Re: pgAdmin for Greenplum database?

Some notes on the current state of the patch:

>
> Hi Chuck,
>
> The patch looks pretty good to me - just a few issues/queries below.
> I've only eyeballed it and built it so far though - I'm in the process
> of downloading Greenplum so I can test more thoroughly, though it's a
> slow download at the moment (btw; you might want to have you're
> website tell the user a password is being emailed - it looks like
> nothing has happened when you first register).
>
> - There is a block of commented out code in dlgExtTable which looks
> like it was copied and pasted from dlgTable. It almost looks like you
> started to write a more comprehensive GUI but then decided to just use
> a textbox instead. What would be seen in the textbox? We would
> normally only provide a UI like that for 'free form' properties such
> as a function body, not for 'rigid form' properties such as those
> defining a table.

Yes, I'd like to make a nicer interface for creating external tables,
But it will take a bit of time to do, as they don't get created like regular tables.
So for now, the current code will have to do.

>
> - In pgPartitions.cpp, GetSQL should return the ALTER TABLE statement
> that might be used to create the partition, not an untranslated
> explanation of how the user should create the partition. The reverse
> engineered SQL we display on the main window is intended as a teaching
> tool as well as a convenience for those that like to use the SQL.

Yes, I plan to do this, but didn't have time to finish it for today, but will have a fix soon.
Note that the partitions are not normally created through any DDL of their own, but rather are created as part of the creation of the base table. So neither showing a CREATE TABLE nor an ALTER TABLE is exactly right, but the alter is better (as you suggested)
>
> - Can you please rename pgPartition/pgExtTable/pgResQueue to gpXXX? We
> use edbXXX for EnterpriseDB-specific objects and should follow that
> convention for Greenplum.

Done.

>
> - There are no dialogues for displaying/editing/creating partitions or
> resource queues. Is there a reason for this? Are those items a user
> will never create but must be able to browse? As a rule (to which
> there are just two closely related exceptions) we want to add complete
> support for object types from the outset.

Partitions are a funny case, as you can display them, but you can't alter them in any way that would change their compatibility with the base table (which rules out most ALTER statements).

Adding partitions to a table is of course possible, and I'd like to automate that too, but couldn't get it ready in time for today.

>
> - pgPartition.h (or gpPartition.h) is missing from precomp.h.
>

fixed

> - The test in isGpApp() is pretty flimsy. Can you call pg_dump --help
> and look for the --gp-syntax option in the help output?
>

I will fix this soon, but it shouldn't stop the patch from going in.

> - There's a change in embed-xrc.bat that probably shouldn't be in the
> patch!

Reverted.

>
> - On frmOptions, the labels 'Greenplum help path' and 'Greenplum bin
> path' don't fit, and are likely to be even longer in some
> translations. We need to use 'GP help path' and 'GP bin path' I guess.
>

Ok, changed to GP help path and GP bin path.

> - The text on the Display tab on frmOptions should be changed from
> 'ResQueues/resource queues' to 'Resource queues'. The dual naming is
> only used for roles where login roles == users and group roles ==
> groups in older versions of PG.

Done.

>
> - Please document the new options in docs/en_US/options-tab1.html. A
> screenshot would help as well, though I can do that if you don't have
> XP.

Done.
>
> Regards, Dave.
>
> --
> Dave Page
> EnterpriseDB UK: http://www.enterprisedb.com