Postgresql trigger to select and write data into new table

Question

I am using PostgreSQL 9.1 on Windows.

I am collecting data automatically from a machine into a PostgreSQL DB and this is working correctly. The data is being saved in several tables and the one I am interested in is called testrecord. I have a second static table called testcode and a join between the tables. I want to select some of the incoming data in testrecord and populate a new table called finaldata, where there is a match on parameter and the testid which is not blank (null).

Question 1

Is an INSERT INTO and JOIN the best way of doing this?

Question 2

All the fields at the moment are varchar. When I add a integer field in the table finaldata (such as auto incrementing serial) I get messages such as column "FdataID" is of type integer but expression is of type character varying or column "FdataID" is of type integer but expression is of type record

The code for my function is below:

INSERT INTO finaldata 
    SELECT 
        "testrecord"."Record", 
        "testrecord"."Sample", 
        "testrecord"."SampleOrig",
        "testrecord"."Parameter", 
        "testrecord"."Result", 
        "testrecord"."ResultOrig", 
        "testrecord"."Units", 
        "testrecord"."OperatorID", 
        "testrecord"."ObsTime", 
        "testrecord"."MsgTime", 
        "testcode"."Machine", 
        "testcode"."TestName", 
        "testcode"."ShortTestName", 
        "testcode"."TestID"
    FROM testrecord 
        LEFT OUTER JOIN testcode 
            ON  ("testrecord"."Parameter" = "testcode"."Parameter") 
            AND ("testrecord"."Machine" = "testcode"."Machine")
    WHERE ("testcode"."TestID") IS NOT NULL; 

Answer 1

You wrote:

INSERT INTO finaldata 
    SELECT "testrecord"."Record", /* some column skipped */ 
        FROM testrecord LEFT OUTER JOIN testcode ON ("testrecord"."Parameter" = "testcode"."Parameter") AND ("testrecord"."Machine" = "testcode"."Machine")
        WHERE ("testcode"."TestID") IS NOT NULL; 

The first thing is: Into which columns shall INSERT write the stuff produced by the SELECT part? You have to specify it like this:

INSERT INTO finaldata ("Record", ...)
    SELECT "testrecord"."Record", /* some column skipped */ 
    ...

Your strange message about wrong types might be caused by this.

The second thing is:

... WHERE ("testcode"."TestID") IS NOT NULL; 

Any you imply that you want to use this code inside a function (the trigger function I guess). Each call to this statement would not just insert one new row into finaldata, it would insert everything. Duplicate entries are then a certain thing.

So:

• Fix the INSERT part
• Fix the missing qualifier in the SELECT part.

Oh, yes, question one: As long as a SELECT query produces the correct data, it is good enough to be used within an INSERT :-) If a JOIN helps to produce the correct data -- that's OK.