Re: PL/pgSQL techniques better than bash for dynamic DO? - Mailing list pgsql-general
From | Pavel Stehule |
---|---|
Subject | Re: PL/pgSQL techniques better than bash for dynamic DO? |
Date | |
Msg-id | CAFj8pRAbQ6BgdnFod3_FX2+wjSg+s1f_VfD+WWEG7T5PfOgSEg@mail.gmail.com Whole thread Raw |
In response to | [MASSMAIL]PL/pgSQL techniques better than bash for dynamic DO? (Ron Johnson <ronljohnsonjr@gmail.com>) |
List | pgsql-general |
út 9. 4. 2024 v 18:33 odesílatel Ron Johnson <ronljohnsonjr@gmail.com> napsal:
PG 9.6.11, if relevant, migrating to PG 14 Real Soon Now.I must purge the oldest X period of records from 70 tables, every Sunday. The field name, interval (X days or months) and date (CURRENT_DATE or CURRENT_TIMESTAMP) varies for each table.Thus, I put all the relevant data in a tab-separated value file, and use bash to read through it, purging one table at a time. This works well, except for Foreign Key constraints; carefully ordering the file to purge records in the correct order eliminates most FK errors, but not all.Therefore, I created an anonymous DO statement to delete the "deletable" old records, while skipping the ones that would fail from a FK constraint. (Eventually, the records in the FK table will get deleted, so eventually the records who's DELETE failed will succeed in getting deleted.)(NOTE: I cannot change the FK constraints to ON DELETE CASCADE, and nor do I want to fight with the 3rd party app vendor, since it defeats the purpose of FK constraints.)Here's the snippet of bash code:local Schema=$1
local Table=$2
local Field=$3
local DtCol=$4 # CURRENT_TIMESTAMP or CURRENT_DATE
local Thresh=$5 # example: '90 day'
local FQTable=${Schema}.${Table}DeS="DO \$\$
DECLARE
delsum INTEGER = 0;
delcnt INTEGER;
skipsum integer = 0;
cur_row CURSOR FOR
SELECT $Field, ${Table}_idfrom ${FQTable}where $Field < (${DtCol} - interval ${Thresh});
BEGIN
FOR arow IN cur_row
LOOP
BEGIN
DELETE FROM ${FQTable} WHERE CURRENT OF cur_row;
GET DIAGNOSTICS delcnt = ROW_COUNT;
delsum = delsum + delcnt; EXCEPTIONWHEN others THEN
skipsum = skipsum + 1;
RAISE NOTICE ' Skipped ${FQTable} WHERE ${Table}_id = %; ${Field} = %',arow.${Table}_id, arow.${Field};
END;
END LOOP;
RAISE NOTICE 'Sum of deleted rows: %', delsum;
RAISE NOTICE 'Sum of skipped rows: %', skipsum;
END \$\$;
"It generates the perfectly functional SQL:DO $$
DECLARE
delsum INTEGER = 0;
delcnt INTEGER;
skipsum integer = 0;
cur_row CURSOR FOR
SELECT modified_on, check_idfrom tms.checkwhere modified_on < (CURRENT_TIMESTAMP - interval '90 day');
BEGIN
FOR arow IN cur_row
LOOP
BEGIN
DELETE FROM tms.check WHERE CURRENT OF cur_row;
GET DIAGNOSTICS delcnt = ROW_COUNT;
delsum = delsum + delcnt;
EXCEPTION
WHEN others THEN
skipsum = skipsum + 1;
RAISE NOTICE ' Skipped tms.check WHERE check_id = %; modified_on = %',arow.check_id, arow.modified_on;
END;
END LOOP;
RAISE NOTICE 'Sum of deleted rows: %', delsum;
RAISE NOTICE 'Sum of skipped rows: %', skipsum;
END $$;Can I do this better in PL/pgSQL with dynamic SQL (that doesn't get hairy with nested quotes, etc)?
you can pass values by GUC instead
pavel@nemesis:~$ psql -v var="AHOJ"
Assertions: on
psql (17devel)
Type "help" for help.
(2024-04-09 19:07:55) postgres=# select set_config('my.var', :'var', false);
┌────────────┐
│ set_config │
╞════════════╡
│ AHOJ │
└────────────┘
(1 row)
(2024-04-09 19:08:46) postgres=# do $$
postgres$# declare myvar varchar default current_setting('my.var');
postgres$# begin
postgres$# raise notice '%', myvar;
postgres$# end;
postgres$# $$;
NOTICE: AHOJ
DO
Assertions: on
psql (17devel)
Type "help" for help.
(2024-04-09 19:07:55) postgres=# select set_config('my.var', :'var', false);
┌────────────┐
│ set_config │
╞════════════╡
│ AHOJ │
└────────────┘
(1 row)
(2024-04-09 19:08:46) postgres=# do $$
postgres$# declare myvar varchar default current_setting('my.var');
postgres$# begin
postgres$# raise notice '%', myvar;
postgres$# end;
postgres$# $$;
NOTICE: AHOJ
DO
Regards
Pavel
pgsql-general by date: