How to calculate a windowed rank function?

[daviewales]

Suppose I have some sorted data like this:

test
A
A
B
C
C
C
D
D

And I want to rank the items in order, but restart the rank each time the group changes:

test	rank
A	1
A	2
B	1
C	1
C	2
C	3
D	1
D	2

Can I do this with Visidata?

The closest I've gotten is using the Window function: w with the argument 1 0, then creating a new =Python column:

=1 if test != test_previous else None

This gives me the first value of each group, but doesn't give me the subsequent ranks.
I saw this issue which suggests it might not currently be possible, but thought I'd ask in case there was another way: #1399

(Note, also working around #2279 by creating a new frozen column with ' and manually setting the value of the first cell to NULL with zd.)

[midichef]

Hi, yes, you can do this with Visidata. It took me a while to remember how. I did it with your method: mark the start of the group. The other key steps are:

1. Get the row number, using i (addcol-incr). Rename that column to i: ^ i
2. Let each row see the row number of the start of its group:
   a) Make a window with addcol-window to use the value of the previous row and the value on the current row: w 1 0
   b) Then, make an expression coumn that will be called first. The expression to use has changed as of September 4, 2024.
   For older visidata versions than that, like visidata <= 3.0.2:
   i) Make an expression column which will be called first: = w[0] != w[1] if len(w) == 2 else True and rename it: ^ first.
   For newer versions >= v3.1dev, with commits after September 4, 2024, instead change the expression:
   ii) Make an expression column which will be called first: = w[0] != w[1] if i > 1 else True and rename it: ^ first.
   c) Make a new expression column = i if first else None. Select all rows with gs, then f to fill nulls. Rename it: ^ filled
   d) Then make a final column with the ranks: = i-filled+1.
   

[daviewales]

That is extremely clever. A bit too clever for me to remember it I'm sure, but it will be here for future reference now. Thanks!