rownum anyone? cumulative sum in one query?

one of the nice (for some things) features of oracle is rownum() function.

i assume you know what it does – if you don't – i think a quick google will show it.

today i will show how to make rownum in postgresql. if you're not interested in rownum itself – please continue reading anyway, as some functions shown here have other uses as well 🙂

Continue reading rownum anyone? cumulative sum in one query?

“hacking” with postgresql

< french translation is now available >

very recently i got a task which involved finding out what happened (what sql queries were called) on remote machine.

the problem was that i didn't have any kind of shell access to server – i could connect only to postgresql. the good side was that i had a superuser-level access.

so the task was quite simple – how to find and read a file from psql. on remote machine. with no access to ssh, ftp, nor anything like this.

this is when i got idea that i'll write more about it. some might say that you shouldn't write this kind of information. my view is that i'm not disclosing any secrets – i will be using only basic (remote) psql with superuser access. all things that i will describe in here are in documentation – you just have to know where to look for them.

Continue reading “hacking" with postgresql

psql prompt issue

i tend to use more (than standard) informative prompts.

whether it's shell, or psql i set it up to match my needs.

for the longest time i used psql prompt:

\set PROMPT1 '(%n@%M:%>) %`date +%H:%M:%S` [%/] %[%012%]%x%# '

(this line should be in .psqlrc).

it gives some information:

(depesz@[local]:5810) 15:26:06 [depesz]
#

at a glance i know what user i am conencted as, to which machine, over which port, what database i'm connected to, and what's current time.

unfortunately this prompt had one issue – whenever i used history, and then something like <ctrl-a> (go to begining of line) something weird happened – cursor got distorted, and it looked like it was on second letter, while in fact it was on first (hard to explain, try it for yourself :).

it was painful, but since i couldn't do anything about it, i just learned to live with it.

now, due to some new environment setup, i decided that i've got enough of it. i tried to find any help in manual (i was looking for something like \[ and \] from bash prompting), but to no avail.

i did find out that the problem lies in %[012%]. i tried to give there literal <enter> (thus making \set statement multi-line), but it didn't work.

and then i simply gave there standard “\n". and it was it – now i have the prompt the way i like it, with no side effects on command line edition.

final prompt1, in case you're interested:

\set PROMPT1 '(%n@%M:%>) %`date +%H:%M:%S` [%/] \n%x%# '

Log Buffer #57: a Carnival of the Vanities for DBAs

A couple of days ago I got email from Dave Edwards of Pythian. Dave invited me to be guest editor of upcoming Log Buffer edition.

After some hesitation (just take a look at list of guys which were editors of previous editions, and you'll understand my shyness) I decided to give it a try. so, here it goes – 57th edition, written by depesz 🙂

First of all, we'll start with some posts about my favorite database – PostgreSQL.

Lewis Cunningham on his An Expert's Guide to Oracle Technology blog writes about new product for EnterpriseDB (commercial DB based on PostgreSQL) – GridSQL, which looks like Oracle's RAC. Check also his writeup on second blog: EnterpriseDB News.

What's interesting is that EnterpriseDB and their PR agency got lately quite controversial, which was mentioned by Joshua Drake on his blog (he also links to mailing list archives with discussions about the situation), and by Peter Eisentraut on his personal blog. this entry is very important as it goes into details about EnterpriseDB's “PostgreSQL distribution for Linux".

zillablog – blog by Robert Treat, has very interesting entry about conforming to standards, and using them as a excuse not to add new functionality.

And as a final note from our (PostgreSQL) area – Dave Page wrote new entry on his blog about setting up SSL access on Windows – apparently there were some issues with pre-built binaries of OpenSSL.

Now, let's enter the territory of “the other" opensource database – MySQL:

Biggest news of the week was change in policies and availability of MySQL products – community server and enterprise server. it all started with Kaj Arnö's post on his “MySQL AB VP Community Whereabouts in a Nutshell".

It generated a lot of followups, including Mike Kruckenberg on his blog, Greg's Postgres Stuff by Greg Sabino Mullane, Lukas Kahwe Smith's Poo-Tee-Weet and Peter Zaitsev on Mysql Performance Blog.

Of course Kaj Arnö had to respond, so there went another post on his blog.

Journal of Patrick Galbraith informs us about his new addition to mysqldump – options specifically designed to simplify dumping replicated databases.

Charlie Cahoon's MySQL – Summer of Code tells us how to use global variables in MySQL Proxy. To be honest I haven't heard about it (MySQL Proxy) before, but now, after reading his post I read more about it, and the software looks great. Something like pgpool (for those of you who know PostgreSQL), but with more functionality.

On the the possible functionalities of Mysql Proxy is using this (together with Charlie's code) to transform it into automatic regression tests generator. Details are shown on Giuseppe Maxia's data charmer – it looks really cool.

Johan Andersson and Jimmy Guerrero both tell us about web seminar (and white papers) about MySQL Cluster. Unfortunately if you're in America – it's too late, your webinar was on 8th, but you can still catch webinar directed at EMEA users on 15th.

