[R] arrow R package: support for stringr::str_replace_all() incomplete

[abfleishman]

Describe the bug, including details regarding any error messages, version, and platform.

The arrow R package claims to support the stringr::str_replace_all() function, however, it does not support using a vector of pattern/replacements as the function says it should.

library(arrow)
#> 
#> Attaching package: 'arrow'
#> The following object is masked from 'package:utils':
#> 
#>     timestamp
library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(stringr)
dat<-data.frame(common_name=c("Cobalt-rumped Parrotlet", 
                              "Striped Woodcreeper", 
                              "Allard's Ground Cricket",
                              "Southern Double-collared Sunbird"))
dat_arrow<-arrow_table(dat)

# if working with a normal data.frame the following code works
dat_df<-dat %>% 
    mutate(common_name_clean = str_replace_all(common_name, c(" |-"="_", "'" = "")))

dat_df
#>                        common_name                common_name_clean
#> 1          Cobalt-rumped Parrotlet          Cobalt_rumped_Parrotlet
#> 2              Striped Woodcreeper              Striped_Woodcreeper
#> 3          Allard's Ground Cricket           Allards_Ground_Cricket
#> 4 Southern Double-collared Sunbird Southern_Double_collared_Sunbird

#but if working with an arrow data.frame it says I have to collect first
dat_arrow<-dat_arrow%>% 
    mutate(common_name_clean = str_replace_all(common_name, c(" |-"="_", "'" = "")))
#> Warning: Expression str_replace_all(common_name, c(` |-` = "_", `'` = "")) not
#> supported in Arrow; pulling data into R
Created on 2023-12-11 with reprex v2.0.2

Component(s)

R

[thisisnic]

Thanks for reporting this @abfleishman! Well-spotted; I think the problem here is that we can only support a single value and not a vector of values, so we should be checking the length of our inputs and providing an error if their lengths are >1. Are you interested in submitting a PR to fix this?

[abfleishman]

@thisisnic While this sounds simple, after a quick search of the code I am not sure where the error message would be added.
Persumidly somewhere near here? I am pretty unclear how these functions work.

The error message could look something like:

if( length(pattern) !=1 || length(replacement) != 1){
    stop("Arrow does not support vectors of length > 1 for the pattern argument for `stringr::str_replace_all` ")
}

[thisisnic]

You got it, yeah!

That arrow_r_string_replace_function() is a generic function which can be used for all of sub/gsub/str_replace/str_replace_all, so if you changed the error message to something more generic like `pattern` must be length 1, then it'd cover multiple cases.

The codebase can be a bit tricky to work through manually, but what I usually do in cases like this is find similar unit tests, then write a unit test for the thing I want to change (in this case it would be attempting to pass in a vector with replacement having length > 1), and then just see if my change makes the new test pass without making other existing tests fail.

If you open a PR, it'll be easier to walk you through any other changes on there :)