`passmissing` and using higher order functions in macros

[ghyatzo]

Hello,

Today I had the need to do something similar:

passmissing(f) = x -> ismissing(x) ? missing : f(x)

f = passmissing(uppercase)
@chain dat begin
	@mutate(COL_A = f(COL_A))
	@distinct(COL_ID)
end

this is the version that works, but I've also tried:

• @mutate(dat, COL_A = passmissing(!!uppercase)(COL_A))
• @mutate(dat, COL_A = passmissing(uppercase)(COL_A))
• @mutate(dat, COL_A = passmissing($uppercase)(COL_A))

but all three fail.

the first with a:

UndefVarError: passmissing($(Expr(:escape, :uppercase))) not defined in TidierData.

the second with

ArgumentError: column name :uppercase not found in the data frame

the third with

UndefVarError: passmissing($uppercase) not defined in TidierData.

Now, I'm definitely stumbling around, but I was wondering if

1. there was a way of signaling that an identifier is a main module symbol and not a column name
2. in a bit of XY-problem fashion, is there a blessed way of replicating the passmissing functionality? skipmissing will return columns that can be smaller, leading to all sort of issues in certain workflows.

[drizk1]

Do you mind putting together a simple ex w a dataframe made so I can play around with it please ?

[ghyatzo]

working version.
(passmissing seems to be reexported by either DataFrames or TidierData. So no need to redefine it)

dat = DataFrame(COL_A = ["a", missing, "b", missing, "a"])`
f = passmissing(uppercase)
@mutate(COL_A = f(COL_A))

would be nice to be able to do:

dat = DataFrame(COL_A = ["a", missing, "b", missing, "a"])`
@mutate(COL_A = passmissing(uppercase)(COL_A))

or even (although less desirable)

@mutate(COL_A = uppercase(COL_A), passmissing=true)

or (like DataFramesMeta.jl)

@mutate(@passmissing COL_A = f(COL_A))

or some variation of this.

Thanks for your time!