the varlena header wasn't garbage, which could not be ensured in all circumstances, as when the would-be varlena bytes are something else entirely, or perhaps are even at an address that is undefined to access. Sites with jbvBinary pointers (e.g. the iterator code) would have to worry about faking varlena. That comment is misleading, and the general idea does need to be explained better. What you propose for jsonb_typeof() looks incorrect to me. Both of those JsonbIteratorNext() calls are required, to get past the container pseudo array to get to the underlying single scalar element. -- Peter Geoghegan -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] jsonb status
On Wed, Mar 19, 2014 at 9:28 AM, Andres Freund and...@2ndquadrant.com wrote: * Jsonb vs JsonbValue is just bad, the latter needs to be renamed, and there needs to be a very clear explanation about why two forms exist and what each is used for. I've pushed some comments to Github that further clarify the distinction: https://github.com/feodor/postgres/commit/5835de0b55bdfc02ee59e2affb6ce25995587dc4 I did not change the name of JsonbValue, because I sincerely could not think of a better one. The distinction is subtle. -- Peter Geoghegan -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] jsonb status
On 03/19/2014 06:57 PM, Peter Geoghegan wrote: On Wed, Mar 19, 2014 at 9:28 AM, Andres Freund and...@2ndquadrant.com wrote: * Jsonb vs JsonbValue is just bad, the latter needs to be renamed, and there needs to be a very clear explanation about why two forms exist and what each is used for. I've pushed some comments to Github that further clarify the distinction: https://github.com/feodor/postgres/commit/5835de0b55bdfc02ee59e2affb6ce25995587dc4 I did not change the name of JsonbValue, because I sincerely could not think of a better one. The distinction is subtle. I didn't like the _state-state stuff either, but I think you changed the wrong name - it's the field name in the struct that needs changing. What you've done is inconsistent with the common idiom in jsonfuncs.c. cheers andrew -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] jsonb status
On Wed, Mar 19, 2014 at 5:10 PM, Andrew Dunstan and...@dunslane.net wrote: I didn't like the _state-state stuff either, but I think you changed the wrong name - it's the field name in the struct that needs changing. What you've done is inconsistent with the common idiom in jsonfuncs.c. Okay. I've changed the variable name back, while changing the struct field as suggested. -- Peter Geoghegan -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] jsonb status
On 03/16/2014 04:10 AM, Peter Geoghegan wrote: On Thu, Mar 13, 2014 at 2:00 PM, Andrew Dunstan and...@dunslane.net wrote: I'll be travelling a good bit of tomorrow (Friday), but I hope Peter has finished by the time I am back on deck late tomorrow and that I am able to commit this on Saturday. I asked Andrew to hold off on committing this today. It was agreed that we weren't quite ready, because there were one or two remaining bugs (since fixed), but also because I felt that it would be useful to first hear the opinions of more people before proceeding. I think that we're not that far from having something committed. Obviously I hope to get this into 9.4, and attach a lot of strategic importance to having the feature, which is why I made a large effort to help land it. Attached patch has a number of notable revisions. Throughout, it has been possible for anyone to follow our progress here: https://github.com/feodor/postgres/commits/jsonb_and_hstore * In general, the file jsonb_support.c (renamed to jsonb_utils.c) is vastly better commented, and has a much clearer structure. This was not something I did much with in the previous revision, and so it has been a definite focus of this one. * Hashing is refactored to not use CRC32 anymore. I felt this was a questionable method of hashing, both within jsonb_hash(), as well as the jsonb_hash_ops GIN operator class. * Dead code elimination. * I got around to fixing the memory leaks in B-Tree support function one. * Andrew added hstore_to_jsonb, hstore_to_jsonb_loose functions and a cast. One goal of this effort is to preserve a parallel set of facilities for the json and jsonb types, and that includes hstore-related features. * A fix from Alexander for the jsonb_hash_ops @operator issue I complained about during the last submission was merged. * There is no longer any GiST opclass. That just leaves B-Tree, hash, GIN (default) and GIN jsonb_hash_ops opclasses. My outstanding concerns are: * Have we got things right with GIN indexing, containment semantics, etc? See my remarks in the patch, by grepping contain within jsonb_util.c. Is the GIN text storage serialization format appropriate and correct? * General design concerns. By far the largest source of these is the file jsonb_util.c. * Is the on-disk format that we propose to tie Postgres to as good as it could be? I've been working through all the changes and fixes that Peter and