Re: [HACKERS] text search vs schemas
On Fri, Aug 17, 2007 at 01:16:22AM -0400, Tom Lane wrote: That seems like it'd fix the problem for expression indexes on to_tsvector calls, but I don't see how it fixes the problem for triggers. We don't have any clear path for making trigger arguments be anything but a list of strings. I'm unsure how it works now, but it seems reasonable that when a regclass/regtype/regetc is passed to a trigger, pass it in OID form. This can be cast back safely inside the trigger itself. Seems a little hacky though... Having it as a type would also help with tracking dependancies. Have a nice day, -- Martijn van Oosterhout [EMAIL PROTECTED] http://svana.org/kleptog/ From each according to his ability. To each according to his ability to litigate. signature.asc Description: Digital signature
Re: [HACKERS] Re: cvsweb busted (was Re: [COMMITTERS] pgsql: Repair problems occurring when multiple RI updates have to be)
Marc G. Fournier wrote: --On Thursday, August 16, 2007 23:16:09 +0200 Magnus Hagander [EMAIL PROTECTED] wrote: But my question still stands - how much work to stop-gap fix it on the old one? rsync should be upgraded now ... Thanks! Hopefully that should fix the short-term problem. I'll try to take a look at the other one as soon as I can, hopefully this weekend - if you have the docs by then. //Magnus ---(end of broadcast)--- TIP 7: You can help support the PostgreSQL project by donating at http://www.postgresql.org/about/donate
Re: [HACKERS] text search vs schemas
Am Freitag, 17. August 2007 05:15 schrieb Tom Lane: Actually ... I'm suddenly not happy about the choice to put text search configurations etc. into schemas at all. We've been sitting here and assuming that to_tsvector('english', my_text_col) has a well defined meaning --- but as the patch stands, *it does not*. The interpretation of the config name could easily change depending on search_path. But that isn't different from any other part of the system. A proper fix would be a mechanism to alleviate the confusion in all places, not simply to remove features that cause such confusion in some places (but not all, thereby causing inconsistencies). It does not seem likely that a typical installation will have so many text search configs that subdividing them into schemas will really be useful. But schemas are not only used to organize objects because there are so many. Altering the search path to get at a different implementation without having to alter the names in every single place is also a useful application. -- Peter Eisentraut http://developer.postgresql.org/~petere/ ---(end of broadcast)--- TIP 4: Have you searched our list archives? http://archives.postgresql.org
Re: [HACKERS] GIT patch
Bruce Momjian wrote: These patches will be held for 8.4: o Grouped Index Tuples (GIT) o Bitmap scan changes o Stream bitmaps (API change for Group Index Tuples) o Maintaining cluster order on insert I believe Heikki is in agreement on this. Yes. -- Heikki Linnakangas EnterpriseDB http://www.enterprisedb.com ---(end of broadcast)--- TIP 5: don't forget to increase your free space map settings
Re: [HACKERS] tsearch patch and namespace pollution
Bruce Momjian wrote: I would be happy if all text search functions began with 'ts', 'ts_', or 'to_ts', and if we don't clean this up now, it is going to be harder in the future. +1 from me. \df is also much more useful then. I think users can expect some migration for text search in 8.3 as a benefit of getting into core and be dump-able. I guess so. Especially if you change some functions, they will have to change source code anyway. So you can as well cleanup all functions that don't fit into a sound naming schema. Best Regards Michael Paesold ---(end of broadcast)--- TIP 6: explain analyze is your friend
[HACKERS] pg_ctl configurable timeout
I'm having trouble with the hardcoded 60 second timeout in pg_ctl. pg_ctl sometimes just times out and there is no way to make it wait a little longer. I would like to add an option to be able to change that, say pg_ctl -w --timeout=120. Comments? -- Peter Eisentraut http://developer.postgresql.org/~petere/ ---(end of broadcast)--- TIP 5: don't forget to increase your free space map settings
Re: [HACKERS] text search vs schemas
Martijn van Oosterhout [EMAIL PROTECTED] writes: On Fri, Aug 17, 2007 at 01:16:22AM -0400, Tom Lane wrote: That seems like it'd fix the problem for expression indexes on to_tsvector calls, but I don't see how it fixes the problem for triggers. We don't have any clear path for making trigger arguments be anything but a list of strings. I'm unsure how it works now, but it seems reasonable that when a regclass/regtype/regetc is passed to a trigger, pass it in OID form. If you insist on a solution that involves attaching type information to trigger arguments, then we can forget about getting tsearch into 8.3. That's a nontrivial amount of new design and code that hasn't even been on the radar screen before. At the moment I feel our thoughts have to revolve not around adding complexity to tsearch, but taking stuff out. If we ship it with no schema support for TS objects in 8.3, we can always add that later, if there proves to be real demand for that (and I note that the contrib version has gotten along fine without it). But we cannot go in the other direction. regards, tom lane ---(end of broadcast)--- TIP 7: You can help support the PostgreSQL project by donating at http://www.postgresql.org/about/donate
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
On Thursday 16 August 2007 15:58, Bruce Momjian wrote: Josh Berkus wrote: All, First off, I'll assert that backup/restore is a serious issue and while the folks who want Tsearch in core now are dismissing it, we'll be fielding the complaints later. Any solution which involves setting a GUC at restore time *which could vary per table or even column* isn't acceptable. We used to do the \SET thing for table ownership with backup/restore, and you *know* how many restore failures that caused. Agreed. Let me summarize where we are now. I talked to Tom on the phone yesterday so we have come up with the following plan: o default_text_search_config stays, not super-user-only, not set in pg_dump output o tsearch functions that don't have a configuration name will be marked so they can't be specified in expression indexes o auto-casts and :: to tsearch data types will also not work in expression indexes (we already do this for timestamp without timezone) o GIN on an text column will not promote to tsvector o No rewrite magic for function calls without configuration names in WHERE clauses to use indexes that do specify configurations (risky) The current documentation explains all this: http://momjian.us/expire/textsearch/HTML/textsearch-tables.html So, we have disabled the ability to create expression indexes that are affected by default_text_search_config, and we have documented other possible problems. tsvector_update_trigger() has to be modified to take a configuration name (and frankly I am not excited about the filter_name capability either, but that is a separate issue). The only remaining problem I see is that the rest of the documentation relies heavily on default_text_search_config when in fact the most common usage with tables and indexes can't use it. tsquery can use the default easily, but I am betting that tsvector usually cannot. What exactly does default_text_search_config buy us? I think it is supposed to simplify things, but it sounds like it adds a bunch of corner cases, special siutations, if's and but's (and candies and nuts), that I fear will lead to more confusion for end users, and make development more difficult in the future as we forced to try and live with backwards compatability. -- Robert Treat Build A Brighter LAMP :: Linux Apache {middleware} PostgreSQL ---(end of broadcast)--- TIP 5: don't forget to increase your free space map settings
