The answer provided here was supposed to work for many columns, how to use ungroup with multiple composite columns in kdb
But it doesn't. For cases like:
tb:([] x: 0 1 10; y:(1 1; 2 2 2; enlist 3); z:("kk";"ttt"; enlist "j");w:("cc";"bbbb";enlist "a")
ungroup1: {[col;tbl] @[tbl where count each tbl col;col;:;raze tbl col]}
This will work:
q)ungroup1[`y;tb]
x y z w
-----------------
0 1 "kk" "cc"
0 1 "kk" "cc"
1 2 "ttt" "bbbb"
1 2 "ttt" "bbbb"
1 2 "ttt" "bbbb"
10 3 ,"j" ,"a"
```q
Again based on single column,
```q
q)ungroup1[`z;tb]
x y z w
-----------------
0 1 1 k "cc"
0 1 1 k "cc"
1 2 2 2 t "bbbb"
1 2 2 2 t "bbbb"
1 2 2 2 t "bbbb"
10 ,3 j ,"a"
This won't work (raises 'length):
ungroup1[`y`z;tb]
Desired result:
x y z w
------------
0 1 k "cc"
0 1 k "cc"
1 2 t "bbbb"
1 2 t "bbbb"
1 2 t "bbbb"
10 3 j ,"a"
May I hear some input from the experts please
Since ungroup is implemented as raze flip each t, you can copy that approach but exclude (and re-join) the columns you don't want to be ungrouped.
q)raze{key[y]#flip[x _y],\:(x:(),x)#y}[`w]each tb
x y z w
-------------
0 1 k "cc"
0 1 k "cc"
1 2 t "bbbb"
1 2 t "bbbb"
1 2 t "bbbb"
10 3 j ,"a"
q)raze{key[y]#flip[x _y],\:(x:(),x)#y}[`y`w]each tb
x y z w
-----------------
0 1 1 k "cc"
0 1 1 k "cc"
1 2 2 2 t "bbbb"
1 2 2 2 t "bbbb"
1 2 2 2 t "bbbb"
10 ,3 j ,"a"
Assumes that the columns you've chosen to ungroup are actually ungroup-able (aka have matching lengths)