I'm having trouble with writing a procedure in PostgreSQL. I can create the procedure, yet when i try to execute i get an error.The error i get
ERROR: "v_all_atts" is not a known variable
the query is:
CREATE OR REPLACE FUNCTION rebuild_views_with_extra_atts()
RETURNS VOID AS $$
DECLARE
v_all_atts varchar(4000);
BEGIN
CREATE OR REPLACE FUNCTION add_column(p_table text, p_column text,p_category text) RETURNS VOID AS $nothing$
declare
v_column_exists bigint := false ;
BEGIN
SELECT
string_agg( CASE WHEN owner='alarm' THEN 'ai' WHEN owner='fault' THEN 'fi'
END ||'.'||lower(alias) , ', ' ORDER BY owner, alias) AS string
INTO STRICT
v_all_atts
FROM
extra_attribute_cfg
WHERE
owner NOT LIKE 'virtual' and enable = true and v_column_exists = true;
IF LENGTH(v_all_atts) is not null THEN
v_all_atts := ', '||v_all_atts;
END IF;
v_view:= q'#
CREATE OR REPLACE VIEW alarm_view AS
SELECT
fi.fault_id, ai.alarm_id,
#'||v_all_atts||q'#
FROM
alarm ai
INNER JOIN fault fi
ON fi.fault_id = ai.fault_id
#';
EXECUTE v_view;
END;
$nothing$ language plpgsql;
end;
$$ LANGUAGE plpgsql;
I have taken a long look at Postgres documentation and cannot find what is wrong and didn't find any answer to this specific situation
Your rebuild_views_with_extra_atts() function is creating the add_column() function.
add_column() uses the v_all_atts variable, but it doesn't exist in that function, it exists only in the rebuild_views_with_extra_atts() function.
To resolve this, it really depends on what you're trying to do. If that variable should exist in the add_column() function, then declare it in there. If you're trying to use the value of v_all_atts when creating add_column() (e.g. so that the content of the function's body is dependent on the value of that variable), then you really need to use dynamic sql to generate a TEXT version of the CREATE OR REPLACE ... code, then EXECUTE it.
Related
I am trying to create the following function in PostgreSQL but get the following error. This is from a MySQL procedure that I need to convert to PostgreSQL. I am failing to convert the syntax to PostgreSQL. I am a beginner in PostgreSQL. Please assist me.
CREATE OR REPLACE FUNCTION public.usp_failed_analyze4()
RETURNS TABLE(status varchar) as
$BODY$
SET #maxdate = (SELECT MAX(analyzetime) FROM wp_analyze_history);
SET #maxdateint = (SELECT DATEDIFF(NOW() ,MAX(analyzetime)) FROM wp_analyze_history);
SET #STATUS = SELECT Status from wp_analyze_history WHERE Status NOT IN ('OK','Table is already up to date','The Analyze task DID NOT run!') AND analyzetime = #maxdate);
SET #STATUSNOTRUN = 'The Analyze task DID NOT run!';
IF #maxdateint > 7
THEN SELECT #STATUSNOTRUN;
ELSE SELECT #STATUS as "";
$BODY$
LANGUAGE sql;
error: ERROR: syntax error at or near "#"
Position: 109
It's hard to tell what you want as you tried to copy the MySQL 1:1.
However, there are several problems in your code:
language sql does not have variables or IF statements. You need to use PL/pgSQL (language plpgsql)
PL/pgSQL requires a declare block to declare all variables and the actual code needs a begin ... end; block as well.
You can use SET for assignment
To store the result of a single row query in a variable use select ... into ... from
The character # is invalid in an SQL identifier and can't be used for variable names (which follow the rules of SQL identifiers). In Postgres it's a common habit to prefix variable with something to avoid ambiguity with column names. I use l_ for local variables (but that's completely a personal preference)
You don't seem to want to return multiple rows, but a single value. So you don't need returns table
To return something from a function, use return not select
Putting that all together it should look something like this:
CREATE OR REPLACE FUNCTION usp_failed_analyze4()
RETURNS varchar -- return a single value
AS
$BODY$
declare
l_maxdate timestamp;
l_maxdatediff interval;
l_status text;
l_statusnotrun text;
begin
select MAX(analyzetime), current_timestamp - MAX(analyzetime)
into l_maxdate, l_maxdatediff
FROM wp_analyze_history;
SELECT Status
into l_status
from wp_analyze_history
WHERE Status NOT IN ('OK','Table is already up to date','The Analyze task DID NOT run!')
AND analyzetime = l_maxdate;
l_statusnotrun := 'The Analyze task DID NOT run!';
IF l_maxdatediff > interval '7 days'
THEN
return l_statusnotrun;
ELSE
return ''; -- strings are enclosed in single quotes in SQL
end if;
end;
$BODY$
LANGUAGE plpgsql;
There is still room for a lot of optimization, but this matches your initial code as much as possible.