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
Robert Treat wrote: The only remaining problem I see is that the rest of the documentation relies heavily on default_text_search_config when in fact the most common usage with tables and indexes can't use it. tsquery can use the default easily, but I am betting that tsvector usually cannot. What exactly does default_text_search_config buy us? I think it is supposed to simplify things, but it sounds like it adds a bunch of corner cases, special siutations, if's and but's (and candies and nuts), that I fear will lead to more confusion for end users, and make development more difficult in the future as we forced to try and live with backwards compatability. Agreed. That was my conclusion long ago but few agreed so I gave up. In fairness the goal was for default_text_search_config to make text search easier for clusters that use a single configuration. If you are using triggers on a separate tsvector column, only the trigger author needs to deal with the configuration name (not queries), but expression indexes require the configuration name to always be used for the tsvector queries, while the tsquery can use the default_text_search_config value. Anyway, again, it is all special-casing this and that, as you said. And, if you are specifying the configuration name for the tsvector but not the tsquery you are more likely to have a configuration mismatch. (Of course you might want different configurations for tsvector and tsquery, but that is for experts.) -- Bruce Momjian [EMAIL PROTECTED] http://momjian.us EnterpriseDB http://www.enterprisedb.com + If your life is a hard drive, Christ can be your backup. + ---(end of broadcast)--- TIP 4: Have you searched our list archives? http://archives.postgresql.org
Re: [HACKERS] text search vs schemas
On Fri, Aug 17, 2007 at 10:42:09AM -0400, Tom Lane wrote: Martijn van Oosterhout [EMAIL PROTECTED] writes: I'm unsure how it works now, but it seems reasonable that when a regclass/regtype/regetc is passed to a trigger, pass it in OID form. If you insist on a solution that involves attaching type information to trigger arguments, then we can forget about getting tsearch into 8.3. That's a nontrivial amount of new design and code that hasn't even been on the radar screen before. Hmm, maybe I didn't explain clearly enough. I meant that if the argument is a regclass for example, to pass it in the TG_ARGV list as the OID in *string form*. That way trigger arguments stay a list of strings, yet the whole thing is schema safe because when trigger body casts the string back to a regclass, it gets exactly what was passed. Hope this makes more sense, -- Martijn van Oosterhout [EMAIL PROTECTED] http://svana.org/kleptog/ From each according to his ability. To each according to his ability to litigate. signature.asc Description: Digital signature
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
Oleg Bartunov [EMAIL PROTECTED] writes: On Thu, 16 Aug 2007, Josh Berkus wrote: First off, I'll assert that backup/restore is a serious issue and while the folks who want Tsearch in core now are dismissing it, we'll be fielding the complaints later. Any solution which involves setting a GUC at restore time *which could vary per table or even column* isn't acceptable. Josh, all my respects to you, but text searching is not about index at all. Text searching is about tsvector and tsquery data type What's your point? The problem is just as bad for an auto-update trigger that computes a stored tsvector column. If the trigger's behavior depends on the GUC settings of the person doing an insert, things will soon be a mess --- do you really want the tsvector contents to change after an update of an unrelated field? After awhile you won't have any idea what's really in the column, because you won't have any good way to know which rows' tsvectors were generated with which configurations. Even if that state of affairs is really what you want, reproducing it after a dump/reload will be tricky. I think that a regular schema-and-data dump would work, because pg_dump doesn't install triggers until after it's loaded the data ... but a data-only dump would *not* work, because the trigger would fire while loading rows. Basically I see no use for a setup in which the configuration used for a particular tsvector value is not fully determined by the table definition. Whether the value is in an index or in the table proper is irrelevant to this argument. regards, tom lane ---(end of broadcast)--- TIP 2: Don't 'kill -9' the postmaster
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
All - we have customers who very much want tsearch2 and will benefit from its inclusion in core. We are also struggling with the update trigger approach for various reasons. Is there a good alternative? Can we embed tsvector updates into the core code efficiently? - Luke Msg is shrt cuz m on ma treo -Original Message- From: Tom Lane [mailto:[EMAIL PROTECTED] Sent: Friday, August 17, 2007 11:28 AM Eastern Standard Time To: Oleg Bartunov Cc: Josh Berkus; pgsql-hackers@postgresql.org Subject:Re: [HACKERS] tsearch2 in PostgreSQL 8.3? Oleg Bartunov [EMAIL PROTECTED] writes: On Thu, 16 Aug 2007, Josh Berkus wrote: First off, I'll assert that backup/restore is a serious issue and while the folks who want Tsearch in core now are dismissing it, we'll be fielding the complaints later. Any solution which involves setting a GUC at restore time *which could vary per table or even column* isn't acceptable. Josh, all my respects to you, but text searching is not about index at all. Text searching is about tsvector and tsquery data type What's your point? The problem is just as bad for an auto-update trigger that computes a stored tsvector column. If the trigger's behavior depends on the GUC settings of the person doing an insert, things will soon be a mess --- do you really want the tsvector contents to change after an update of an unrelated field? After awhile you won't have any idea what's really in the column, because you won't have any good way to know which rows' tsvectors were generated with which configurations. Even if that state of affairs is really what you want, reproducing it after a dump/reload will be tricky. I think that a regular schema-and-data dump would work, because pg_dump doesn't install triggers until after it's loaded the data ... but a data-only dump would *not* work, because the trigger would fire while loading rows. Basically I see no use for a setup in which the configuration used for a particular tsvector value is not fully determined by the table definition. Whether the value is in an index or in the table proper is irrelevant to this argument. regards, tom lane ---(end of broadcast)--- TIP 2: Don't 'kill -9' the postmaster
