Spurious snake_case error when using `$` in assignment

[nsoranzo]

get_deseq_dataset.R:60:17: style: Variable and function name style should be snake_case.
        tx2gene$TXNAME <- sub("^transcript:", "", tx2gene$TXNAME)
                ^~~~~~

[russHyde]

Since TXNAME is all caps, rather than in snake_case, the linter seems to be functioning correctly here.

[nsoranzo]

@russHyde TXNAME is not the variable name though, tx2gene is. TXNAME is used with the $ operator to subset a data.frame column by name, it is equivalent to tx2gene[["TXNAME"]] , which lints fine.
Column names are often coming from third-party libraries, so I'd suggest to lint only the variable name (i.e. stop at the $).

[russHyde]

Sorry, I hope it didn't sound glib. It's a difficult call whether subelements of a list/data.frame are variables, since they define variables that are nested in an environment defined by the top-level variable.

It might need quite a bit of thought to make lintr distinguish linting when i) assigning to top-level variables from ii) assigning to their subelements.

If it were possible, and accepting that there are some people who would want the subelement names to be linted by object_name_linter how should the linter API work?

# ? something like this ?
my_linter <- object_name_linter(case = "snake_case", lint_subelements = FALSE)
lint(my_file, linters = list(snake_case_linter = my_linter))

It's a bit verbose but could be thought about.

[AshesITR]

#759 contains the XPath necessary to disambiguate symbols used in OP-DOLLAR contexts vs. as variable names.