ads' corner - Code Snippets

Single line input for psql copy command

nospam@example.com (Andreas 'ads' Scherbaum) — Wed, 08 Oct 2008 22:26:05 +0200

Andreas 'ads' Scherbaum

I have the following problem:

A database server and another client machine. I have to execute several \copy commands in psql on the client machine to retrieve data and store the output in CSV files. The PostgreSQL COPY command is not an option because it stores the results on the database server, not the client machine. All \copy commands are executed from several files containing SQL commands, which are concated together by a makefile and feeded into psql. So far so good.

My SELECT commands used in the \copy are a bit more complicated and \copy requires to have the entire command on a single line. That can be done in two ways:

- Write the entire command in a single line. But that's hard to debug :-(

- Write the command as i like it and remove the linebreaks before feeding the command into psql.

Continue reading "Single line input for psql copy command"

role ... cannot be dropped because some objects depend on it

nospam@example.com (Andreas 'ads' Scherbaum) — Wed, 10 Sep 2008 20:11:07 +0200

Andreas 'ads' Scherbaum

I want to drop a ROLE (previously USER/GROUPs) which is still referenced by object(s) in my database. But PostgreSQL has objections:

ERROR: role "xyz" cannot be dropped because some objects depend on it

DETAIL: access to schema public

Continue reading "role ... cannot be dropped because some objects depend on it"

table size, database size

nospam@example.com (Andreas 'ads' Scherbaum) — Tue, 04 Dec 2007 11:07:00 +0100

Andreas 'ads' Scherbaum

I have seen this question more than once in the past: "How much storage space do i need for my table?"
PostgreSQL can give you this information.

The storage space (in bytes) for one specific table:

ads=# select pg_relation_size('size_test');
pg_relation_size
------------------
5668864
(1 row)

Same query, but with the result in a human-readable format:

ads=# select pg_size_pretty(pg_relation_size('size_test'));
pg_size_pretty
----------------
5536 kB(1 row)

Remember that this is only the size of the table, not included an index or additional stuff.
The size including any index can be found out with:

ads=# select pg_size_pretty(pg_total_relation_size('size_test'));
pg_size_pretty
----------------
7656 kB
(1 row)

The size of a complete database:

ads=# select pg_size_pretty(pg_database_size('ads'));
pg_size_pretty
----------------
11 MB
(1 row)

You can even find out, how much space a specific value needs:

ads=# select pg_column_size(5::smallint);
pg_column_size
----------------
2
(1 row)

ads=# select pg_column_size(5::int);
pg_column_size
----------------
4
(1 row)

ads=# select pg_column_size(5::bigint);
pg_column_size
----------------
8
(1 row)

ads=# select pg_column_size('This is a string'::varchar);
pg_column_size
----------------
20
(1 row)

ads=# select length('This is a string'::varchar);
length
--------
16
(1 row)

The 4 byte difference between the column size and the string length is the storage overhead needed for varchar columns.

The size for a tablespace can be found out with pg_tablespace_size(), but i don't have a machine with configured tablespace at hand to show an example.

stop psql on first error

nospam@example.com (Andreas 'ads' Scherbaum) — Sat, 20 Oct 2007 00:27:16 +0200

Andreas 'ads' Scherbaum

Because i always forget how to stop psql on the first error:

\set ON_ERROR_STOP

You can include this at the top of your script/file, psql will stop right after the first error and will not try to continue any other commands in the chain.

Avoid Unique Key violation

nospam@example.com (Andreas 'ads' Scherbaum) — Fri, 28 Jul 2006 11:18:51 +0200

Andreas 'ads' Scherbaum

If you have a table with an unique key field, want to insert new values but you don't know, if the value may already be in the table, you could use the following query:

CREATE TABLE uk_test (id SERIAL, col1 VARCHAR(20) NOT NULL UNIQUE, col2 VARCHAR(20) NOT NULL);

INSERT INTO uk_test (col1, col2) SELECT 'content 1', 'content 2' WHERE NOT EXISTS

(SELECT col1 FROM uk_test WHERE col1='content 1');

This does avoid unique key violation errors in a way, that no insert will happen if the value is already in the table.

I would like to thank depesz from #postgresql for the idea.

Reverse a text in PostgreSQL

nospam@example.com (Andreas 'ads' Scherbaum) — Mon, 24 Jul 2006 23:36:40 +0200

Andreas 'ads' Scherbaum

I'm always searching for a reverse function, so here is the link:

http://www.varlena.com/GeneralBits/53.php

code snippet is available in plperl and plpgsql.