Re: [HACKERS] text search vs schemas
Peter Eisentraut [EMAIL PROTECTED] writes: Am Freitag, 17. August 2007 05:15 schrieb Tom Lane: Actually ... I'm suddenly not happy about the choice to put text search configurations etc. into schemas at all. But that isn't different from any other part of the system. A proper fix would be a mechanism to alleviate the confusion in all places, not simply to remove features that cause such confusion in some places (but not all, thereby causing inconsistencies). Well, we are already inconsistent about this. PL languages and index access methods, for example, don't have schema-ified names. It does not seem likely that a typical installation will have so many text search configs that subdividing them into schemas will really be useful. But schemas are not only used to organize objects because there are so many. Altering the search path to get at a different implementation without having to alter the names in every single place is also a useful application. This is isomorphic to the argument about whether default_text_search_config is a good idea; indeed, I think that default_text_search_config pretty much solves this problem already, for the places where it's sane to have the configuration-in-use depend upon context. The problem with using schemas for TS configs is that we can't prevent the search result from changing in contexts where it mustn't change. At least, not short of requiring fully-qualified config names in those places, which doesn't sound like an advance in usability. regards, tom lane ---(end of broadcast)--- TIP 2: Don't 'kill -9' the postmaster
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
Robert Treat [EMAIL PROTECTED] writes: What exactly does default_text_search_config buy us? I think it is supposed to simplify things, but it sounds like it adds a bunch of corner cases, Well, the main thing we'd lose if we remove it is all trace of upward compatibility from the contrib version of tsearch. People are accustomed to using query functions that rely on a default configuration setting. Even though I want to prohibit use of a default in the definition of an index or auto-update trigger, I don't see a good reason to forbid it in queries. regards, tom lane ---(end of broadcast)--- TIP 3: Have you checked our extensive FAQ? http://www.postgresql.org/docs/faq
Re: [HACKERS] text search vs schemas
Martijn van Oosterhout [EMAIL PROTECTED] writes: On Fri, Aug 17, 2007 at 10:42:09AM -0400, Tom Lane wrote: If you insist on a solution that involves attaching type information to trigger arguments, then we can forget about getting tsearch into 8.3. Hmm, maybe I didn't explain clearly enough. I meant that if the argument is a regclass for example, to pass it in the TG_ARGV list as the OID in *string form*. Are you expecting the *user* to deal with that? If not, how is the system supposed to know which trigger arguments to do it to? What about dump and reload? regards, tom lane ---(end of broadcast)--- TIP 4: Have you searched our list archives? http://archives.postgresql.org
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
Luke Lonergan wrote: All - we have customers who very much want tsearch2 and will benefit from its inclusion in core. We are also struggling with the update trigger approach for various reasons. Is there a good alternative? Can we embed tsvector updates into the core code efficiently? No, doing it automatically adds too much complexity for little benefit. If you want more concrete suggestions, you will have to provide more information about the problems you are having. --- - Luke Msg is shrt cuz m on ma treo -Original Message- From: Tom Lane [mailto:[EMAIL PROTECTED] Sent: Friday, August 17, 2007 11:28 AM Eastern Standard Time To: Oleg Bartunov Cc: Josh Berkus; pgsql-hackers@postgresql.org Subject: Re: [HACKERS] tsearch2 in PostgreSQL 8.3? Oleg Bartunov [EMAIL PROTECTED] writes: On Thu, 16 Aug 2007, Josh Berkus wrote: First off, I'll assert that backup/restore is a serious issue and while the folks who want Tsearch in core now are dismissing it, we'll be fielding the complaints later. Any solution which involves setting a GUC at restore time *which could vary per table or even column* isn't acceptable. Josh, all my respects to you, but text searching is not about index at all. Text searching is about tsvector and tsquery data type What's your point? The problem is just as bad for an auto-update trigger that computes a stored tsvector column. If the trigger's behavior depends on the GUC settings of the person doing an insert, things will soon be a mess --- do you really want the tsvector contents to change after an update of an unrelated field? After awhile you won't have any idea what's really in the column, because you won't have any good way to know which rows' tsvectors were generated with which configurations. Even if that state of affairs is really what you want, reproducing it after a dump/reload will be tricky. I think that a regular schema-and-data dump would work, because pg_dump doesn't install triggers until after it's loaded the data ... but a data-only dump would *not* work, because the trigger would fire while loading rows. Basically I see no use for a setup in which the configuration used for a particular tsvector value is not fully determined by the table definition. Whether the value is in an index or in the table proper is irrelevant to this argument. regards, tom lane ---(end of broadcast)--- TIP 2: Don't 'kill -9' the postmaster -- Bruce Momjian [EMAIL PROTECTED] http://momjian.us EnterpriseDB http://www.enterprisedb.com + If your life is a hard drive, Christ can be your backup. + ---(end of broadcast)--- TIP 7: You can help support the PostgreSQL project by donating at http://www.postgresql.org/about/donate
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
Joshua D. Drake [EMAIL PROTECTED] writes: Tom Lane wrote: Well, the main thing we'd lose if we remove it is all trace of upward compatibility from the contrib version of tsearch. I don't think this is all that big of a deal. In fact I would expect it going from contrib to core and never had any illusion to the effect that I would be able to just upgrade from 8.2 (8.1) Tsearch2 to 8.3. I would hope that what we do with contrib/tsearch2 is rewrite it as a compatibility wrapper. This at least will provide an answer to anyone who complains that we renamed the functions. But if there are fundamental things missing in the core implementation, and we try to make the wrapper supply them, then we haven't really eliminated the problem ... just moved it over a little. regards, tom lane ---(end of broadcast)--- TIP 6: explain analyze is your friend
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
Tom Lane wrote: Josh Berkus [EMAIL PROTECTED] writes: Here's something not to forget in this whole business: the present TSearch2 implementation permits you to have a different tsvector configuration for each *row*, not just each column. That is, applications can be built with per-cell configs. Certainly. That's actually the easiest case to deal with, because you're going to put the tsvector config identity into another column of the table, and the trigger or index just references it there. It hasn't been part of the discussion because it's not a problem. I added an example of that in the documentation (second query): http://momjian.us/expire/textsearch/HTML/textsearch-tables.html#TEXTSEARCH-TABLES-INDEX -- Bruce Momjian [EMAIL PROTECTED] http://momjian.us EnterpriseDB http://www.enterprisedb.com + If your life is a hard drive, Christ can be your backup. + ---(end of broadcast)--- TIP 6: explain analyze is your friend
[HACKERS] tsearch still has external configuration files
A couple months ago I wrote: Lastly, I'm unhappy that the patch still keeps a lot of configuration information, such as stop word lists, in the filesystem rather than the database. It seems to me that the single easiest and most useful part of a configuration to change is the stop word list; but this setup guarantees that no one but a DBA can do that, and what's more that pg_dump won't record your changes. What's the point of having any non-superuser configuration capability at all, if stop words aren't part of what you can change? It appears that nothing has been done about this objection in the current patch. It is too late to redesign stop word handling for 8.3, but right now I have a more limited complaint: the patch allows unprivileged users to specify stopword files with absolute paths. This is a serious security breach, since it allows unprivileged users to read arbitrary files with the permissions of the postgres user. Now they maybe would have some difficulty determining the exact contents of such a file, but it would certainly be easy to test for the existence of particular words in it. What I think we should do about this is the same as we do for timezone abbreviation sets: the user-given stopword specification is just a name, which we insist can't contain dots or directory separators, and then we look up $SHAREDIR/dict_data/NAME.stop (or other suffixes for the other kinds of configuration files). This closes the security hole and also gives us a chance at an upward-compatible redesign later --- for instance, in a future release the name might refer to an entry in some other system catalog, rather than a file. BTW, I'm inclined to rename the installation directory to $SHAREDIR/tsearch_data/ ... any objections to that? regards, tom lane ---(end of broadcast)--- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
Folks, Here's something not to forget in this whole business: the present TSearch2 implementation permits you to have a different tsvector configuration for each *row*, not just each column. That is, applications can be built with per-cell configs. I know of at least one out there: Ubuntu's Rosetta. I'm sure there are others. Therefore there are two cases we're trying to solve: (1) The simple case: someone wants to build a database with text search entirely in one UTF8 language. All vectors are in that language, and so are all queries. The user wants the simplest syntax possible. (2) The Rosetta case: different configs are used for each cell and all searches have to be language-qualified. In both cases, the databases need to backup and restore cleanly. From this, I'd first of all say that I don't see the point of a Superuser default_tsvector_search_config. There are too many failure conditions with the default once you get away from the simplest case, so I don't see how setting it to Superuser-only protects anything. Might as well make it a userset and then it will be more useful. Unfortunately, the way I see it the only permanent solution for this is to alter the TSvector structure to include a config OID at the beginning of it. That doesn't sound like it's doable in time for 8.3, though; is there a way we could work around that until 8.4? And why does this sound exactly like the issues we've had with per-column encodings and the currency type? -- Josh Berkus PostgreSQL @ Sun San Francisco ---(end of broadcast)--- TIP 4: Have you searched our list archives? http://archives.postgresql.org
Re: [HACKERS] Re: cvsweb busted (was Re: [COMMITTERS] pgsql: Repair problems occurring when multiple RI updates have to be)
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 - --On Friday, August 17, 2007 08:40:11 +0200 Magnus Hagander [EMAIL PROTECTED] wrote: Marc G. Fournier wrote: --On Thursday, August 16, 2007 23:16:09 +0200 Magnus Hagander [EMAIL PROTECTED] wrote: But my question still stands - how much work to stop-gap fix it on the old one? rsync should be upgraded now ... Thanks! Hopefully that should fix the short-term problem. I'll try to take a look at the other one as soon as I can, hopefully this weekend - if you have the docs by then. If you need information, just ask for it ... - Marc G. Fournier Hub.Org Networking Services (http://www.hub.org) Email . [EMAIL PROTECTED] MSN . [EMAIL PROTECTED] Yahoo . yscrappy Skype: hub.orgICQ . 7615664 -BEGIN PGP SIGNATURE- Version: GnuPG v2.0.4 (FreeBSD) iD8DBQFGxdqt4QvfyHIvDvMRAps7AKCOeK/Nnl+QHP6s4dowwueVlJKCKgCgpdGV mmvsY+qa7gszdye6ftAc++4= =5WQB -END PGP SIGNATURE- ---(end of broadcast)--- TIP 1: if posting/reading through Usenet, please send an appropriate subscribe-nomail command to [EMAIL PROTECTED] so that your message can get through to the mailing list cleanly
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Tom Lane wrote: Robert Treat [EMAIL PROTECTED] writes: What exactly does default_text_search_config buy us? I think it is supposed to simplify things, but it sounds like it adds a bunch of corner cases, Well, the main thing we'd lose if we remove it is all trace of upward compatibility from the contrib version of tsearch. I don't think this is all that big of a deal. In fact I would expect it going from contrib to core and never had any illusion to the effect that I would be able to just upgrade from 8.2 (8.1) Tsearch2 to 8.3. Sincerely, Joshua D. Drake - -- === The PostgreSQL Company: Command Prompt, Inc. === Sales/Support: +1.503.667.4564 24x7/Emergency: +1.800.492.2240 PostgreSQL solutions since 1997 http://www.commandprompt.com/ UNIQUE NOT NULL Donate to the PostgreSQL Project: http://www.postgresql.org/about/donate PostgreSQL Replication: http://www.commandprompt.com/products/ -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.6 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFGxdYTATb/zqfZUUQRAo6gAJ9JDNGdTvYopOdw0Dp7rknffEZqewCaAkR9 d4EmQLv6iMpZ/iWR8Ksy1Ek= =aEft -END PGP SIGNATURE- ---(end of broadcast)--- TIP 6: explain analyze is your friend
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
Josh Berkus [EMAIL PROTECTED] writes: Here's something not to forget in this whole business: the present TSearch2 implementation permits you to have a different tsvector configuration for each *row*, not just each column. That is, applications can be built with per-cell configs. Certainly. That's actually the easiest case to deal with, because you're going to put the tsvector config identity into another column of the table, and the trigger or index just references it there. It hasn't been part of the discussion because it's not a problem. regards, tom lane ---(end of broadcast)--- TIP 4: Have you searched our list archives? http://archives.postgresql.org
[HACKERS] pgparam extension to the libpq api
Attached are some new functions that extend the libpq api to make calling the parameterized interfaces easier, especially when making binary calls. IMO, this fills one of the two big missing parts of the libpq api when making binary database calls, the other being client side handling of complex structures (arrays, etc). The code covers two major areas of functionality and isolated for separate inclusion: * PGparam (param.c) * get/set functions for the pgresult (result_ext.c) We are happy with both pieces but they can be adopted separately or not at all. The attached code is basically a cleaned up version of wrappers put in place in our own applications, plus a functional test. The major ideas were: * introduce a new opaque structure, PGparam, that handles some of the more difficult aspects of memory management associated with binary calls. * remove the requirement of client side code having to do byte swapping * make binary calls as efficient as possible, with a minimal amount of memory allocations * introduce, as much as possible, no additional portability issues or additional dependencies to the libpq api. Here are the interesting and/or possibly controversial pieces: * For portability purposes, we had the 64 bit integer put function take a pointer where the other putters take value types. We couldn't think of any other way to do it because there is not 64 bit portable integer type in libpq. * The synchronous execution functions (for example PQparamExec), takes a pointer to a result and return error status, which is _not_ how the other flavors of Exec operate, but is very convenient however. If you pass in NULL the result is discarded for you. We are stuck on this approach, but we like it. * The getters check the returned type oid to make sure it is sane. For this reason, we have to include catalog/pg_type.h and postgres.h to get to the OID defines (these are not exposed to the interface however). I don't see a reason why this is not ok. The 64 bit integer is handled as a pointer in the get/set functions because as far as we can tell there is no 64 bit integer type we can count on without introducing compatibility issues. We considered putting the PGparam struct into the PGconn structure. In this case, a PGconn pointer would be passed to the PQparamXXX functions instead of a PGparam, and would lazy allocate the structure and free it on PQfinish. We are curious for opinions on this. Writing credits to myself and Andrew Chernow. If this proposal is accepted, we will write all the documentation and make suitable changes necessary for inclusion, presumably for the 8.4 release. To compile the changes see the attached makefile. What we would really like is to use the backend input and output functions for data types, rather than reimplementing this within the client ... ie pqformat.c and similar files. For this reason, we did not re-implement get/put functions for the geometric types (we thought about it), etc. Merging the client and the server marshaling may require some abstraction of the server so formatting functions can be called from the client api. Hopefully this will open up the binary interfaces to more developers. For certain types of queries, binary calls can be a huge win in terms of efficiency. merlin makefile Description: Binary data #include stdlib.h #include string.h #include pg.h #include libpq-int.h /* Supports 250 columns worth of params. If more are needed, * memory is allocated ... very rare case. */ #define COLSTACKSIZE 4096 #define CHKPARAMPTR(p) do{ \ if(!(p)) \ { \ errno = EINVAL; \ strcpy((p)-errmsg, libpq_gettext(PGparam pointer is NULL)); \ return 0; \ } \ }while(0) #define PARAM_ARRAY_DECL \ char _stackbuffer[COLSTACKSIZE]; \ char *buf = _stackbuffer; \ char **vals = NULL; \ int *lens = NULL; \ int *fmts = NULL #define PARAM_ARRAY_ASSIGN do{ \ if(param) \ { \ int n = (int)((sizeof(void *) * param-vcnt) + \ ((sizeof(int) * 2) * param-vcnt)); \ if(n COLSTACKSIZE) \ { \ buf = (char *)malloc(n); \ if(!buf) \ { \ printfPQExpBuffer(conn-errorMessage, \ libpq_gettext(cannot allocate parameter column arrays\n)); \ return 0; \ } \ } \ vals = (char **)buf; \ lens = (int *)(buf + (sizeof(void *) * param-vcnt)); \ fmts = lens + param-vcnt; \ for(n=0; n param-vcnt; n++) \ { \ vals[n] = param-vals[n].data; \ lens[n] = param-vals[n].datal; \ fmts[n] = param-vals[n].format; \ } \ } \ }while(0) #define PARAM_ARRAY_FREE do{ \ if(buf != _stackbuffer) \ free(buf); \ }while(0) typedef struct { int ptrl; void *ptr; int datal; char *data; int format; } PGvalue; struct pg_param { int vcnt; int vmax; PGvalue *vals; int slabsize; char *slab; char errmsg[128]; }; PGparam *PQparamCreate(void) { return (PGparam *)calloc(1, sizeof(PGparam)); } void PQparamReset(PGparam *param) { if(param) param-vcnt = 0; } char *PQparamErrorMessage(PGparam *param) {
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
Josh Berkus wrote: Folks, Here's something not to forget in this whole business: the present TSearch2 implementation permits you to have a different tsvector configuration for each *row*, not just each column. That is, applications can be built with per-cell configs. I know of at least one out there: Ubuntu's Rosetta. I'm sure there are others. Therefore there are two cases we're trying to solve: (1) The simple case: someone wants to build a database with text search entirely in one UTF8 language. All vectors are in that language, and so are all queries. The user wants the simplest syntax possible. (2) The Rosetta case: different configs are used for each cell and all searches have to be language-qualified. In both cases, the databases need to backup and restore cleanly. From this, I'd first of all say that I don't see the point of a Superuser default_tsvector_search_config. There are too many failure conditions with the default once you get away from the simplest case, so I don't see how setting it to Superuser-only protects anything. Might as well make it a userset and then it will be more useful. Per my email yesterday, default_tsvector_search_config is _not_ super-user-only: o default_text_search_config stays, not super-user-only, not set in pg_dump output Unfortunately, the way I see it the only permanent solution for this is to alter the TSvector structure to include a config OID at the beginning of it. That doesn't sound like it's doable in time for 8.3, though; is there a way we could work around that until 8.4? Oh, so you want the config inside each tsvector value. Interesting idea. And why does this sound exactly like the issues we've had with per-column encodings and the currency type? Yes, this is a very similar issue except we are trying to allow multiple encodings. -- Bruce Momjian [EMAIL PROTECTED] http://momjian.us EnterpriseDB http://www.enterprisedb.com + If your life is a hard drive, Christ can be your backup. + ---(end of broadcast)--- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
Re: [HACKERS] tsearch still has external configuration files
