Thread: pgadmin III initial comments and bug reports

Hello,

I've finished a first, raw translation of pgadmin3 into Danish and I now
want to actually use the program. I'm using a pgadmin3 based on code
checked out today.

First, some bugs found. In order not to confuse things, these errors
have been confirmed using the en_US locale:

- I tried to create a relatively simple table, but
  the resulting SQL failed because of misplaced and/or
  missing commas:
   CREATE TABLE public.misc_event
   (,
      misc_event_id serial
      shop_id int4
      time_create timestamptz NOT NULL
      remote_ip inet NOT NULL
      remote_id int4CONSTRAINT misc_event_pkey PRIMARY KEY
    (misc_event_id)
   ) WITHOUT OIDS;

 - I tried opening the "Query builder", but got a debug window:
     src/gtk/window.cpp(2852): assert "(m_widget != NULL)"
     failed: invalid window
     Do you want to stop the program?
     You can also choose [Cancel] to suppress further warnings.
   After hitting "no":
     src/gtk/window.cpp(2853): assert "(m_parent != NULL)" failed:
     wxWindowGTK::SetSize requires parent.
     Do you want to stop the program?
     You can also choose [Cancel] to suppress further warnings.
   (After this, I chose 'cancel') and the program crashed.
   The output of "bt" in gdb is attached.

- When adding a new column ("New Column" dialog), I'm not
  allowed to bring focus to neither the "Length" or the
  "Precision" fields, so the varchar length cannot be
  specified. This seems to be a general problem for all
  types with lengths which may be defined.


Some inconveniences:

 - I have to explicitly name primary keys which is annoying
   because I normally don't care what primay keys are called.

   It would be cool if pgadmin had pre-selected a name with
   a pattern as <table_name>_pkey as suggested primary key name.

   Like when this is run:

   create table test(test_id int not null primary key);

   then PostgreSQL auto-names the primary key "test_pkey".

 - When creating a new table, it's inconvenient that the
   primary key has to be defined separately. Example:
   In the CLI, I would do:
   create table test(test_id int not null primary key, ...);
   i.e. define the primary key in connection with the column
   which will act as primary key.
   It would be great of pgadmin had a selection box called
   "primary key" in the "Create column" dialog when a
   primary key hasn't already been defined for the table.

 - Indexes related to primary keys are not listed under
   the table's "Indexes" sub-tree.

 - Foreign keys have to be explicitly named which is a bit
   annoying compared to the CLI situation where PostgreSQL
   just chooses a name like $1, $2, ...

 - When creating a new column, I cannot simultaneously
   to look at another column in another table to be
   inpired by data type, etc: The "New Column" dialog
   is "modal".


Finally, a feature suggestion:

 - I haven't been able to find a sort of front-end to
   pg_dump. It would be nice if that existed in pgadmin.

--
Greetings from Troels Arvin, Copenhagen, Denmark

Troels Arvin wrote:

>- I tried to create a relatively simple table, but
>  the resulting SQL failed because of misplaced and/or
>  missing commas:
>
>
Fixed (invented last night when implementing table inheritance)

> - I tried opening the "Query builder", but got a debug window:
>     src/gtk/window.cpp(2852): assert "(m_widget != NULL)"
>
>
I finally commented out the query builder menu entry completely, not
only for release builds. This bug is really-really-really-well-known,
and documented in BUGS.txt, but nobody reads it...
In case anybody's going to debug that beast, please be invited to
uncomment the menu and get it running.

>- When adding a new column ("New Column" dialog), I'm not
>  allowed to bring focus to neither the "Length" or the
>  "Precision" fields, so the varchar length cannot be
>  specified. This seems to be a general problem for all
>  types with lengths which may be defined.
>
>
Typo disabling all length types, fixed.

>Some inconveniences:
>
> - I have to explicitly name primary keys which is annoying
>   because I normally don't care what primay keys are called.
>
Fixed (was implemented and got lost)

>   It would be cool if pgadmin had pre-selected a name with
>   a pattern as <table_name>_pkey as suggested primary key name.
>
Not necessary, if name is empty PostgreSQL will do this.

> - When creating a new table, it's inconvenient that the
>   primary key has to be defined separately. Example:
>   In the CLI, I would do:
>   create table test(test_id int not null primary key, ...);
>   i.e. define the primary key in connection with the column
>   which will act as primary key.
>   It would be great of pgadmin had a selection box called
>   "primary key" in the "Create column" dialog when a
>   primary key hasn't already been defined for the table.
>
We can put this on the wishlist. Not so trivial because we want to
inhibit duplicate pk definition.

> - Indexes related to primary keys are not listed under
>   the table's "Indexes" sub-tree.
>
The PK is a special unique index, which is listed under constraints. It
can't be handled like an ordinary index, so I'd consider it wrong if it
was listed there.

> - Foreign keys have to be explicitly named which is a bit
>   annoying compared to the CLI situation where PostgreSQL
>   just chooses a name like $1, $2, ...
>
No, they don't.

> - When creating a new column, I cannot simultaneously
>   to look at another column in another table to be
>   inpired by data type, etc: The "New Column" dialog
>   is "modal".
>
I hate modal dialogs, but need this in that situation to prevent problems.
I'll keep this in the back of my mind.

>Finally, a feature suggestion:
>
> - I haven't been able to find a sort of front-end to
>   pg_dump. It would be nice if that existed in pgadmin.
>
>
On my wishlist too. You'll have to wait for V2...

Regards,
Andreas