group by all non-aggregate columns

Question

I'm currently using postgres for a datawarehouse and are running into some big queries which are set up the following way:

SELECT
    col_1,
    col_2,
    col_3,
    ...,
    col_41,
    SUM(col_42)
FROM table
GROUP BY
1,2,3,4,..., 41

I'm not fond of the group by syntax using the numbers, but with the sheer number of columns I can see why it would be used.

My main question is: Is there a way to simply group by all columns without an aggregating function?

Something along the lines of:

SELECT
    col_1,
    col_2,
    col_3,
    ...,
    col_41,
    SUM(col_42)
FROM table
GROUP BY ALL

or even

SELECT
    col_1,
    col_2,
    col_3,
    ...,
    col_41,
    SUM(col_42)
FROM table
GROUP BY 1 to 41

would be an improvement.

Answer 1

You can almost do what you want . . . if you are content to have the first column be a tuple rather than a column reference:

SELECT (col1, col2, col3), SUM(x)
FROM t
GROUP BY 1

I don't think there is a way to extract the columns from a tuple without listing them. For instance, the following does not work:

SELECT t.cols.*, sum_x
FROM (SELECT (col1, col2, col3) as cols, SUM(x) as sum_x
      FROM t
      GROUP BY 1
     ) t;

You can reconstruct the columns -- but you are back to verbosity. For an anonymous row type:

SELECT (cols).f1 as col1, (cols).f2 as col2, (cols).f3 as f3, sum_x
FROM (SELECT (col1, col2, col3) as cols, SUM(x) as sum_x
      FROM t
      GROUP BY 1
     ) t;