Tom Lane wrote: A couple months ago I wrote: Lastly, I'm unhappy that the patch still keeps a lot of configuration information, such as stop word lists, in the filesystem rather than the database. It seems to me that the single easiest and most useful part of a configuration to change is the stop word list; but this setup guarantees that no one but a DBA can do that, and what's more that pg_dump won't record your changes. What's the point of having any non-superuser configuration capability at all, if stop words aren't part of what you can change? It appears that nothing has been done about this objection in the current patch. It is too late to redesign stop word handling for 8.3, Yes, I thought we agreed that for 8.3 we would use external files with UTF8 encoding. BTW, I'm inclined to rename the installation directory to $SHAREDIR/tsearch_data/ ... any objections to that? Seems clear to me. -- Bruce Momjian [EMAIL PROTECTED] http://momjian.us EnterpriseDB http://www.enterprisedb.com + If your life is a hard drive, Christ can be your backup. + ---(end of broadcast)--- TIP 7: You can help support the PostgreSQL project by donating at http://www.postgresql.org/about/donate
Re: [HACKERS] pgparam extension to the libpq api
On 8/17/07, Merlin Moncure [EMAIL PROTECTED] wrote: Attached are some new functions that extend the libpq api to make after sending the mail, we noticed some dead code that might be confusing...in PQparamClear there was some legacy code referring to 'slab' which has no effect...ignore. Also slab and slabsize members of PGparam are not supposed to be there. * The synchronous execution functions (for example PQparamExec), takes a pointer to a result and return error status, which is _not_ how the other flavors of Exec operate, but is very convenient however. If you pass in NULL the result is discarded for you. We are stuck on this approach, but we like it. Also, we are _not_ stuck in the **PGresult concept :-). (typo) merlin ---(end of broadcast)--- TIP 1: if posting/reading through Usenet, please send an appropriate subscribe-nomail command to [EMAIL PROTECTED] so that your message can get through to the mailing list cleanly
Re: [HACKERS] Re: cvsweb busted (was Re: [COMMITTERS] pgsql: Repair problems occurring when multiple RI updates have to be)
Marc G. Fournier wrote: I'll try to take a look at the other one as soon as I can, hopefully this weekend - if you have the docs by then. If you need information, just ask for it ... Magnus has repeatedly asked you to document it on PMT as we have been doing as a matter of course for everything for quite some time now. We're not doing that to be a pita, but to help us run the kind of professional infrastructure that the community has come to expect. That means everything is documented, systems are built in standard ways whereever possible, everything is monitored constantly, and backed up left right and center. So please help maintain that level of professionalism and document the new VM you've built so it can be properly maintained in the future. Regards, Dave. ---(end of broadcast)--- TIP 7: You can help support the PostgreSQL project by donating at http://www.postgresql.org/about/donate
[HACKERS] A small rant about coding style for backend functions
I don't want to pick on Teodor in particular, because I've seen other people do this too, but which of these functions do you find more readable? Datum to_tsquery_byname(PG_FUNCTION_ARGS) { PG_RETURN_DATUM(DirectFunctionCall2( to_tsquery_byid, ObjectIdGetDatum(name2cfgId((text *) PG_GETARG_POINTER(0), false)), PG_GETARG_DATUM(1) )); } Datum to_tsquery_byname(PG_FUNCTION_ARGS) { text *cfgname = PG_GETARG_TEXT_P(0); text *txt = PG_GETARG_TEXT_P(1); Oid cfgId; cfgId = name2cfgId(cfgname, false); PG_RETURN_DATUM(DirectFunctionCall2(to_tsquery_byid, ObjectIdGetDatum(cfgId), PointerGetDatum(txt))); } The main drawback to the V1-call-convention function call mechanism, compared to ordinary C functions, is that you can't instantly see what the function arguments are supposed to be. I think that good coding style demands ameliorating this by declaring and extracting all the arguments at the top of the function. The above example is bad enough, but when you have to dig through a function of many lines looking for GETARG calls in order to know what arguments it expects, it's seriously annoying and unreadable. And another thing: use the correct extraction macro for the argument's type, rather than making something up on the fly. Quite aside from helping the reader see what the function expects, the first example above is actually *wrong*, as it will crash on toasted input. OK, I'm done venting ... back to patch-fixing. regards, tom lane ---(end of broadcast)--- TIP 1: if posting/reading through Usenet, please send an appropriate subscribe-nomail command to [EMAIL PROTECTED] so that your message can get through to the mailing list cleanly
Re: [HACKERS] Re: cvsweb busted (was Re: [COMMITTERS] pgsql: Repair problems occurring when multiple RI updates have to be)
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Dave Page wrote: Marc G. Fournier wrote: I'll try to take a look at the other one as soon as I can, hopefully this weekend - if you have the docs by then. If you need information, just ask for it ... Magnus has repeatedly asked you to document it on PMT as we have been doing as a matter of course for everything for quite some time now. +1, even I am following up with the team standard of trying to document on PMT. We're not doing that to be a pita, but to help us run the kind of professional infrastructure that the community has come to expect. That means everything is documented, systems are built in standard ways whereever possible, everything is monitored constantly, and backed up left right and center. +1 So please help maintain that level of professionalism and document the new VM you've built so it can be properly maintained in the future. As our infrastructure continues to grow this is going to be vital. We keep getting bigger, and the only way to track this stuff appropriately is through documentation. Sincerely, Joshua D. Drake Regards, Dave. ---(end of broadcast)--- TIP 7: You can help support the PostgreSQL project by donating at http://www.postgresql.org/about/donate - -- === The PostgreSQL Company: Command Prompt, Inc. === Sales/Support: +1.503.667.4564 24x7/Emergency: +1.800.492.2240 PostgreSQL solutions since 1997 http://www.commandprompt.com/ UNIQUE NOT NULL Donate to the PostgreSQL Project: http://www.postgresql.org/about/donate PostgreSQL Replication: http://www.commandprompt.com/products/ -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.6 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFGxf4mATb/zqfZUUQRAsgQAJ4i/4j9a94WFHK2i1Xe/mA1yWi4gQCffQAI AsOsgbdFuqvYjLpFZRpby34= =FErs -END PGP SIGNATURE- ---(end of broadcast)--- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
Re: [HACKERS] Problem with locks
Gregory Stark wrote: I switched the code over to the sysv_sema style api. It's gotten a bit grotty and I would clean it up if it weren't a temporary test program. If we find a real problem perhaps I should add a test case like this to the smoke test in ipc_test.c so people can check their OS. So did you discover anything? I ran your test program and it worked successfully for several different configurations. Not enough times maybe, though. -- Alvaro Herrerahttp://www.CommandPrompt.com/ The PostgreSQL Company - Command Prompt, Inc. ---(end of broadcast)--- TIP 7: You can help support the PostgreSQL project by donating at http://www.postgresql.org/about/donate
Re: [HACKERS] More logging for autovacuum
