如何在 postgres 的文本列上添加唯一约束（忽略特殊字符）？

Question

How to add an unique constraint (ignoring special characters) on text column in Postgres?

CREATE TABLE my_table(
    SomeTextColumn citext
CONSTRAINT person_u_1 UNIQUE (SomeTextColumn)
);

In the above table, I'm trying to add an unique constraint that will look for uniqueness by ignoring special characters in the incoming data

For example:
1. HelloWorld --> Gets inserted successfully
2. Hello World --> Should fail with duplicate constraint
2. Hello%$^&*W^%orld --> Should fail with duplicate constraint

Answer 1

You can create a unique index that implements the check:

create unique index t_txt_unique on t(regexp_replace(txt, '\W', '', 'g'));

The regexp removes all non-word characters from the string, retaining only alphanumeric characters and the undescore _ . You can adjust the character class to as needed.

Demo on DB Fiddle :

create table t (id int, txt citext);
create unique index t_txt_unique on t(regexp_replace(txt, '\W', '', 'g'));

insert into t values(1, 'HelloWorld');
-- ok

insert into t values(1, 'Hello World');
-- ERROR:  duplicate key value violates unique constraint "t_txt_unique"
-- DETAIL:  Key (regexp_replace(txt, '\W'::text, ''::text, 'g'::text))=(HelloWorld) already exists.

insert into t values(1, 'Hello%$^&*W^%orld');
-- ERROR:  duplicate key value violates unique constraint "t_txt_unique"
-- DETAIL:  Key (regexp_replace(txt, '\W'::text, ''::text, 'g'::text))=(HelloWorld) already exists.

insert into t values(1, 'Hello Mars');
-- ok

Answer 2

The question is older but I think some additional notes might be useful...

• Uniqueness on TEXT is always problematic because text is case-sensitive (not only in PostgreSQL).
  You can get duplicates because "HelloWorld" is not the same as "HELLOWORLD". Because of that you might want to add the UPPER function if you create a UNIQUE INDEX on text fields of a database:
  CREATE UNIQUE INDEX t_txt_unique ON t(REGEXP_REPLACE(UPPER(txt), '\\W', '', 'g'));
• You might want to use the following REGEXP_REPLACE option to only keep the characters 0-9 and AZ (removing umlauts):
  CREATE UNIQUE INDEX t_txt_unique ON t(REGEXP_REPLACE(UPPER(txt), '[^0-9A-Z]', '\\1', 'g'));
• If you want to preserve umlauts as general character (Ü --> U, Ä --> A, etc.) you could throw in the UNACCENT function ( https://www.postgresql.org/docs/current/unaccent.html - it's an extension you need to add to PostgreSQL but I might also help you to search for something - keep in mind that only the superuser or other admin user might be able/allowed to add EXTENSIONs):
  CREATE EXTENSION IF NOT EXISTS unaccent WITH SCHEMA public;
CREATE UNIQUE INDEX t_txt_unique ON t(REGEXP_REPLACE(UPPER(UNACCENT(txt)), '[^0-9A-Z]', '\\1', 'g'));

SQL-Code

--
-- Check REGEXP_REPLACE to remove unwanted characters.
-- Test-String: '^°!"§$%&/()=?`+#äöüÜÖÄ',.-_:;@ABµC123abc123'
--
SELECT '^°!"§$%&/()=?`+#äöüÜÖÄ'',.-_:;@ABµC123abc123' AS original_text
      ,REGEXP_REPLACE( UPPER( '^°!"§$%&/()=?`+#äöüÜÖÄ'',.-_:;@ABµC123abc123' ), '\W', '', 'g') AS replace_word
      ,REGEXP_REPLACE( UPPER( '^°!"§$%&/()=?`+#äöüÜÖÄ'',.-_:;@ABµC123abc123' ), '[^0-9A-Z]', '\1', 'g') AS replace_upper_keep_num_a_to_z
      ,REGEXP_REPLACE( UPPER(UNACCENT( '^°!"§$%&/()=?`+#äöüÜÖÄ'',.-_:;@ABµC123abc123' )), '[^0-9A-Z]', '\1', 'g') AS replace_keep_num_a_to_z_umlaut
;

Result of the above SQL code as HTML table It was to complicated for the markdown table in the final result (preview was working).

 <table border="1"><tr BGCOLOR="#CCCCFF"><th>original_text</th><th>replace_word</th><th>replace_upper_keep_num_a_to_z</th><th>replace_keep_num_a_to_z_umlaut</th></tr> <tr><td>^°!&quot;§$%&amp;/()=?`+#äöüÜÖÄ',.-_:;@&lt;&gt;ABµC123abc123</td><td>ÄÖÜÜÖÄ_AB?C123ABC123</td><td>ABC123ABC123</td><td>AOUUOAABC123ABC123</td></tr> </table>

As you might have noticed the character 'µ' has been transformed to upper case 'M' in the second column using the '\\W' replacement which is a bit strange.