How can I add a column that's a shift from another column?

[jabellcu]

I have a column with dates: "Date". I would like to create a new column, "PrevDate" with each row's previous Date value. I have tried to add an expression column with = and Date.shift(), but that doesn't work. Ultimately, I would like to add a column with the duration between two consecutive rows. Is there a way to acomplish this? Many thanks.

[midichef]

Hi, to use the values of rows that are before or after the current row, use the addcol-window command, which is bound to w.

For your example: move the cursor to be in the column "Date", then w then 1 0, and then = and Date_window[1] - Date_window[0]

[jabellcu]

This works, thank you very much. I didn't know about addcol-window. Still, is there a way to add a column applying pandas functions to other columns as pd.Series ie. = and then OtherColumn.shift() ?

[midichef]

I don't use pandas, so I'm unfamiliar with the details of how Series works.

Yes, it can be done, but it is a bit inconvenient. I'll give some example commands for a column named test.
Navigate to the test column, then execute:
Space exec-python vals = [vd.sheet.cursorCol.getTypedValue(r) for r in sheet.rows]
That makes a Python list with the values in the column, accessible as vals.
Then create the pandas series:
Space exec-python s = pandas.Series(vals).shift()
Then, to turn it into a column:
za test_shifted
gs
Space setcol-iter s

(update: I edited to simplify the command sequence, using setcol-iter instead of addcol-incr and =)
(update: I changed the command sequence to no longer use z+ list)

[jabellcu]

Great answer, thank you. I was under the impression visidata's tables were pandas dataframes (and columns would then be series objects).

[midichef]

The answer I gave you was flawed. I've edited it to fix it.

The flaw is that it would give bad results if the starting column had any cells that held null, or contained errors. By errors, I mean the kind of Exception-related errors that would be shown byerror-cell.

The reason is, z+ list ignores any cells that were null or had errors. So vd.memory['test_list'] would have fewer elements than needed to fill in the rows of test_shifted. The rows at the end would be filled in by setcol-iter using repeated values from the start of vd.memory['test_list'].

The edited version, using vals = [vd.sheet.cursorCol.getTypedValue(r) for r in sheet.rows], fixes this problem.