seq_linter() should cover seq(dim(x)[1])

[IndrajeetPatil]

Follow-up on #1467

This should produce a lint:

library(lintr)

lint(
  text = "seq(dim(x)[1])",
  linter = seq_linter()
)

^{Created on 2022-07-26 by the reprex package (v2.0.1)}

Because

seq(dim(data.frame())[1])
#> [1] 1 0

[IndrajeetPatil]

Test to use:

expect_lint(
    "function(x) { seq(dim(x)[1]) }",
    rex("seq(dim(...))", anything, "Use seq_len(...)"),
    linter
)

[IndrajeetPatil]

I saw this test, which should actually fail:

lintr/tests/testthat/test-seq_linter.R

Lines 101 to 107 in a99b3c1

	test_that("reverse seq is ok", {
	expect_lint(
	"function(x) { length(x):1 }",
	rex("length(...):1", anything, "Use seq_along"),
	seq_linter()
	)
	})

Since this is actually fine even in empty edge cases:

length(data.frame()):1
#> [1] 0 1

length(data.frame()):0
#> [1] 0

^{Created on 2022-07-26 by the reprex package (v2.0.1)}

[MichaelChirico]

sorry, I don't follow why length(data.frame()):1 is ok. the reasoning is the same as for 1:length(data.frame()). The point is the is usually used to index columns, but 1 is not a valid column number for a 0-column data.frame.

So

for (jj in length(myDF):1) {
  myDF[[jj]] = foo(myDF, jj + 1) # some recursive definition depending on previous columns
}

is likely a mistake when ncol(myDF) is 0

[IndrajeetPatil]

sorry, I don't follow why length(data.frame()):1 is ok.

I am not saying it is okay. I just saw this already existing test and realized that it is wrong, and so wanted to keep track of it in this issue.

[MichaelChirico]

Sorry, still not following why the test is wrong