The stack function of frame can turn a data frame like this
cola colb colc
1 -> 10 <missing> aaa
3 -> 20 5.5 bb
5 -> 30 <missing> <missing>
6 -> 40 <missing> ccc
INTO
Row Column Value
0 -> 1 cola 10
1 -> 1 colc aaa
2 -> 3 cola 20
3 -> 3 colb 5.5
4 -> 3 colc bb
5 -> 5 cola 30
6 -> 6 cola 40
7 -> 6 colc ccc
However normally it is needed to use one of the column value and the other column's heading as the join key for the new column while doing a unpivot. how can I achieve a result like:
0 -> 10 colb <missing>
1 -> 10 colc aaa
2 -> 20 colb 5.5
3 -> 20 colc bb
4 -> 30 colb <missing>
5 -> 30 colc <missing>
6 -> 40 colb <missing>
7 -> 40 colc ccc
the original cola's value and the column headings for colb and colc have become a combined key to point to colb value and colc value.
How can I achieve this with Deedle?
I don't think we have any built-in function to do this automatically in Deedle, but you can do that by iterating over the rows of the frame and then iterating over the columns:
Assuming f is the sample input frame from your question, the following should do the trick:
[ for r in f.Rows.Values do
for c in r.Keys do
if c <> "cola" then
yield r.Get("cola"), c, r.TryGet(c) ]
|> Frame.ofRecords