Dynamically mutate a column

[ozgurkalan]

Hi,
I try to mutate a column, which's name is assigned to a var.

col = "Mkw2"

if col.startswith("Mkw"):
    t = t.mutate(..mutate..the..col...)

[ozgurkalan]

this is my workaround:

col = "Mkw2"
    if col.startswith("Wkw"):
        t=(t
           .mutate(x=...some..mutate...)
           .drop(col)
           .relabel({"x":col})
          )

[cpcloud]

You should be able to use the across selector here:

t.mutate(s.across(s.c(col), some_mutation))

[jcrist]

No need to drop and relabel the column. You can control the name of the mutated columns using either the .name method on a column expression, or by passing in the mutated columns via keywords as names.

In [1]: import ibis

In [2]: ibis.options.interactive = True

In [3]: from ibis import _

In [4]: t = ibis.memtable({"x": [1, 2, 3], "y": [-1, -2, -3]})

In [5]: t
Out[5]: 
┏━━━━━━━┳━━━━━━━┓
┃ x     ┃ y     ┃
┡━━━━━━━╇━━━━━━━┩
│ int64 │ int64 │
├───────┼───────┤
│     1 │    -1 │
│     2 │    -2 │
│     3 │    -3 │
└───────┴───────┘

In [6]: col = "y"

In [7]: t.mutate(_[col].abs().name(col))  # mutate col, and rename expr back to col
Out[7]: 
┏━━━━━━━┳━━━━━━━┓
┃ x     ┃ y     ┃
┡━━━━━━━╇━━━━━━━┩
│ int64 │ int64 │
├───────┼───────┤
│     1 │     1 │
│     2 │     2 │
│     3 │     3 │
└───────┴───────┘

In [8]: t.mutate(**{col: _[col].abs()})  # same thing, but using keywords as names
Out[8]: 
┏━━━━━━━┳━━━━━━━┓
┃ x     ┃ y     ┃
┡━━━━━━━╇━━━━━━━┩
│ int64 │ int64 │
├───────┼───────┤
│     1 │     1 │
│     2 │     2 │
│     3 │     3 │
└───────┴───────┘