Re: proposal - psql - show longest tables

On Mon, Jul 22, 2013 at 2:03 PM, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
> On Sun, Jul 21, 2013 at 12:47 AM, Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com> wrote:
>> Hello all
>>
>> I very often use a little bit adjusted psql system queries to
>> detection TOP N sized tables. I am thinking so it can be useful for
>> all users
>>
>> I propose a few new commands
>>
>> \dts [N|size] ... show N largest tables | show tables larger than size
>> ordered by size
>> \dis [N|size] ... show N largest indexes | show indexes larger than
>> size ordered by size
>> \dtst [N|size] ... show N largest total size | show tables where total
>> size is larger than size ordered by total size
>> \dtr [N] ... show N largest tables (ordered by rows)
>>
>> example:
>>
>> \dts 10 --- top 10 tables ordered by size
>> \dts 10MB -- tables larger than 10MB ordered by size
>>
>> Schema | Name | Type | Owner | Size
>> --------+-------------+-------+-------------+-------+-------------
>> public | eshop_users | table | eshop_owner | 16 kB
>> public | zamestnanci | table | eshop_owner | 16 kB
>>
>> What do you think about this proposal? Comments, notes?
>
> I think our \d commands are in inscrutable morass of indecipherable
> gobbledygook as it is, and this is only one more step down the road to
> complete insanity. :-(
>
> Rather than just continuing to add more imposible-to-remember syntax,
> we really need a better design here.

These type of administrative tasks should be implemented as stored
procedures or functions, not enhancements to psql. That way non-psql
clients can leverage them and you can integrate them to other queries.
Another advantage is that they can be implemented as extension.

SELECT * from top5();

Is a little more of a pain to type. But with my psql-fu I can cut
that down with \i if I'm so inclined.

merlin