How to sum columns that match column selector

[tkbadamdorj]

For example, if I have a list of columns in my table ["height", "width", "depth", "name"], I want to automatically get the numeric columns (age, weight, and height) and add a new column that is their sum without having to type out each column name.

The expected syntax is something ilke:

summed_output = penguins.mutate(
    numeric_sum=sum(s.numeric())
)

or possibly:

summed_output = penguins.mutate(
    numeric_sum=s.apply(s.numeric(), sum)
)

# apply takes a column selector and general function

I was able to work out a solution that's a bit syntactically long:

summed_output = penguins.mutate(
    numeric_sum=sum(
        [
            penguins[col] for col in penguins.select(s.numeric()).columns
        ]
    )
)

is there a cleaner alternative that I'm missing?

thank you!

[tkbadamdorj]

or perhaps there's just a way to get the names of a table's columns without doing this: penguins.select(s.numeric()).columns?

[gforsyth]

Hey @tkbadamdorj! We've talked about how to make these "sum across" operations cleaner but haven't spent the time to really dig in. Your solution is the correct one with the current API. I have a version that uses map and a lambda but I think it's a bit more awkward.

[tkbadamdorj]

thanks @gforsyth !

do you think it would be complicated as a first PR? I could try taking a look at how we could do it if you point me in the right direction. It'd be a good way for me to dive into how ibis works

[gforsyth]

Hey @tkbadamdorj -- sorry for the delay! I missed this notification.

We (the Ibis maintainers) have a very bad track record when it comes to identifying "good first issues", etc.

I think this might be a pretty hard PR to start with. In any case, since it's going to be a larger API discussion, if you want to try it out, I would suggest opening an issue first, so that we can hash out what the API might look like before implementation starts.