Sample code trimmed down the the bare essentials to demonstrate question:
CREATE OR REPLACE FUNCTION mytest4() RETURNS TEXT AS $$
DECLARE
wc_row wc_files%ROWTYPE;
fieldName TEXT;
BEGIN
SELECT * INTO wc_row FROM wc_files WHERE "fileNumber" = 17117;
-- RETURN wc_row."fileTitle"; -- This works. I get the contents of the field.
fieldName := 'fileTitle';
-- RETURN format('wc_row.%I',fieldName); -- This returns 'wc_row."fileTitle"'
-- but I need the value of it instead.
RETURN EXECUTE format('wc_row.%I',fieldName); -- This gives a syntax error.
END;
$$ LANGUAGE plpgsql;
How can I get the value of a dynamically generated field name in this situation?
Use a trick with the function to_json(), which for a composite type returns a json object with column names as keys:
create or replace function mytest4()
returns text as $$
declare
wc_row wc_files;
fieldname text;
begin
select * into wc_row from wc_files where "filenumber" = 17117;
fieldname := 'filetitle';
return to_json(wc_row)->>fieldname;
end;
$$ language plpgsql;
You don't need tricks. EXECUTE does what you need, you were on the right track already. But RETURN EXECUTE ... is not legal syntax.
CREATE OR REPLACE FUNCTION mytest4(OUT my_col text) AS
$func$
DECLARE
field_name text := 'fileTitle';
BEGIN
EXECUTE format('SELECT %I FROM wc_files WHERE "fileNumber" = 17117', field_name)
INTO my_col; -- data type coerced to text automatically.
END
$func$ LANGUAGE plpgsql;
Since you only want to return a scalar value use EXECUTE .. INTO ... - optionally you can assign to the OUT parameter directly.
RETURN QUERY EXECUTE .. is for returning a set of values.
Use format() to conveniently escape identifiers and avoid SQL injection. Provide identifiers names case sensitive! filetitle is not the same as fileTitle in this context.
Are PostgreSQL column names case-sensitive?
Use an OUT parameter to simplify your code.
I'm using sql fiddle...PostgreSQL 9.3:
CREATE TABLE HotelStays
(roomNum INTEGER NOT NULL,
arrDate DATE NOT NULL,
depDate DATE NOT NULL,
guestName CHAR(30) NOT NULL,
PRIMARY KEY (roomNum, arrDate))
;
CREATE OR REPLACE FUNCTION new_customer() RETURNS void AS
$BODY$
DECLARE
depatureDate DATE;\
BEGIN
SELECT depDate INTO depatureDate FROM HotelStays WHERE OLD.roomNum = NEW.roomNum;
IF (depatureDate <= NEW,arrDate)
INSERT INTO HotelStays (roomNum, arrDate, depDate, guestName)
VALUES (:NEW.roomNum, :NEW.arrDate, :NEW.depDate, :NEW.guestName);
END IF;
RETURN;
END
$BODY$
LANGUAGE 'plpgsql' ;
CREATE TRIGGER;
INSERT INTO HotelStays(roomNum, arrDate, depDate, guestName)
VALUES
(123, to_date('20160202', 'YYYYMMDD'), to_date('20160206','YYYYMMDD'), 'A');
Problem I am trying to solve: a new entry (for a new guest) could be put in for a room number, even before the existing guest has checked out.
I'm trying to solve this question using triggers. Please help me out. Thanks in advance.
As author mentioned, he used SQL Fiddle. I had the same problem with db-fiddle.com and resolved it by replacing $$ or $BODY$ with single quotes ' (and doubling single quotes elsewhere in between.
For example, db-fiddle
CREATE OR REPLACE FUNCTION update_datem()
RETURNS trigger AS
'
BEGIN
NEW.dateM = DATE_TRUNC(''MONTH'', NEW.date);
RETURN NEW;
END;
'
LANGUAGE plpgsql;
There are several errors in your code. First the backslash in depatureDate DATE;\. You are also missing a THEN for the IF clause and new does not need a : in front of it. You also have a , instead of a . in NEW,arrDate. And the final END is missing a ;.
Not an error, but the language name is an identifier, do not put it in single quotes.
The line CREATE TRIGGER; is also wrong. If you want to create trigger your function also needs to be declared as returns trigger and has to return the new row if it is a "before" trigger. If you intend to use an after trigger you still need to return something from that.
I am not sure what the condition WHERE OLD.roomNum = NEW.roomNum; is supposed to select. If you want to get the room number of the changed row, just use new.depdate. The select .. into ... will fail if that query returns more then one row. You probably meant to use where roomnum = new.roomnum or something similar.
So the function should be something like this:
CREATE OR REPLACE FUNCTION new_customer()
RETURNS trigger
AS
$BODY$
DECLARE
depatureDate DATE;
BEGIN
SELECT depDate
INTO depatureDate
FROM HotelStays
WHERE roomNum = NEW.roomNum;
IF (depatureDate <= NEW.arrDate) THEN
INSERT INTO HotelStays (roomNum, arrDate, depDate, guestName)
VALUES (NEW.roomNum, nEW.arrDate, NEW.depDate, NEW.guestName);
END IF;
RETURN NEW; -- this is important for a trigger
END;
$BODY$
LANGUAGE plpgsql;
And the code to create the trigger would be something like this:
CREATE TRIGGER check_stays
before update or insert on hotelstays
execute procedure new_customer();
Is it possible to get the current OID within a function? Like:
CREATE FUNCTION foo()
RETURNS numeric
LANGUAGE plpgsql
AS '
BEGIN
return THIS_FUNCTIONS_OID;
END
';
I need this, because I created function foo within different schemas so the functions name is not helpful here.