While we're at learning – Jay Pipes presents materials from his workshop about tuning mysql queries – a must read for every MySQL DBA.

On his diary, Erik Wetterberg, wrote about generating XML content directly from MySQL. Check also comment to this post, as it contains link to valuable resource .

Sergey Zhuravlev on Xaprb wrote about new extension for MySQL replication – slave delay. Interesting idea with some obvious, and some not-so-obvious uses.

At the end of MySQL section, I couldn't help but write about two posts highlighting 2 interesting bugs in MySQL code.

First is Vadim Tkachenko on MySQL performance blog which writes about issues with query cache while dealing with column-based privileges.

The other post highlights what looks to me as quite serious bug. It's Parvesh Garg's post on Optim MySQL, and it's about InnoDB engine repeating (in some situations) auto_increment values.

Oracle is not the database I use, but 2 posts got my attention:

Mark Brunelli in Eye on Oracle: a searchoracle.com blog let us know that white papers for newest Oracle (11g) are available for download.

Second post made me sad that we don't have such a thing in PostgreSQL. It's on Coskan's Approach to Oracle by Coskan Gundogar and tells about permanent (guaranteed) restore points. Great feature.

At the end one more post – not technically database related, but it deals with the human side of databases and (DB) conferences: Mark Atwood's Journal entry about what to do when you're going to speak at a conference.

That would be all in this edition of Log Buffer. I hope you liked my choice of posts, and if not, don't worry, in a week there will be another Log Buffer, this time hosted by Jay Pipes

simple howto about restoring damaged template1

this is nothing new, but i blog it to have a place to point people to.

let's assume you accidentally loaded dump file in template1 database.

this is definitely not something one could have wanted (i mean i can see some uses for this but it's not really likely).

so, now you want to cleanup your template1.

how? that's easy.

connect with your superuser account (postgres) to some database other than template1.

and then issue these queries:

  1. # update pg_database set datistemplate = false where datname = ‘template1';
  2. # drop database template1;
  3. # create database template1 with template template0;
  4. # update pg_database set datistemplate = true where datname = ‘template1';

be sure that there is no connection to template1 at the moment you're dealing with it.

and that's all. nothing really complicated.

set returning functions in 8.3

just couple of days ago i read about a new, great addition to postgresql 8.3 – “return query" in pl/pgsql.

what does it do?

in set returning functions, when you wanted to return multiple rows from a given query you had to:

FOR record IN SELECT ..... LOOP
    RETURN NEXT record;
END LOOP;

now, you can simply:

RETURN QUERY SELECT ...;

what's more – since RETURN QUERY doesn't terminate function (just like return next) you can:

RETURN QUERY SELECT something;
RETURN QUERY SELECT something ELSE;

and then you'll get (more or less) “union all" of the queries.

additionally – return query is supposed to be faster then return next/loop.

so, let's test it.
Continue reading set returning functions in 8.3

indexable ” field like ‘%something'”

for the long time everybody knew that you can't use index on “LIKE" operations.

then came text_pattern_ops, so we could use indexes for prefix searches:

# \d depesz_test
                         Table "public.depesz_test"
 Column |  Type   |                        Modifiers
--------+---------+----------------------------------------------------------
 id     | integer | not null default nextval('depesz_test_id_seq'::regclass)
 email  | text    | not null
Indexes:
    "depesz_test_pkey" PRIMARY KEY, btree (id)
    "x" UNIQUE, btree (email text_pattern_ops)
# EXPLAIN analyze SELECT COUNT(*) FROM depesz_test WHERE email LIKE 'dep%';
                                                      QUERY PLAN
-----------------------------------------------------------------------------------------------------------------------
 Aggregate  (cost=96.71..96.72 ROWS=1 width=0) (actual TIME=0.983..0.985 ROWS=1 loops=1)
   ->  Bitmap Heap Scan ON depesz_test  (cost=4.68..96.65 ROWS=24 width=0) (actual TIME=0.184..0.641 ROWS=155 loops=1)
         FILTER: (email ~~ 'dep%'::text)
         ->  Bitmap INDEX Scan ON x  (cost=0.00..4.67 ROWS=24 width=0) (actual TIME=0.158..0.158 ROWS=155 loops=1)
               INDEX Cond: ((email ~>=~ 'dep'::text) AND (email ~<~ 'deq'::text))
 Total runtime: 1.067 ms
(6 ROWS)

but what if i'd like to search for ‘%something'? not prefix, but suffix. in my example – what can i do to use indexes when searching for people from given domain?

Continue reading indexable " field like ‘%something'"

update account set password = ‘new_password’; oops?!

how many times did you (or somebody in your environment) did something like this? update (or delete) without proper “where"?

it happened at least couple of times to me.

of course – using transactions solved the problem. but – on the other hand – if i can't trust myself to add proper where, how can i trust myself to add “begin"?

so, is there no hope?

are we doomed to always repeat the same mistake, and spend uncountable hours on recovering from damaged tables?

fortunatelly there is a help.
Continue reading update account set password = ‘new_password'; oops?!