others have made, and they look pretty good to me. There are a few mostly cosmetic changes I want to make, but nothing that would be worth holding up committing this for. I'm fairly keen to get this committed, get some buildfarm coverage and get more people playing with it and testing it. Like Peter, I would like to see more comments from people on the GIN support, especially. The one outstanding significant question of substance I have is this: given the commit 5 days ago of provision for triConsistent functions for GIN opclasses, should be be adding these to the two GIN opclasses we are providing, and what should they look like? Again, this isn't an issue that I think needs to hold up committing what we have now. Regarding Peter's last question, if we're not satisfied with the on-disk format proposed it would mean throwing the whole effort out and starting again. The only thing I have thought of as an alternative would be to store the structure and values separately rather than with values inline with the structure. That way you could have a hash of values more or less, which would eliminate redundancy of storage of things like object field names. But such a structure might well involve at least as much computational overhead as the current structure. And nobody's been saying all along hold on, we can do better than this. So I'm pretty inclined to go with what we have. cheers andrew -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] jsonb status
Alexander will take a look on TriConsistent function. On Mon, Mar 17, 2014 at 9:48 PM, Andrew Dunstan and...@dunslane.net wrote: On 03/16/2014 04:10 AM, Peter Geoghegan wrote: On Thu, Mar 13, 2014 at 2:00 PM, Andrew Dunstan and...@dunslane.net wrote: I'll be travelling a good bit of tomorrow (Friday), but I hope Peter has finished by the time I am back on deck late tomorrow and that I am able to commit this on Saturday. I asked Andrew to hold off on committing this today. It was agreed that we weren't quite ready, because there were one or two remaining bugs (since fixed), but also because I felt that it would be useful to first hear the opinions of more people before proceeding. I think that we're not that far from having something committed. Obviously I hope to get this into 9.4, and attach a lot of strategic importance to having the feature, which is why I made a large effort to help land it. Attached patch has a number of notable revisions. Throughout, it has been possible for anyone to follow our progress here: https://github.com/feodor/postgres/commits/jsonb_and_hstore * In general, the file jsonb_support.c (renamed to jsonb_utils.c) is vastly better commented, and has a much clearer structure. This was not something I did much with in the previous revision, and so it has been a definite focus of this one. * Hashing is refactored to not use CRC32 anymore. I felt this was a questionable method of hashing, both within jsonb_hash(), as well as the jsonb_hash_ops GIN operator class. * Dead code elimination. * I got around to fixing the memory leaks in B-Tree support function one. * Andrew added hstore_to_jsonb, hstore_to_jsonb_loose functions and a cast. One goal of this effort is to preserve a parallel set of facilities for the json and jsonb types, and that includes hstore-related features. * A fix from Alexander for the jsonb_hash_ops @operator issue I complained about during the last submission was merged. * There is no longer any GiST opclass. That just leaves B-Tree, hash, GIN (default) and GIN jsonb_hash_ops opclasses. My outstanding concerns are: * Have we got things right with GIN indexing, containment semantics, etc? See my remarks in the patch, by grepping contain within jsonb_util.c. Is the GIN text storage serialization format appropriate and correct? * General design concerns. By far the largest source of these is the file jsonb_util.c. * Is the on-disk format that we propose to tie Postgres to as good as it could be? I've been working through all the changes and fixes that Peter and others have made, and they look pretty good to me. There are a few mostly cosmetic changes I want to make, but nothing that would be worth holding up committing this for. I'm fairly keen to get this committed, get some buildfarm coverage and get more people playing with it and testing it. Like Peter, I would like to see more comments from people on the GIN support, especially. The one outstanding significant question of substance I have is this: given the commit 5 days ago of provision for triConsistent functions for GIN opclasses, should be be adding these to the two GIN opclasses we are providing, and what should they look like? Again, this isn't an issue that I think needs to hold up committing what we have now. Regarding Peter's last question, if we're not satisfied with the on-disk format proposed it would mean throwing the whole effort out and starting again. The only thing I have thought of as an alternative would be to store the structure and values separately rather than with values inline with the structure. That way you could have a hash of values more or less, which would eliminate redundancy of storage of things like object field names. But such a structure might well involve at least as much computational overhead as the current structure. And nobody's been saying all along hold on, we can do better than this. So I'm pretty inclined to go with what we have. cheers andrew -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] jsonb status