Gregory Stark wrote: I'm having trouble following what's going on with autovacuum and I'm finding the existing logging insufficient. In particular that it's only logging vacuum runs *after* the vacuum finishes makes it hard to see what vacuums are running at any given time. Also, I want to see what is making autovacuum decide to forgo vacuuming a table and the log with that information is at DEBUG2. So did this idea go anywhere? -- Alvaro Herrera Developer, http://www.PostgreSQL.org/ Officer Krupke, what are we to do? Gee, officer Krupke, Krup you! (West Side Story, Gee, Officer Krupke) ---(end of broadcast)--- TIP 5: don't forget to increase your free space map settings
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
Joshua D. Drake wrote: Tom Lane wrote: Robert Treat [EMAIL PROTECTED] writes: What exactly does default_text_search_config buy us? I think it is supposed to simplify things, but it sounds like it adds a bunch of corner cases, Well, the main thing we'd lose if we remove it is all trace of upward compatibility from the contrib version of tsearch. I don't think this is all that big of a deal. In fact I would expect it going from contrib to core and never had any illusion to the effect that I would be able to just upgrade from 8.2 (8.1) Tsearch2 to 8.3. FWIW, I also would _not_ have expected compatibility between contrib and core. In fact, I would have expected contrib tsearch to be a place where experimental APIs existed and that the single biggest difference between contrib vs core was that the core APIs removed any cruft that might have been in contrib. If default_text_search_config makes things more confusing or more fragile, I'd rather see it gone than kept around for backward-compatibility-to-pre-core reasons. ---(end of broadcast)--- TIP 2: Don't 'kill -9' the postmaster
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
Bruce, Oh, so you want the config inside each tsvector value. Interesting idea. Yeah, hasn't anyone suggested this before? It seems like the obvious solution. A TSvector constructed with en_US is NOT the same as a vector constructed with fr_FR and it's silly to pretend that they are comparable. Sticking the config name at the beginning of the field would allow for the use of single-parameter functions, and default_config would only be used for SELECT queries. Backup/restore issues should go away completely ... EXCEPT this would introduce issues if the config is changed or deleted after being used. However, I'd imagine that we have those anyway -- certainly we would at restore time. -- --Josh Josh Berkus PostgreSQL @ Sun San Francisco ---(end of broadcast)--- TIP 1: if posting/reading through Usenet, please send an appropriate subscribe-nomail command to [EMAIL PROTECTED] so that your message can get through to the mailing list cleanly
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
Tom, It might be an obvious solution, but to some other problem than the one we have. The problem we are trying to address is how to know which config to use to construct a *new* tsvector. Oh, right. Back to the circular arguments then ... -- --Josh Josh Berkus PostgreSQL @ Sun San Francisco ---(end of broadcast)--- TIP 6: explain analyze is your friend
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
Josh Berkus [EMAIL PROTECTED] writes: Oh, so you want the config inside each tsvector value. Interesting idea. Yeah, hasn't anyone suggested this before? It seems like the obvious solution. It might be an obvious solution, but to some other problem than the one we have. The problem we are trying to address is how to know which config to use to construct a *new* tsvector. A TSvector constructed with en_US is NOT the same as a vector constructed with fr_FR and it's silly to pretend that they are comparable. Um, actually I think Oleg and Teodor believe that they *are* comparable. If we try to force them not to be then we'll break multi-language situations. regards, tom lane ---(end of broadcast)--- TIP 2: Don't 'kill -9' the postmaster
Re: [HACKERS] A small rant about coding style for backend functions
On 8/18/07, Tom Lane [EMAIL PROTECTED] wrote: The main drawback to the V1-call-convention function call mechanism, compared to ordinary C functions, is that you can't instantly see what the function arguments are supposed to be. I think that good coding style demands ameliorating this by declaring and extracting all the arguments at the top of the function. The above example is bad enough, but when you have to dig through a function of many lines looking for GETARG calls in order to know what arguments it expects, it's seriously annoying and unreadable. And another thing: use the correct extraction macro for the argument's type, rather than making something up on the fly. Quite aside from helping the reader see what the function expects, the first example above is actually *wrong*, as it will crash on toasted input. This is all useful guidance. My question is why it's not part of the developer documentation. Which brings me around to a minor rant of my own. All the developer FAQ has to say about coding style is that we use 4-space tabs for indentation, and that you should merge seamlessly into the surrounding code. That isn't much solace when the surrounding code is itself nigh unreadable or doesn't contain examples of what you are trying to do. For postgres hacking newbies (such as myself), the lack of any obvious published coding standards for the project is daunting, and is bound to lead to those developers filling in the blanks with their own coding style biases. Which means the patch reviewers need to spend time pointing out the flaws, and the submitter needs to spend time adjusting, testing and resubmitting ... it's all quite avoidable. I humbly suggest that if the sort of valuable information posted by Tom here was documented instead of ranted to the mailing list, maybe you guys wouldn't have to do so much ranting =) Cheers BJ ---(end of broadcast)--- TIP 5: don't forget to increase your free space map settings
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Fri, Aug 17, 2007 at 04:06:15PM -0700, Josh Berkus wrote: Bruce, Oh, so you want the config inside each tsvector value. Interesting idea. Yeah, hasn't anyone suggested this before? It seems like the obvious solution. A TSvector constructed with en_US is NOT the same as a vector constructed with fr_FR and it's silly to pretend that they are comparable. Except that (as I understand Oleg) it even seems to make sense sometimes to compare a tsvectors constructed with different configs -- so it might be important not to prevent this use case eihter. Oleg? Otherwise your proposal makes the most sense... Regards - -- tomás -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.1 (GNU/Linux) iD8DBQFGxm+DBcgs9XrR2kYRAn7RAJ4u508XQB/W6fMTmTchizlsvKEkEwCfTtTK R0DMLqNil2VQolFBWE69ZU0= =Tvh/ -END PGP SIGNATURE- ---(end of broadcast)--- TIP 2: Don't 'kill -9' the postmaster
Re: [HACKERS] Problem with locks
