sqlpostgresqlpivotpivot-table

PostgreSQL says "return and sql tuple descriptions are incompatible"


I have the following data:

ID  CLASS   VALUE
1   NHB    700905.7243
1   HBW    164216.1311
1   HBO    700905.7243
2   NHB    146023.3792
2   HBW    89543.2972
2   HBO    82152.072
3   NHB    1409818.328
3   HBW    220430.7922
3   HBO    323512.9391
4   NHB    48711.3814
4   HBW    163385.1575
4   HBO    363352.3441

I want to reorganize it into the following:

ID     HBO             HBW              NHB
1   700905.7243    164216.1311      700905.7243
2   82152.072      89543.2972       146023.3792
3   323512.9391    220430.7922      1409818.328
4   363352.3441    163385.1575      48711.3814

Please note that the values in columns HBW, HBO and NHB are totals (sum).

Here is the script I am using to create the output:

-- CREATE EXTENSION tablefunc;

SELECT *
FROM  CROSSTAB
(
    'SELECT _tlfd.id,   
    _tlfd."class",
    _tlfd."value"
    FROM public._tlfd
    WHERE _tlfd."class" = ''HBW'' or _tlfd."class" = ''HBO'' or _tlfd."class" = ''NHB'' 
    ORDER BY 1,2'
) 
    AS
(
    "class" int, 
    "HBW" text,
    "HBO" text,
    "NHB" text,
    --"Purpose" varchar, 
    "value" double precision
);

When I run the script I get this error:

ERROR:  return and sql tuple descriptions are incompatible. 

What am I doing wrong in the script?


Solution

  • This works for me on Postgres 9.3:

    SELECT *
    FROM   crosstab (
     $$SELECT id, class, "value"
       FROM   _tlfd
       WHERE  class = ANY ('{HBW, HBO, NHB}')
       ORDER  BY 1,2$$
       ) AS t (
            class int,                   -- needs a table alias!
            "HBW" float8,                -- resulting columns are double precision!
            "HBO" float8,
            "NHB" float8
            -- "value" double precision  -- column does not exist in result!
        );

    Produces the desired output.

    Essential changes

    • the table alias (bold t)
    • the removed surplus column "value"
    • the correct data type for your data columns (double precision a.k.a. float8)

    The rest is a matter of taste and style. I wouldn't use value as column name though, since it is a reserved word in SQL.

    Basics for crosstab() queries here: