Обсуждение: HISTORY file
I find the HISTORY file to be distressingly poor to peruse. Reasons: A large proportion of the items don't convey any useful information. Examples: | PLpgSQL fix for SELECT... FOR UPDATE (Tom) What did this fix? Does SELECT FOR UDPATE now work whereas it didn't use to? => "SELECT ... FOR UPDATE now works in PL/pgSQL" | Fix for PL/pgSQL PERFORM returning multiple rows (Tom) What did this fix? Can you return multiple rows now or does it merely give an error message that you cannot where it used to crash? | Fix for inherited CHECK constraints (Stephan Szabo) ditto | PL/pgSQL Allow IS and FOR in cursors (Bruce) If I didn't happen to know exactly what this meant, I wouldn't have a clue. | Allow NULL to appear at beginning/end based on ORDER BY (Tom) It doesn't "allow", it just "does". | Pltcl add spi_lastoid capability (bob@redivi.com) Capability = command, function, type, ...? | Allow column renaming in views ALTER VIEW foo RENAME COLUMN -- huh? | New option to output SET SESSION AUTHORIZATION commands (Peter E) Option to what to output where? | New postgresql.conf option to enable/disable "col = NULL" comparisons This is not correct. | Cachability fixes (Thomas, Tom) I don't think cachability as such was "fixed", or even "changed". The item probably related to some iscacheable pg_proc entries which were temporarily broken. The categories Bug Fixes, Enhancements, Types, Performance, Interfaces, Source Code could be split better, and they're not used very consistently. An example from each category that doesn't fit: Bug Fixes: Disallow access to pg_statistic for non-super user (Tom) This was not a bug, but a consequence of a change. Enhancements: Fix TCL COPY TO/FROM (ljb) If it is "fixed" then it was broken before. Types: New function bit_length() (Peter E) No comment. Performance: Dynahash portability improvements (Tom) Interfaces: Obviously, anything done in the interfaces is also either a bug fix or an enhancement. And what exactly constitutes an interface is not clear to me. Source code: Remove OID's from some system tables (Tom) Maybe this is an enhancement. Some changes are "must know", because they are incompatible, such as | Load pg_hba.conf only on startup and SIGHUP (Bruce) This should be made clear somewhere. Finally, | Remove configure --enable-pltcl-utf option There was never such an option in a previous release. -- Peter Eisentraut peter_e@gmx.net http://funkturm.homeip.net/~peter
> | Fix for inherited CHECK constraints (Stephan Szabo) > > ditto If this is what I think it is, I think the actual fix was the following (although I don't know what a particularly good wording is) ALTER TABLE ADD CONSTRAINT now properly adds check constraints to children of the specified table, which is consistant to the behavior of check constraints in inheritance trees created at create time.
Peter Eisentraut wrote:
>
> I find the HISTORY file to be distressingly poor to peruse. Reasons:
>
I noticed ODBC related items.
It seems plain to change as follows at least for me.
ODBC Remove query limit (Hiroshi) Remove query size limit
Remove text field size limit (Hiroshi) Fix for SQLPrimaryKeys() (Hiroshi) Fix for SQLPrimaryKeys in
multibytemode
Procedure calls (Hiroshi) Allow ODBC procedure calls FETCH first fix (Aidan Mountford)
??? maybe the following ? Improve boolean handing Updatable cursors (Hiroshi)
This isn't true. Please remove from the HISTORY list.
