DataFrames version of R's which() function

Specific Domains Data
1

TLDR; is there a Julia version of R’s which() function?

I’m rewriting some code from R to Julia. I’ve gotten comfortable with the DataFrames subset function but in addition to returning rows I occassionally need to access relative to a specific row that meets a given condition. My data records a series of simulated aircraft flights. They always start at a given position but vary in end point and duration. To find the beginning of each run I subset the data as follows:

@subset(df,:XPOSITION.== df.XPOSITION[1])

I’d like to be able to get the last row of each run as well. In R I would use the which() function which returns a vector of indices in the original dataframe that match the given condition. Then I know that the end positions abutt the start positions and I can shift the correct elements of the start position vector to get the end position vector.

Is there a Julia equivelant to R’s which()?

2

You can do

@subset(df, :XPOSITION .== :XPOSITION[1])

which works just fine (and will be faster)

I think you may want to use a grouped data frame? This might not be the most elegant solution, maybe someone else has an answer

julia> df = DataFrame(flight = [1, 1, 1, 1, 2, 2, 2, 2], location = [1, 2, 3, 2, 43, 44, 43, 42])
8×2 DataFrame
 Row │ flight  location
     │ Int64   Int64
─────┼──────────────────
   1 │      1         1
   2 │      1         2
   3 │      1         3
   4 │      1         2
   5 │      2        43
   6 │      2        44
   7 │      2        43
   8 │      2        42

julia> @chain df begin
           groupby(:flight)
           combine(_) do sdf
               sdf[[1, nrow(sdf)], :]
           end
       end
4×2 DataFrame
 Row │ flight  location
     │ Int64   Int64
─────┼──────────────────
   1 │      1         1
   2 │      1         2
   3 │      2        43
   4 │      2        42

If each data frame is one flight, it’s just

df[[1, nrow(df), :]

I can’t think of a great way to write it with DataFramesMeta.jl verbs, to be honest. Maybe someone else will think of something.

To just get the indices, I would recommend using @with rom DataFramesMeta.jl

@with df :XPOSITION .== :XPOSITION[1]

Some R code you use would also be helpful

3

So at the moment my solution is

row = 1:length(df.XPOSITION)
df.row = row
startRows = @subset(df, :XPOSITION.== :XPOSITION[1])
startIndex = startRows.row
endIndex = vcat(startIndex[2:end].-1,last(df.row))

This works but adding in a column of row indices when my data is 6.2+ million rows long seems like a really inefficient way of getting the indices I need.

Unfortunately the data does not have the indivdual flights labeled. That is what I’m going to use the start and end index lists to do.

4

I’m not sure what this last line is doing. But it sounds like you need the function findall. You could do, for example

@with df findall(==(:XPOSITION[1]), :XPOSITION)
5

Yes, I think @rwalters31 is looking for findall.

6

@pdeffebach & @bkamins findall is the function I was hunting for. Thanks!

7

To get the index of the last row of I flight I know that it is the index one prior to the start of the next flight so for example if startIndex = [1,10,17,25] then the end index for flight 1 would be 9 (the index prior to the start of flight 2) and so forth so in this case endIndex = [9,16,24,nrow(df)].

1 Like