R argument switching mutants are equivalent

[tdhock]

hi @agroce
I'm trying to make PRs based on the R data.table mutants Rdatatable/data.table#6114

And I noticed that many of the mutants involve switching the order of arguments, here are two examples.

setops.R:8 by.y becomes cols in either case.

by.y = colnamesInt(y, by.y, check_dups=TRUE)
by.y = colnamesInt(y, check_dups=TRUE, by.y)

transpose.R:63 type.convert becomes x in either case.

indxs = unlist(type.convert, recursive=FALSE, use.names=FALSE)
indxs = unlist( recursive=FALSE,type.convert, use.names=FALSE)

All of these argument switching mutants in R code are equivalent, because R first matches the named arguments, then assigns the positional arguments.

For future mutation testing of R code, it would be useful to avoid generating such mutants, perhaps by adding a rule that if the argument to switch has an equals sign (name=value, so it is a named argument), then do not try switching.

[agroce]

@debsourav33 @kjain14 you guys are the rule hacking experts now; is this hard? It'd actually be better for Python too, since swapping named args is always equivalent

[debsourav33]

Shouldn't be very hard to achieve.. Just need to find a clever way to skip the arguments with '=' in it.. I will try updating the rules and make a PR

[tdhock]

perhaps an alternative to turning it off would be switching values or names in two named arguments

fun(x=1, y=2)
fun(x=2, y=1)

etc

[debsourav33]

Addressed both issues (skip swapping named arguments and swap their values/names instead) in the PR #27

[agroce]

Closed via #27

[tdhock]

yay thanks

[agroce]

@tdhock not on pip yet, though, note

[agroce]

(on pip now)