[postgresql] PostgreSQL: How to make "case-insensitive" query

Is there any way to write case-insensitive queries in PostgreSQL, E.g. I want that following 3 queries return same result.

SELECT id FROM groups where name='administrator'

SELECT id FROM groups where name='ADMINISTRATOR'

SELECT id FROM groups where name='Administrator'

You can also read up on the ILIKE keyword. It can be quite useful at times, albeit it does not conform to the SQL standard. See here for more information: http://www.postgresql.org/docs/9.2/static/functions-matching.html

You can use ILIKE. i.e.

SELECT id FROM groups where name ILIKE 'administrator'

The most common approach is to either lowercase or uppercase the search string and the data. But there are two problems with that.

1. It works in English, but not in all languages. (Maybe not even in most languages.) Not every lowercase letter has a corresponding uppercase letter; not every uppercase letter has a corresponding lowercase letter.
2. Using functions like lower() and upper() will give you a sequential scan. It can't use indexes. On my test system, using lower() takes about 2000 times longer than a query that can use an index. (Test data has a little over 100k rows.)

There are at least three less frequently used solutions that might be more effective.

1. Use the citext module, which mostly mimics the behavior of a case-insensitive data type. Having loaded that module, you can create a case-insensitive index by CREATE INDEX ON groups (name::citext);. (But see below.)
2. Use a case-insensitive collation. This is set when you initialize a database. Using a case-insensitive collation means you can accept just about any format from client code, and you'll still return useful results. (It also means you can't do case-sensitive queries. Duh.)
3. Create a functional index. Create a lowercase index by using CREATE INDEX ON groups (LOWER(name));. Having done that, you can take advantage of the index with queries like SELECT id FROM groups WHERE LOWER(name) = LOWER('ADMINISTRATOR');, or SELECT id FROM groups WHERE LOWER(name) = 'administrator'; You have to remember to use LOWER(), though.

The citext module doesn't provide a true case-insensitive data type. Instead, it behaves as if each string were lowercased. That is, it behaves as if you had called lower() on each string, as in number 3 above. The advantage is that programmers don't have to remember to lowercase strings. But you need to read the sections "String Comparison Behavior" and "Limitations" in the docs before you decide to use citext.

You could also use POSIX regular expressions, like

SELECT id FROM groups where name ~* 'administrator'

SELECT 'asd' ~* 'AsD' returns t

using ILIKE instead of LIKE

SELECT id FROM groups WHERE name ILIKE 'Administrator'

Using ~* can improve greatly on performance, with functionality of INSTR.

SELECT id FROM groups WHERE name ~* 'adm'

return rows with name that contains OR equals to 'adm'.