R stringr unexpected behavior with str_replace & str_pad. Bug or Layer-8 problem?

[AR1337]

Hi, I am using R 4.1.3 and the stringr-package 1.4.0 and get some unexpected results from this code:

stringr::str_replace(string = "5", 
                     pattern = "([0-9]+)", 
                     replacement = stringr::str_pad(string = "\\1", width = 3,  side = "left",  pad = "0"))

Expected: "005"; Result: "05".

All the parts generate the expected results:

(1) The padding

stringr::str_pad(string = "5", width = 3, side = "left", pad = "0") Returns "005"

(2) The regex match

stringr::str_replace(string = "5", pattern = "([0-9]+)", replacement = "\\1") Returns "5".

Only the combination of these two leads to unexpected behavior.

Using an anonymous function works, too:

stringr::str_replace(string = "5", 
                     pattern = "([0-9]+)", 
                     replacement = {\(x) stringr::str_pad(string = x, width = 3,  side = "left",  pad = "0")})

It seems that "\1" refers to the content of the capture group, but the character length is determined from the literal "\1".

stringr::str_replace(string = "5", 
                     pattern = "([0-9]+)", 
                     replacement = {\(x) as.character(nchar(x))})

stringr::str_replace(string = "5", 
                     pattern = "([0-9]+)", 
                     replacement = as.character(nchar("\\1")))

These two examples return "1" and "2". The second example always returns "2" as replacement for the captured group, independend of its content.

Is this intended behavior and am I using the package wrong or did I stumble across an actual issue of the package?

Thx for developing and maintaining such a great library.

For reference: I originally asked this on Stackoverflow.

[hadley]

That all looks ok to me — str_pad() can't know what \\1 is, so you have to use it as a replacement function.