Forum Discussion
Array of arrays using Lambda helper functions
PeterBartholomew1Just to get rid of that pesky blank row:
=REDUCE( "", ListCountry#, LAMBDA(acc,val, IF( TYPE(acc)<>64,TRANSPOSE(FILTER(StartList[ATHLETE],StartList[COUNTRY]=val)), VSTACKλ(,acc,TRANSPOSE(FILTER(StartList[ATHLETE],StartList[COUNTRY]=val))) ) ) )I think since MAP will iterate over an array of any shape, the results have to fit back into the same shape; I'd be interested to know if this is a mistaken assumption on my part.
Thank you so much for the reply and the excellent work you present. REDUCE was the next thing I was going to try but only after I had hit a dead end with MAP. I decided to pause to see what ideas you and others might have (and it was coming up to midnight).
I can understand that support for 'arrays of arrays' is out of scope for Excel, but should it do more with what it has. The idea that a vertical array of row arrays/ranges should map to a 2D array (with a specified fill characters for unused cells) would not appear to be that revolutionary; after all the implementations of H- and VSTACK that I have seen go much further.
The problem with the blank row is probably something that should be addressed within VSTACKλ but I wrote it in a hurry to replace Charles Williams's VSTACK function. It would be perfectly reasonable to assume that, should an array be identifiable as 'null', it may be stacked to leave the second array unaltered. Your test has some class, though a slightly 'techie' for a spreadsheet solution.
lori_m Sorry, I meant to include you within the OP but couldn't locate your user id.
PeterBartholomew1 The FastExcel xll you provided with the ACCUMULATE example seems to give me access to the other functions as well (intended or not??), but without the user interface bells and whistles. I started trying to replicate a bunch of those functions with LAMBDA and the new helpers (with the help of the online PDF user guide), but haven't gotten to his _STACK functions just yet; working through DIFF and sent me on a tangent to be able to handle some of its nuances more cleanly.
As you say, it could be handled in STACK just as well as in the REDUCE LAMBDA call, but I think having the inclusive vs. exclusive qualities of REDUCE/SCAN like were in my pre-helper fold/reduce/scan versions would've been nice, though there are probably quite goods reasons why its not set up that way.