On Sun, March 16, 2014 09:10, Peter Geoghegan wrote: [ jsonb-11.patch.gz ] This doesn't quite compile: [...] patching file src/include/catalog/pg_amop.h patching file src/include/catalog/pg_amproc.h Hunk #3 FAILED at 358. 1 out of 3 hunks FAILED -- saving rejects to file src/include/catalog/pg_amproc.h.rej patching file src/include/catalog/pg_cast.h patching file src/include/catalog/pg_opclass.h [...] cat src/include/catalog/pg_amproc.h.rej *** src/include/catalog/pg_amproc.h --- src/include/catalog/pg_amproc.h *** DATA(insert ( 3659 3614 3614 2 3656 )) *** 358,364 DATA(insert ( 3659 3614 3614 3 3657 )); DATA(insert ( 3659 3614 3614 4 3658 )); DATA(insert ( 3659 3614 3614 5 2700 )); ! /* sp-gist */ DATA(insert ( 3474 3831 3831 1 3469 )); --- 360,373 DATA(insert ( 3659 3614 3614 3 3657 )); DATA(insert ( 3659 3614 3614 4 3658 )); DATA(insert ( 3659 3614 3614 5 2700 )); ! DATA(insert ( 4036 3802 3802 1 3480 )); ! DATA(insert ( 4036 3802 3802 2 3482 )); ! DATA(insert ( 4036 3802 3802 3 3483 )); ! DATA(insert ( 4036 3802 3802 4 3484 )); ! DATA(insert ( 4037 3802 3802 1 351 )); ! DATA(insert ( 4037 3802 3802 2 3485 )); ! DATA(insert ( 4037 3802 3802 3 3486 )); ! DATA(insert ( 4037 3802 3802 4 3487 )); /* sp-gist */ DATA(insert ( 3474 3831 3831 1 3469 )); -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] jsonb status - JsonbValue has no member named size
On Sun, March 16, 2014 09:50, Peter Geoghegan wrote: On Sun, Mar 16, 2014 at 1:28 AM, Erik Rijkers e...@xs4all.nl wrote: [ jsonb-11.patch.gz ] This doesn't quite compile: Sorry. I guess Andrew's earlier merging of master was insufficient. Attached revision fixes bitrot. Patch applies, but there is still something wrong; now during the actual compile (of contrib): (this is actually the same error I get when pulling straight from the git repo (since yesterday somewhere, I think, or maybe a bit earlier still)) (git repo being: on https://github.com/feodor/postgres.git, branch jsonb_and_hstore) -- [2014.03.16 12:29:27 jsonb/0] make contrib hstore_io.c: In function hstore_to_jsonb: hstore_io.c:1398:6: error: JsonbValue has no member named size key.size = sizeof(JEntry); ^ hstore_io.c:1402:6: error: JsonbValue has no member named size key.size += key.string.len; ^ hstore_io.c:1408:7: error: JsonbValue has no member named size val.size = sizeof(JEntry); ^ hstore_io.c:1413:7: error: JsonbValue has no member named size val.size = sizeof(JEntry); ^ hstore_io.c:1417:7: error: JsonbValue has no member named size val.size += val.string.len; ^ hstore_io.c: In function hstore_to_jsonb_loose: hstore_io.c:1450:6: error: JsonbValue has no member named size key.size = sizeof(JEntry); ^ hstore_io.c:1454:6: error: JsonbValue has no member named size key.size += key.string.len; ^ hstore_io.c:1458:6: error: JsonbValue has no member named size val.size = sizeof(JEntry); ^ hstore_io.c:1524:8: error: JsonbValue has no member named size val.size += VARSIZE_ANY(val.numeric) +sizeof(JEntry); ^ hstore_io.c:1528:8: error: JsonbValue has no member named size val.size = sizeof(JEntry); ^ hstore_io.c:1532:8: error: JsonbValue has no member named size val.size += val.string.len; ^ make[1]: *** [hstore_io.o] Error 1 make: *** [all-hstore-recurse] Error 2 -- make returned 2 - abort thanks, Erik Rijkers -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] jsonb status - 'JsonbValue' has no member named 'size'
On Sun, March 16, 2014 13:23, Andrew Dunstan wrote: [ jsonb-12.patch ] patch applies; compiles, and builds, but contrib installs with this error: make[1]: *** No rule to make target `hstore--1.2.sql', needed by `installdata'. Stop. make: *** [install-hstore-recurse] Error 2 After that an instance can be started but hstore is not available: create extension hstore; ERROR: could not stat file /home/aardvark/pg_stuff/pg_installations/pgsql.jsonb/share/extension/hstore--1.3.sql: No such file or directory -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
[HACKERS] jsonb status
Peter Geoghegan has been doing a lot of great cleanup of the jsonb code, after moving in the bits we wanted from nested hstore. You can see the current state of the code at https://github.com/feodor/postgres/tree/jsonb_and_hstore I've been working through some of his changes, I will probably make a couple of minor tweaks, but basically they look pretty good. I'll be travelling a good bit of tomorrow (Friday), but I hope Peter has finished by the time I am back on deck late tomorrow and that I am able to commit this on Saturday. cheers andrew -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
