Обсуждение: RFC: Removal of Revision Control
I am in the process of making *major* changes to the pgAdmin code to support Schemas in the upcoming PostgreSQL 7.3 release. I would like to remove the Revision Control feature from pgAdmin 1.3 for the following reasons: 1) It is difficult to maintain. Removing the code will make Schema implementation significantly easier. 2) It is of limited use (for example, rolling back a table will lose its data, rolling back other objects like functions or views may break functions, views, triggers, indexes etc that are dependant on that object). 3) It isn't particuarly reliable. RC Status is often reported incorrectly. 4) I suspect few, if any users actually make use of the feature (including myself!). Does anyone have any objections or comments? Regards, Dave.
Thats fair. The ability to rollback most changes isn't overly useful. What is useful is the ability to replay commands on another database. I suggest a good replacement would be a simple 'log' mode which writes text based queries to a log file. PGAdmin should be able to play an arbitrary text file against a database connection. This would preserve most of the functionality and should be straight forward to implement. The only catch is it shouldn't have the select queries in it which the PGAdmin uses to determine structure. -- Rod ----- Original Message ----- From: "Dave Page" <dpage@vale-housing.co.uk> To: <pgadmin-hackers@postgresql.org>; <pgadmin-support@postgresql.org> Sent: Wednesday, May 01, 2002 5:09 PM Subject: [pgadmin-hackers] RFC: Removal of Revision Control > I am in the process of making *major* changes to the pgAdmin code to > support Schemas in the upcoming PostgreSQL 7.3 release. > > I would like to remove the Revision Control feature from pgAdmin 1.3 for > the following reasons: > > 1) It is difficult to maintain. Removing the code will make Schema > implementation significantly easier. > > 2) It is of limited use (for example, rolling back a table will lose its > data, rolling back other objects like functions or views may break > functions, views, triggers, indexes etc that are dependant on that > object). > > 3) It isn't particuarly reliable. RC Status is often reported > incorrectly. > > 4) I suspect few, if any users actually make use of the feature > (including myself!). > > Does anyone have any objections or comments? > > Regards, Dave. > > ---------------------------(end of broadcast)--------------------------- > TIP 2: you can get off all lists at once with the unregister command > (send "unregister YourEmailAddressHere" to majordomo@postgresql.org) >
> -----Original Message----- > From: Rod Taylor [mailto:rbt@zort.ca] > Sent: 02 May 2002 00:31 > To: Dave Page; pgadmin-hackers@postgresql.org; > pgadmin-support@postgresql.org > Subject: Re: [pgadmin-hackers] RFC: Removal of Revision Control > > > Thats fair. The ability to rollback most changes isn't > overly useful. What is useful is the ability to replay > commands on another database. > > I suggest a good replacement would be a simple 'log' mode > which writes text based queries to a log file. PGAdmin > should be able to play an arbitrary text file against a > database connection. This would preserve most of the > functionality and should be straight forward to > implement. The only catch is it shouldn't have the select queries in > it which the PGAdmin uses to determine structure. Umm, I already implemented that when you requested it some months ago :-). It was one of the first features to be added to the 1.3 dev version. All queries are now tagged with a type - User (ones you type in), Datagrid (generated by the datagrid when you edit data) & System (the ones you want to ignore) and you can select any combination to be logged to a text file. Regards, Dave.
Le Mercredi 1 Mai 2002 23:09, Dave Page a écrit : > Does anyone have any objections or comments? Dear all, There were discussions on hackers months ago about rollback using WAL (point to point recovery?) and object tracing (something like SET TRACE ON ...). I don't think it will be implemented in 7.3. By the way, do you know if CREATE OR REPLACE VIEW will be part of 7.3? Cheers, Jean-Michel
> -----Original Message----- > From: Jean-Michel POURE [mailto:jm.poure@freesurf.fr] > Sent: 02 May 2002 09:18 > To: Dave Page; pgadmin-hackers@postgresql.org; > pgadmin-support@postgresql.org > Subject: Re: [pgadmin-hackers] RFC: Removal of Revision Control > > > Le Mercredi 1 Mai 2002 23:09, Dave Page a écrit : > > Does anyone have any objections or comments? > > Dear all, > > There were discussions on hackers months ago about rollback > using WAL (point > to point recovery?) and object tracing (something like SET > TRACE ON ...). I > don't think it will be implemented in 7.3. What do you mean by object tracing? If you mean dependency tracking, I believe Rod's working on that now. > By the way, do you know if CREATE OR REPLACE VIEW will be part of 7.3? Gavin(?) said he'd try to do it, but at least we have pseudo modification now. Regards, Dave.
> -----Original Message----- > From: Jean-Michel POURE [mailto:jm.poure@freesurf.fr] > Sent: 02 May 2002 10:07 > To: Dave Page > Subject: Re: [pgadmin-hackers] RFC: Removal of Revision Control > > > Le Jeudi 2 Mai 2002 10:25, vous avez écrit : > > What do you mean by object tracing? If you mean dependency > tracking, I > > believe Rod's working on that now. > > I guess it is the ability to log all queries involving a > particular object, > like in MS SQL Server. For example : > > (I don't remember the syntax exactly:) > SET TRACE ON table foo; > will log all queries on table foo. > > I don't think anyone is working on such a feature, > but hackers discussed about it. Oh, I see. That sounds like a nice feature. Probably possible in pgAdmin, though I think schema support is going to keepme busy for a while. pgAdmin dies horribly in many places with the current 7.3 code :-( Regards, Dave.
> > I suggest a good replacement would be a simple 'log' mode > > which writes text based queries to a log file. PGAdmin > > should be able to play an arbitrary text file against a > > database connection. This would preserve most of the > > functionality and should be straight forward to > > implement. The only catch is it shouldn't have the select queries in > > it which the PGAdmin uses to determine structure. > > Umm, I already implemented that when you requested it some months ago > :-). It was one of the first features to be added to the 1.3 dev > version. All queries are now tagged with a type - User (ones you type > in), Datagrid (generated by the datagrid when you edit data) & System > (the ones you want to ignore) and you can select any combination to be > logged to a text file. Erm, sorry. I don't use the dev version ;) Thanks :)