regards,
Hiroshi Inoue
> > > | Fix for inherited CHECK constraints (Stephan Szabo) > > > > ditto > > If this is what I think it is, I think the actual fix was the > following (although I don't know what a particularly good wording > is) > > ALTER TABLE ADD CONSTRAINT now properly adds check constraints > to children of the specified table, which is consistant to > the behavior of check constraints in inheritance trees created > at create time. Changed to: Fix for ALTER TABLE ADD CONSTRAINT ... CHECK for inherited children (Stephan Szabo) -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 853-3000+ If your life is a hard drive, | 830 Blythe Avenue + Christ can be your backup. | Drexel Hill, Pennsylvania19026
[ Sorry I am two days late in responding to this.] > I find the HISTORY file to be distressingly poor to peruse. Reasons: While I do my best to generate the HISTORY file, it is far from perfect. I need comments like this to help me improve it. Peter, glad you took the time to review the list. Let me address each one and modify the HISTORY file accordingly: > > A large proportion of the items don't convey any useful information. > Examples: > > | PLpgSQL fix for SELECT... FOR UPDATE (Tom) > > What did this fix? Does SELECT FOR UDPATE now work whereas it didn't use > to? => "SELECT ... FOR UPDATE now works in PL/pgSQL" Part of the problem here is that I have to guess from the commit message as to what was actually changed. The entries have to be: o conciseo understandable to noviceso combine entries fixing the same problem I could use some more information on this one. I should add that certain committers, particularly to interfaces, have commit messages that just say "Committed patch from Fred" and this does not help me generate a proper HISTORY file. Usually, copying something from the original message helps. > > | Fix for PL/pgSQL PERFORM returning multiple rows (Tom) > > What did this fix? Can you return multiple rows now or does it merely > give an error message that you cannot where it used to crash? Again, I don't know. > | Fix for inherited CHECK constraints (Stephan Szabo) > > ditto I don't know the details. Can you give them to me? > > | PL/pgSQL Allow IS and FOR in cursors (Bruce) > > If I didn't happen to know exactly what this meant, I wouldn't have a > clue. I can fix this one: PL/pgSQL Allow IS and FOR keywords in cursors, for compatibility (Bruce) > | Allow NULL to appear at beginning/end based on ORDER BY (Tom) > > It doesn't "allow", it just "does". Uh, yes, this is better: Make NULL appear at beginning/end based on ORDER BY (Tom) > | Pltcl add spi_lastoid capability (bob@redivi.com) > > Capability = command, function, type, ...? Got it, capability -> function: Pltcl add spi_lastoid function (bob@redivi.com) > > | Allow column renaming in views > > ALTER VIEW foo RENAME COLUMN -- huh? Actually, yes, it modifies the AS label of the column. Was that what you meant? create view x as select * from pg_class;alter table x rename column relname to jj;select jj from x; I suppose it didn't work before. > | New option to output SET SESSION AUTHORIZATION commands (Peter E) > > Option to what to output where? I now see the entire command was added in 7.2. I missed the earlier CVS commit: New SET SESSION AUTHORIZATION command (Peter E) > | New postgresql.conf option to enable/disable "col = NULL" comparisons > > This is not correct. Uh, it isn't? Can you give me some new text? > > | Cachability fixes (Thomas, Tom) > > I don't think cachability as such was "fixed", or even "changed". The > item probably related to some iscacheable pg_proc entries which were > temporarily broken. Do you have other wording? Seems there were was a cachability bug report and we "fixed" it in the catalogs. > The categories Bug Fixes, Enhancements, Types, Performance, Interfaces, > Source Code could be split better, and they're not used very consistently. > An example from each category that doesn't fit: > > Bug Fixes: Disallow access to pg_statistic for non-super user (Tom) > This was not a bug, but a consequence of a change. I considered it a bug. If there was a salary column, any user in 7.1 could see the max value in the column. Seemed like a security bug to me. > > Enhancements: Fix TCL COPY TO/FROM (ljb) > If it is "fixed" then it was broken before. Now: Add TCL COPY TO/FROM (ljb) Fixed. :-) > Types: New function bit_length() (Peter E) > No comment. Uh, I started to put some of the type-specific additions into Types. Is that OK? Particularly the multi-byte ones so they are all in one place. > > Performance: Dynahash portability improvements (Tom) Good point. Moved. > > Interfaces: Obviously, anything done in the interfaces is also either a > bug fix or an enhancement. And what exactly constitutes an interface is > not clear to me. Not clear to me either. I wanted to get jdbc and odbc into separate lists because they are so large. Seemed like a good idea. > Source code: Remove OID's from some system tables (Tom) > Maybe this is an enhancement. Uh, yes. > Some changes are "must know", because they are incompatible, such as > > | Load pg_hba.conf only on startup and SIGHUP (Bruce) > > This should be made clear somewhere. Added to Migration section: Also, pg_hba.conf only loads on SIGHUP now. > Finally, > > | Remove configure --enable-pltcl-utf option > > There was never such an option in a previous release. Oh, did that come in and out in 7.2? Removed. Let me know what else you see. Thanks. -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 853-3000+ If your life is a hard drive, | 830 Blythe Avenue + Christ can be your backup. | Drexel Hill, Pennsylvania19026 /usr/local/bin/mime: cannot create /dev/ttyp5: permission denied
All changed made. Thanks. --------------------------------------------------------------------------- > Peter Eisentraut wrote: > > > > I find the HISTORY file to be distressingly poor to peruse. Reasons: > > > > I noticed ODBC related items. > It seems plain to change as follows at least for me. > > ODBC > Remove query limit (Hiroshi) > Remove query size limit > > Remove text field size limit (Hiroshi) > Fix for SQLPrimaryKeys() (Hiroshi) > Fix for SQLPrimaryKeys in multibyte mode > > Procedure calls (Hiroshi) > Allow ODBC procedure calls > FETCH first fix (Aidan Mountford) > ??? maybe the following ? > Improve boolean handing > Updatable cursors (Hiroshi) > This isn't true. Please remove from the HISTORY list. > > regards, > Hiroshi Inoue > -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 853-3000+ If your life is a hard drive, | 830 Blythe Avenue + Christ can be your backup. | Drexel Hill, Pennsylvania19026