New yoda_test_linter

[MichaelChirico]

Part of #962

[AshesITR]

Maybe it's worth to craft up an "extended literal" XPath that contains a few whitelisted functions and use that throughout lintr if we want to match a "constant" expression?

First thought would be something like not(SYMBOL_FUNCTION_CALL[not(text() = 'c' or ...)]) etc.

[MichaelChirico]

sounds great, I was thinking something similar. We have something like that in StringsAsFactorsLinter as well (which also has some logic about mixing in rep()), and it can be re-used in the expect_equal($LITERAL, typeof(x)) family of lints too.

[MichaelChirico]

Let's mark this as a TODO for now. we have a messy XPath that works for a limited number of cases, I'd like to generalize it if possible.

For reference here's the XPath for StringsAsFactorsLinter (so we only look for STR_CONST):

//expr[
  expr[SYMBOL_FUNCTION_CALL[text() = 'data.frame']]
  and expr[
    (
      STR_CONST
      or (
        expr[SYMBOL_FUNCTION_CALL[text() = 'c']]
        and expr[STR_CONST]
        and not(expr[SYMBOL or expr])
      )
      or expr[
        SYMBOL_FUNCTION_CALL[text() = 'rep']
        and (
          following-sibling::expr[1][
            STR_CONST
            or (
              expr[SYMBOL_FUNCTION_CALL[text() = 'c']]
              and expr[STR_CONST]
              and not(expr[SYMBOL or expr])
            )
          ]
        )
      ] or expr[SYMBOL_FUNCTION_CALL[
        text() = 'character'
        or text() = 'as.character'
        or text() = 'paste'
        or text() = 'sprintf'
        or text() = 'format'
        or text() = 'formatC'
        or text() = 'prettyNum'
        or text() = 'toString'
        or text() = 'encodeString'
      ]]
    )
    and not(preceding-sibling::SYMBOL_SUB[1][text() = 'row.names'])
  ]
  and not(SYMBOL_SUB[text() = 'stringsAsFactors'])
]

[AshesITR]

LGTM, needs a manual merge though, so not approving for now.