9
\$\begingroup\$

I want to perform a search in several columns of a table. I use the following query:

select *
from Tabela t
inner join "TabelaPai" tp on tp."ID" = t."RefTabelaPai" and tp."RefProject" = 'projectid'
where not t."Deleted" 
 and (t.Col1 ~ '.*__param1__.*' or t.Col2 ~ '.*__param1__.*' or t.Col3 ~ '.*__param1__.*'
   or t.Col4 ~ '.*__param1__.*' or t.Col5 ~ '.*__param1__.*' or t.Col6 ~ '.*__param1__.*' 
   or t.Col7 ~ '.*__param1__.*' or t.Col8 ~ '.*__param1__.*' or t.Col9 ~ '.*__param1__.*');

This will search for the keyword __param1__ in any of the columns and it's working fine.

But I don't like the way the query looks like. Any suggestion on how to refactor the query so it can look 'prettier' (without those ~ '.*__param1__.*' repetitions, for example)?

Edit: A little of context about the query:

What leads to this usage is that I can parameterize the data in the table. For example, I have a column in a table where scripts are saved. My application allows the users to parametrize the script using something like __param1__. If the user wants to rename the parameter I'll have to search for the usage of the parameter in every column that is parameterizable, and this is the query that finds where the parameter is used.

\$\endgroup\$
5
  • 2
    \$\begingroup\$ I think this is more of a database design issue rather than a query issue. How is your database used such that you need to search for the same value across several columns to find something? What's your database schema? \$\endgroup\$
    – user4549
    May 25, 2011 at 20:09
  • \$\begingroup\$ I think your kind of right.. \$\endgroup\$ May 26, 2011 at 8:18
  • \$\begingroup\$ Question updated with context. \$\endgroup\$ May 26, 2011 at 8:24
  • \$\begingroup\$ With such a task maybe you should keep variable name separately from its usage. Instead of storing script with variable name you can store it with some placeholder like {param1}. This will allow renaming parameters easily but you will have to replace placeholders in return. \$\endgroup\$
    – Snowbear
    May 26, 2011 at 8:59
  • \$\begingroup\$ The placehoders are the __ in the beggining and in the end. The user interacts with the name param1. \$\endgroup\$ May 26, 2011 at 12:29

3 Answers 3

5
\$\begingroup\$

I must admit, I don't really see what's wrong with the repetition — assuming it is what you're wanting to do (and your columns aren't actually named t.Colx!). If I came across this query in a project, I'd know pretty quickly what it's doing I think: searching a bunch of columns for a single supplied value (e.g. searching name, address, phone, etc. with a single search box, perhaps).

As for the matter of storing scripts and their parameters in a database: I'd probably go for a second key-value table, something like:

scripts { id, name, body }
script_parameters { id, script_id, name, value }

And you'd fetch the script and parameters and substitute the latter into the former in the app.

But then, I'm probably quite missing the point of what you're trying to do! :-)

\$\endgroup\$
1
  • \$\begingroup\$ Well, maybe I'm just being picky... When I see an example like this with so many repetitions I think immediately that it can be generalized. But as you said maybe there is nothing wrong with this query and just have to let it go :) \$\endgroup\$ May 27, 2011 at 8:15
10
\$\begingroup\$

Something "prettier"? And so many repetitions call for generalization?

SELECT *
FROM   tabela AS t
JOIN   "TabelaPai" tp ON tp."ID" = t."RefTabelaPai"
WHERE  NOT t."Deleted" 
AND    tp."RefProject" = 'projectid'
AND    t::text LIKE '%\_\_param1\_\_%';

Or cleaner:

...
AND    t.*::text LIKE '%\_\_param1\_\_%';

A nested column of the same name (t in this case) would take precedence. The more verbose syntax t.* makes it an unambiguous reference to the table row.

You can reference the composite type of any relation in the SELECT list. The manual:

Whenever you create a table, a composite type is also automatically created, with the same name as the table, to represent the table's row type.

You can cast the whole row (the composite type) to its text representation in one fell swoop, which is a very convenient syntactical shorthand. The resulting filter in my query is guaranteed to find every occurrence in the whole row.

The LIKE operator is generally faster than regular expression pattern matching (~). Regular expressions are far more powerful, but whenever LIKE can do the job, use it. Its syntax is simpler, too.

Underscores (_) have a special meaning for the LIKE operator, so you need to escape literal _. Default escape character is \.

Since Postgres 9.1, the setting standard_conforming_strings is on by default. Else, or with E'' syntax to declare Posix escape strings explicitly, escape \ like:

t::text LIKE E'%\\_\\_param1\\_\\_%'

If the separator in text representation (, by default) or double quotes (which can enclose strings) can be part of the search pattern, there can be false positives (across columns). So this is corner-case "dirty".

Asides:

It is cleaner to write AND tp."RefProject" = 'projectid' as WHERE clause, as it has no connection to tabela. Else you could put NOT t."Deleted" into the JOIN condition as well. Either way, same result.

Avoid CaMeL-case spelling of identifiers in Postgres like Tabela or RefTabelaPai. Without double-quotes, all identifiers are folded to lower case automatically.
My standing advice is to use legal. lower case identifiers exclusively in PostgreSQL and avoid avoid double-quoting and possible confusion.

\$\endgroup\$
1
  • \$\begingroup\$ Jesus Crist man! This is powerful ! \$\endgroup\$
    – Magno C
    Jul 8, 2015 at 16:58
1
\$\begingroup\$

http://www.postgresql.org/docs/current/interactive/textsearch-tables.html#TEXTSEARCH-TABLES-SEARCH

SELECT *
FROM "TAble"
WHERE to_tsvector("QUARTER1"||' '||"QUARTER2") @@ to_tsquery('abc');

selects any row that contains abc in column QUARTER1 or QUARTER2

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.