Alvaro Herrera [EMAIL PROTECTED] writes: Gregory Stark wrote: I switched the code over to the sysv_sema style api. It's gotten a bit grotty and I would clean it up if it weren't a temporary test program. If we find a real problem perhaps I should add a test case like this to the smoke test in ipc_test.c so people can check their OS. So did you discover anything? I ran your test program and it worked successfully for several different configurations. Not enough times maybe, though. I haven't been able to find any kernel problem which would explain the timeouts. The test program seems to work fine on all the machines I've tested it on except one where it turned up seemingly unrelated (and far worse) problems. But looking over the old test results from other machines I can see occasional transaction response times which exactly match the deadlock_timeout even though there should be no deadlocks. Apparently this happens with older releases of Postgres too. So I am fairly stumped here. There's really no way I can see where we would have the deadlock signal handler firing, not doing anything, but causing a semaphore wait to return. I've updated the kernel and will be running more benchmarks with the updated kernel next week. But I don't expect the results to change. -- Gregory Stark EnterpriseDB http://www.enterprisedb.com ---(end of broadcast)--- TIP 5: don't forget to increase your free space map settings
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
On 8/18/07, [EMAIL PROTECTED] [EMAIL PROTECTED] wrote: -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Fri, Aug 17, 2007 at 04:06:15PM -0700, Josh Berkus wrote: Bruce, Oh, so you want the config inside each tsvector value. Interesting idea. Yeah, hasn't anyone suggested this before? It seems like the obvious solution. A TSvector constructed with en_US is NOT the same as a vector constructed with fr_FR and it's silly to pretend that they are comparable. Except that (as I understand Oleg) it even seems to make sense sometimes to compare a tsvectors constructed with different configs -- so it might be important not to prevent this use case eihter. Oleg? Configs are not simply about languages, they are also about stopword lists and stemmers and parsers, and there's no reason to think that one would be using only one configuration to create a single tsvector. Different fields from within one document may require different treatment. Take for instance title, with stopwords included, and body, with them removed. Those two initial tsvectors can then be concatenated together with different weights to provide a very rich, and simple (relatively speaking) search infrastructure. --miker ---(end of broadcast)--- TIP 2: Don't 'kill -9' the postmaster
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
On Sat, 18 Aug 2007, [EMAIL PROTECTED] wrote: -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Fri, Aug 17, 2007 at 04:06:15PM -0700, Josh Berkus wrote: Bruce, Oh, so you want the config inside each tsvector value. б═Interesting idea. Yeah, hasn't anyone suggested this before? It seems like the obvious solution. A TSvector constructed with en_US is NOT the same as a vector constructed with fr_FR and it's silly to pretend that they are comparable. Except that (as I understand Oleg) it even seems to make sense sometimes to compare a tsvectors constructed with different configs -- so it might be important not to prevent this use case eihter. Oleg? yes, for example, you have tsvectors obtained from different sources, which require different processing. Otherwise your proposal makes the most sense... Regards - -- tomц║s -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.1 (GNU/Linux) iD8DBQFGxm+DBcgs9XrR2kYRAn7RAJ4u508XQB/W6fMTmTchizlsvKEkEwCfTtTK R0DMLqNil2VQolFBWE69ZU0= =Tvh/ -END PGP SIGNATURE- Regards, Oleg _ Oleg Bartunov, Research Scientist, Head of AstroNet (www.astronet.ru), Sternberg Astronomical Institute, Moscow University, Russia Internet: [EMAIL PROTECTED], http://www.sai.msu.su/~megera/ phone: +007(495)939-16-83, +007(495)939-23-83 ---(end of broadcast)--- TIP 5: don't forget to increase your free space map settings
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
Tom and Bruce, what version of patch you're using ? Bruce complained about using OID in arguments of functions, but AFAIR, it was removed in 0.58 version of patch. I and Teodor are very busy and just can't follow all discussions, so we have to rely on people's wisdom. If we have so many problem with integration, that probably we could just integrate support of data types (tsquery, tsvector), index support for them and set of support functions like to_tsquery, to_tsvector and leave everything remaining in contrib/tsearch2 as an example of text search engine design. Then, after fixing design problem as well as some backend's issues we could come with much better conclusions. Oleg On Fri, 17 Aug 2007, Tom Lane wrote: Josh Berkus [EMAIL PROTECTED] writes: Oh, so you want the config inside each tsvector value. Interesting idea. Yeah, hasn't anyone suggested this before? It seems like the obvious solution. It might be an obvious solution, but to some other problem than the one we have. The problem we are trying to address is how to know which config to use to construct a *new* tsvector. A TSvector constructed with en_US is NOT the same as a vector constructed with fr_FR and it's silly to pretend that they are comparable. Um, actually I think Oleg and Teodor believe that they *are* comparable. If we try to force them not to be then we'll break multi-language situations. regards, tom lane ---(end of broadcast)--- TIP 2: Don't 'kill -9' the postmaster Regards, Oleg _ Oleg Bartunov, Research Scientist, Head of AstroNet (www.astronet.ru), Sternberg Astronomical Institute, Moscow University, Russia Internet: [EMAIL PROTECTED], http://www.sai.msu.su/~megera/ phone: +007(495)939-16-83, +007(495)939-23-83 ---(end of broadcast)--- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
Re: [HACKERS] tsearch2 in PostgreSQL 8.3?
On Sat, 18 Aug 2007, Mike Rylander wrote: On 8/18/07, [EMAIL PROTECTED] [EMAIL PROTECTED] wrote: -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Fri, Aug 17, 2007 at 04:06:15PM -0700, Josh Berkus wrote: Bruce, Oh, so you want the config inside each tsvector value. Interesting idea. Yeah, hasn't anyone suggested this before? It seems like the obvious solution. A TSvector constructed with en_US is NOT the same as a vector constructed with fr_FR and it's silly to pretend that they are comparable. Except that (as I understand Oleg) it even seems to make sense sometimes to compare a tsvectors constructed with different configs -- so it might be important not to prevent this use case eihter. Oleg? Configs are not simply about languages, they are also about stopword lists and stemmers and parsers, and there's no reason to think that one would be using only one configuration to create a single tsvector. Different fields from within one document may require different treatment. Take for instance title, with stopwords included, and body, with them removed. Those two initial tsvectors can then be concatenated together with different weights to provide a very rich, and simple (relatively speaking) search infrastructure. I can't say better, Mike ! Regards, Oleg _ Oleg Bartunov, Research Scientist, Head of AstroNet (www.astronet.ru), Sternberg Astronomical Institute, Moscow University, Russia Internet: [EMAIL PROTECTED], http://www.sai.msu.su/~megera/ phone: +007(495)939-16-83, +007(495)939-23-83 ---(end of broadcast)--- TIP 5: don't forget to increase your free space map settings