I'm trying create a update for a new structure of tables from a old table, but not using functions. I'm trying create a script to it. The old table is like this:
Old table (OldTable
)
name | col_a | col_b | col_c |
---|---|---|---|
ABC | 0 | NULL | 1 |
DEF | 1 | 1 | 1 |
GHI | NULL | 1 | 0 |
And, the new tables:
Users
:
ID | NAME |
---|---|
1 | ABC |
2 | DEF |
3 | GHI |
Rules
:
ID | RULE_NAME |
---|---|
1 | col_a |
2 | col_b |
3 | col_c |
4 | col_d |
UserRule
ID_USER | ID_RULE |
---|
So, I'm needing populate the table UserRules with the result from a select on TableMix where the user name is equals to the user name on table User and the value for column <col_name> in TableMix is equals 1. (select <col_name> from OldTable where OldTable.name = Users.name). Well, I'm trying this:
DO $$
DECLARE rules CURSOR FOR SELECT column_name FROM information_schema.columns
WHERE table_schema = 'public' AND table_name = 'OldTable'
AND data_type = 'numeric' AND column_name NOT IN ('foo','bar');
-- "foo" and "bar" are another numeric cols, but not define rules.
DECLARE users CURSOR FOR SELECT name FROM public.Users;
BEGIN
FOR ruleName IN rules LOOP
FOR userName IN users LOOP
EXECUTE format('SELECT COALESCE(%I,col,1ドル) FROM public.OldTable
WHERE name = ''2ドル''', ruleName, username);
-- insert on... populate the table Rules after get the id of the rule in table Rules if the result of select in OldTable equals 1 (or true if has a "where ... = 1")
END LOOP;
END LOOP;
END $$
After this, I don't have progress.
UPDATE
This way I can get the real value:
DO $$
DECLARE
rules CURSOR FOR SELECT column_name FROM information_schema.columns
WHERE table_schema = 'public' AND table_name = 'oldtable' AND data_type = 'numeric'
AND column_name NOT IN ('foo','bar');
names CURSOR FOR SELECT username FROM public.users;
res integer;
BEGIN
FOR rulename IN rules LOOP
EXECUTE format('SELECT %s FROM oldtable WHERE nome = %L', rulename, 'USER NAME') INTO res;
-- IF res > 0 THEN
RAISE NOTICE '%', res;
--END IF;
END LOOP;
END $$
Result the real integer value.
But if I use a loop for usernames, it's doesn't work.
DO $$
DECLARE
rules CURSOR FOR SELECT column_name FROM information_schema.columns
WHERE table_schema = 'public' AND table_name = 'oldtable' AND data_type = 'numeric'
AND column_name NOT IN ('foo','bar');
names CURSOR FOR SELECT username FROM public.users;
res integer;
BEGIN
FOR name IN names LOOP
FOR rulename IN rules LOOP
EXECUTE format('SELECT %s FROM oldtable WHERE nome = %L', rulename, name) INTO res;
-- IF res > 0 THEN
RAISE NOTICE '%', res;
--END IF;
END LOOP;
END LOOP;
END $$
So, the result is always . Where am I going wrong?
-
Copied and pasted to stackoverflow.com/q/26122833/398670Craig Ringer– Craig Ringer2014年09月30日 14:06:45 +00:00Commented Sep 30, 2014 at 14:06
-
PS: The first error occurs because the column name is printed with parentheses. "SELECT (<col_name>) from ...."GustavoAdolfo– GustavoAdolfo2014年09月30日 14:07:15 +00:00Commented Sep 30, 2014 at 14:07
-
Yes, because there are people who are not in both groups.GustavoAdolfo– GustavoAdolfo2014年09月30日 14:11:15 +00:00Commented Sep 30, 2014 at 14:11
1 Answer 1
I think you might be looking for the USING
clause:
EXECUTE format('SELECT COALESCE(%I,col,1ドル) FROM public.TableMix
WHERE name = 2ドル', ruleName)
USING (username, whateverTheSecondParameterIs);
You use 1ドル
, 2ドル
, etc for placement-parameters, passed by USING
. You use %I
for itentifiers expanded by format
.
Update: If you want to coalesce the column name ruleName
if it's null, perhaps you want something more like:
EXECUTE format('SELECT %I FROM public.TableMix
WHERE name = 1ドル', coalesce(ruleName, 'col'))
USING (username);
??
You need to think about order of evaluation. What's part of the generated SQL, and what's evaluated in order to create that SQL?
-
Thanks! But the initial problem persist: column name is printed with parentheses. "SELECT COALESCE(<col_name>) from ...."GustavoAdolfo– GustavoAdolfo2014年09月30日 14:49:35 +00:00Commented Sep 30, 2014 at 14:49
-
@GustavoAdolfo Hang on, are you trying to coalesce the column name, i.e. "if the column name is null, use the column name
col
instead"? If so, you have to do that outside the dynamic SQL string, in theformat
argument list.Craig Ringer– Craig Ringer2014年09月30日 14:57:22 +00:00Commented Sep 30, 2014 at 14:57 -
pfff!! Yes man! Just the obvious... :) Sorry guys, after a while I could not see the obvious... And, I do not even know pgsql (yet). :( Now... how get the result and progress this work? I no have idea how to do this without giving a lot of laps.GustavoAdolfo– GustavoAdolfo2014年09月30日 15:13:10 +00:00Commented Sep 30, 2014 at 15:13
-
@GustavoAdolfo
EXECUTE ... INTO
. See the pl/pgsql manual, "dynamic SQL" section.Craig Ringer– Craig Ringer2014年09月30日 15:28:55 +00:00Commented Sep 30, 2014 at 15:28 -
@GustavoAdolfo Probably best accept this answer if it's helped then, and follow up with a new question if you're stuck on something else later.Craig Ringer– Craig Ringer2014年10月01日 01:46:17 +00:00Commented Oct 1, 2014 at 1:46
Explore related questions
See similar questions with these tags.