特定字符的Postgresql更新列功能

Question

I want to make a function which does update the column and replace specific values from column.

For example i have a column like

'187-2026-00130740 167'

The first number 187 and the last number 167 is random numbers. With function i take them with declared variables. Then going to update and remove them from column.

The _colid gets all id's from table as array. Then foreach id's in that array i get the first and last number. After that i updated these numbers with blank.

My function is basically looks like this.

CREATE OR REPLACE FUNCTION fix() RETURNS integer as $$
Declare _colid varchar[] := (SELECT '{{' || string_agg(concat_ws(',',id), '},{') || '}}' as a
FROM News n);
Declare _newsslice varchar[];
Declare _firstnumber varchar;
Declare _lastnumber varchar;
BEGIN
FOREACH _newsslice SLICE 1 IN ARRAY _colid 
LOOP
_firstnumber := (select regexp_replace(substring(split_part(token, '-', 1)from '[0-9]+'), '\s+$', '') AS x FROM news where id = _newsslice[1]::int and length(split_part(token, '-', 1)) < 4);

_lastnumber :=(select regexp_replace(substring(regexp_replace(token, '^. ', '')from '[0-9]+'), '\s+$', ' ') from news where id = _newsslice[1]::int and length(regexp_replace(substring(regexp_replace(token, '^. ', '')from '[0-9]+'), '\s+$', ' ')) < 4);

IF(_firstnumber IS NOT NULL) THEN
  update news set token = regexp_replace(regexp_replace(regexp_replace(trim(from token, ' ', ''),_colnumber,''),_firstnumber , '') where id = _newsslice[1]::int;
END IF;
END LOOP;
RETURN 1;
END; $$
LANGUAGE plpgsql;

This does update and delete the first number as i want. But did not delete the last number. What should i do, how do i approach to this?

Answer 1

This is a problem that is better solved using SQL, rather than a function.

update news
set
    token = substring(token from position('-' in token) + 1)
where
    position('-' in token) <= 4;

update news
set
    token = substring(token from 1 for (length(token) - 4 + position(' ' in substring(token from length(token) - 3)))) 
where
    position(' ' in substring(token from length(token) - 3)) > 0;

This solution assumes that:

• token may have a prefix or a suffix or both or neither.
• the prefix is up to 3 digits followed by a hyphen
• the suffix is a space followed by up to 3 digits

I get this from your description and the code. The actual requirement may be simpler, if for example all prefixes are exactly 3 digits.

If the prefix and suffix have to be checked to be numeric, that can be added, still using SQL.

The SQL solution is better because:

• It is shorter and simpler
• It avoids the looping boundary conditions that we all make mistakes with (me too)
• It will run faster

It would be interesting to know how many rows you are dealing with.

I would be inclined to create a new table with the tokens as required, do some comparisons between the old and the new and then drop the old and rename the new table. This may also be faster than doing lots of updates. If you do use the update approach make sure to vacuum afterwards.

My test data is:

create table news
(
    id        int,
    token     varchar(32)
);

insert into news (id, token) values (1, '187-2026-00130740 167');
insert into news (id, token) values (2, '18-2026-00130740 167');
insert into news (id, token) values (3, '1-2026-00130740 167');
insert into news (id, token) values (4, '-2026-00130740 167');
insert into news (id, token) values (5, '2026-00130740 167');
insert into news (id, token) values (6, '187-2026-00130740 16');
insert into news (id, token) values (7, '187-2026-00130740 1');
insert into news (id, token) values (8, '187-2026-00130740 ');
insert into news (id, token) values (9, '187-2026-00130740');

and the results are:

 id |     token      
----+----------------
  9 | 2026-00130740
  5 | 2026-00130740 
  1 | 2026-00130740 
  2 | 2026-00130740 
  3 | 2026-00130740 
  4 | 2026-00130740 
  6 | 2026-00130740 
  7 | 2026-00130740 
  8 | 2026-00130740 
(9 rows)