1.2 String Manipulation

Last updated: September 5, 2024

See the Postgres docs Section 9.7 for a more in-depth description of the below.

LIKE, NOT LIKE¶

The LIKE/NOT LIKE keywords, which are often used in WHERE clauses to find matches between a string and a given pattern. Using this keyword, two specific characters are used to substitute for one or more characters in a pattern string:

• % matches zero or more characters

• _ matches a single character

If the pattern does not contain any percent sign (%) or underscore (_), then LIKE/NOT LIKE function as exact match.

Advanced string functionality¶

There are many string manipulation functions, such as substring, STRPOS (starting position of a substring), concatenation (||), and more. For example, the SUBSTRING function extracts the substring from a string with start and count parameters.

SELECT 'Hello' || 'World',
       STRPOS('Hello', 'el'),
       SUBSTRING('Hello', 2, 3);

In SQL the || operator means concatenation.

  ?column?  | strpos | substring 
------------+--------+-----------
 HelloWorld |      2 | ell

REGEXP_REPLACE¶

PostgreSQL can also handle (POSIX) regular expressions, e.g., with REGEXP_REPLACE.

General syntax:

REGEXP_REPLACE(source, pattern, replacement[, flags])

• source: source string, or column name, etc.

• pattern: POSIX regular expression

• replacement: replacement string when pattern matches

• flags: Optional parameter. Pass in g for to process all matches. See PostgresSQL docs Table 9.24 for the full list of options.

This follows standard POSIX regular expressions, and therefore you can use capture groups in both pattern (with parentheses to denote the subpattern to capture) and replacement (with \<number>, where number is the one-indexed index of the capture group.

To refresh on regular expressions, here are some examples:

Here are some example queries:

1. Extract first and last name of all people:

SELECT REGEXP_REPLACE(name, '(.*) (.*)', '\2')
         as lastname,
       REGEXP_REPLACE(name, '(.*) (.*)', '\1')
         as firstname
FROM people;

2. Count number of first names

SELECT REGEXP_REPLACE(name, '(.*) (.*)', '\1')
         as firstname,
       COUNT(*) as countname
FROM people
GROUP BY firstname
ORDER BY countname desc;

3. Compute length of longest first name

WITH firstnames AS (
    SELECT REGEXP_REPLACE(name, '(.*) (.*)', '\1')
             as firstname
    FROM people
)
SELECT MAX(LENGTH(firstname)), firstname
FROM firstnames;