I guess you are looking smth like
return select oid from pg_proc where proname='$0';
I doubt you can get it as variable. You can get the name from current_query(), but it will be very not reliable... Unless you define function name as first argument each time you call it :), then you can use $1, but it is not much reliable either...
I don't know what your are doing, but I am sure you don't do it well :). Usually, these strange requirements are related to a strange design and result in code that's hard to maintain.
But you can get the oid of the current function easily with PostgreSQL 9.4 and higher. (This info is easily accessible in C PL functions, but it is hidden in PLpgSQL.) Much easier if your functions are from other schemas than public:
CREATE OR REPLACE FUNCTION omega.inner_func()
RETURNS oid AS $$
DECLARE
stack text; fcesig text;
BEGIN
GET DIAGNOSTICS stack = PG_CONTEXT;
fcesig := substring(stack from 'function (.*?) line');
RETURN fcesig::regprocedure::oid;
END;
$$ LANGUAGE plpgsql;
For functions from the public schema it is a little bit more difficult - there is an inconsistency and without explicitly appending the prefix "public" the cast to regprocedure should not work when public is not in search_path. A generic solution needs a few more lines:
CREATE OR REPLACE FUNCTION omega.inner_func()
RETURNS oid AS $$
DECLARE
stack text; fcesig text; retoid oid;
BEGIN
GET DIAGNOSTICS stack = PG_CONTEXT;
fcesig := substring(stack from 'function (.*?) line');
retoid := to_regprocedure(fcesig::cstring);
IF retoid IS NOT NULL THEN RETURN retoid; END IF;
RETURN to_regprocedure(('public.' || fcesig)::cstring);
END;
$$ LANGUAGE plpgsql;
Question: How can I declare a variable of the same type a parameter in a stored function?
The simple answer is use %TYPE, this works:
CREATE OR REPLACE FUNCTION test_function_1(param1 text)
RETURNS integer AS
$BODY$
DECLARE
myVariable param1%TYPE;
BEGIN
return 1;
END;
$BODY$
LANGUAGE plpgsql VOLATILE
COST 100;
But the problem is when param1 is a composite type:
CREATE TYPE comp_type as
(
field1 text
)
CREATE OR REPLACE FUNCTION test_function_2(param1 comp_type)
RETURNS integer AS
$BODY$
DECLARE
myVariable param1%TYPE;
BEGIN
return 1;
END;
$BODY$
LANGUAGE plpgsql VOLATILE
COST 100;
This doesn't work:
ERROR: type comp_type does not exist [SQL State=42704]
So how can I do when param1 is a composite type?
(Note: Just myVariable comp_type is not a good option because my function is slightly more complex.)
Edited:
I had a mistake on copy&paste, the real error is:
ERROR: invalid type name "param1%TYPE"
Position: 130 [SQL State=42601]
And using param1%ROWTYPE the error is:
ERROR: relation "param1" does not exist
Where: compilation of PL/pgSQL function "test_function_2" near line 3 [SQL State=42P01]
Use %ROWTYPE in that case.
Simple case
Tests by A.H. and DavidEG have shown this won't work. Interesting problem!
You could try a workaround. As long as your definition is like the example you can simply resort to
CREATE FUNCTION test(param1 comp_type)
RETURNS integer
LANGUAGE plpgsql VOLATILE AS
$func$
DECLARE
myvar comp_type;
BEGIN
RETURN 1;
END
$func$;
But your real problem is probably not as simple?
The real problem
As expected, the real problem turns out to be more complex: a polymorphic input type.
Workaround for that scenario was harder, but should work flawlessly:
CREATE FUNCTION test(param1 anyelement, OUT a integer, OUT myvar anyelement)
RETURNS record
LANGUAGE plpgsql VOLATILE AS
$func$
BEGIN
myvar := $1; -- myvar has the required type now.
-- do stuff with myvar.
myvar := NULL; -- reset if you don't want to output ..
a := 1;
END;
$func$;
Call:
SELECT a FROM test('("foo")'::comp_type); -- just retrieve a, ignore myvar
See full output:
SELECT * FROM test('("foo")'::comp_type);
Note for Postgres 9.0+
There has been a crucial update in Postgres 9.0. The release notes:
Allow input parameters to be assigned values within PL/pgSQL functions
(Steve Prentice)
Formerly, input parameters were treated as being declared CONST, so
the function's code could not change their values. This restriction
has been removed to simplify porting of functions from other DBMSes
that do not impose the equivalent restriction. An input parameter now
acts like a local variable initialized to the passed-in value.
In addition to my workaround, you can (ab)use input variables directly now.
Dynamic field names
See:
How to clone a RECORD in PostgreSQL
How to set value of composite variable field using dynamic SQL