Add a GHA for unexported, unused objects

.dev/unused_helpers_test.R

@@ -0,0 +1,55 @@
+# Look for unexported objects which are not used internally
+# TODO(#1513): delete this after we have a dedicated linter for it
+
+ns_contents <- parseNamespaceFile("lintr", "..")
+
+export_names <- ns_contents$exports
+registered_s3_methods <- apply(ns_contents$S3methods[, 1L:2L], 1L, paste, collapse = ".")
+
+pkgload::load_all()
+
+internal_names <- setdiff(
+  ls(asNamespace("lintr"), all.names = TRUE),
+  c(export_names, registered_s3_methods)
+)
+
+# .__global__, etc. 
+internal_names <- grepv("^[.]__", internal_names, invert = TRUE)
+# other generic R names
+internal_names <- setdiff(internal_names, c(".onLoad", ".onAttach", ".packageName"))
+# known false positives
+#   - %||% is a mask for certain versions of R
+#   - addin_lint* are used by RStudio even though they're not exported
+#   - regexes_rd is used in ?object_name_linter by an \Sexpr execution
+internal_names <- setdiff(internal_names, c("%||%", "addin_lint", "addin_lint_package", "regexes_rd"))
+
+is_operator <- make.names(internal_names) != internal_names
+
+operator_linter <- undesirable_operator_linter(internal_names[is_operator])
+function_linter <- undesirable_function_linter(internal_names[!is_operator])
+
+# suppressWarnings: #nolint for linters not present here
+usages <- as.data.frame(suppressWarnings(lint_dir("R", linters = list(operator_linter, function_linter))))
+usages$used_call <- gsub('.*["`]([^"`]*)["`].*', "\\1", usages$message)
+# TODO(#2815): these should be excluded by the linter itself
+usages <- subset(usages, !mapply(\(nm, l) grepl(sprintf("\\b%s <-", rex::rex(nm)), l), used_call, line))
+
+# TODO(#2004): can this just come from get_source_expressions(), and/or will
+#   the above linters "just work" once we're roxygen2-aware?
+options(keep.source = TRUE)
+roxy_usage <- character()
+for (robj in roxygen2::parse_package()) {
+  for (tag in robj$tags) {
+    if (tag$tag == "evalRd") {
+      roxy_usage <- c(roxy_usage,
+        subset(getParseData(tag$val), token == "SYMBOL_FUNCTION_CALL", select = "text", drop = TRUE)
+      )
+    }
+  }
+}
+
+unused_names <- setdiff(internal_names, c(usages$used_call, roxy_usage))
+
+if (length(unused_names) > 0L) {
+  stop("These objects are in the {lintr} namespace but don't have any usage: ", toString(unused_names))
+}

.github/workflows/repo-meta-tests.yaml

@@ -45,3 +45,8 @@ jobs:
         run: |
           callr::rscript(".dev/defunct_linters_test.R")
         shell: Rscript {0}
+
+      - name: Ensure any objects in the namespace are actually needed
+        run: |
+          callr::rscript(".dev/unused_helpers_test.R")
+        shell: Rscript {0}

R/methods.R

@@ -85,39 +85,6 @@ print.lints <- function(x, ...) {
   invisible(x)
 }
 
-trim_output <- function(x, max = 65535L) {
-  # if x is less than the max, just return it
-  if (length(x) <= 0L || nchar(x) <= max) {
-    return(x)
-  }
-
-  # otherwise trim x to the max, then search for the lint starts
-  x <- substr(x, 1L, max)
-
-  re <- rex(
-    "[", except_some_of(":"), ":", numbers, ":", numbers, ":", "]",
-    "(", except_some_of(")"), ")",
-    space,
-    "*", or("style", "warning", "error"), ":", "*",
-    except_some_of("\r\n"), newline,
-    except_some_of("\r\n"), newline,
-    except_some_of("\r\n"), newline,
-    except_some_of("\r\n"), newline,
-    except_some_of("\r\n"), newline
-  )
-
-  lint_starts <- re_matches(x, re, global = TRUE, locations = TRUE)[[1L]]
-
-  # if at least one lint ends before the cutoff, cutoff there, else just use
-  # the cutoff
-  last_end <- lint_starts[NROW(lint_starts), "end"]
-  if (!is.na(last_end)) {
-    substr(x, 1L, last_end)
-  } else {
-    x
-  }
-}
-
 #' @export
 names.lints <- function(x, ...) {
   vapply(x, `[[`, character(1L), "filename")

R/path_utils.R

@@ -57,10 +57,6 @@ is_root_path <- function(path) {
   re_matches(path, root_path_regex)
 }
 
-is_relative_path <- function(path) {
-  re_matches(path, relative_path_regex)
-}
-
 is_path <- function(path) {
   re_matches(path, path_regex)
 }

R/utils.R

@@ -166,11 +166,6 @@ read_lines <- function(file, encoding = settings$encoding, ...) {
   lines
 }
 
-# nocov start
-# support for usethis::use_release_issue(). Make sure to use devtools::load_all() beforehand!
-release_bullets <- function() {}
-# nocov end
-
 # see issue #923, PR #2455 -- some locales ignore _ when running sort(), others don't.
 #   We want to consistently treat "_" < "n" = "N"; C locale does this, which 'radix' uses.
 platform_independent_order <- function(x) order(tolower(x), method = "radix")

tests/testthat/test-absolute_path_linter.R

@@ -52,28 +52,6 @@ test_that("is_absolute_path", {
   expect_identical(f(x), y)
 })
 
-
-test_that("is_relative_path", {
-  f <- lintr:::is_relative_path
-
-  x <- character()
-  y <- logical()
-  expect_identical(f(x), y)
-
-  x <- c("/",   "c:\\", "~/",  "foo", "http://rseek.org/", "'./'")
-  y <- c(FALSE, FALSE,  FALSE, FALSE, FALSE,               FALSE)
-  expect_identical(f(x), y)
-
-  x <- c("/foo", "foo/", "foo/bar", "foo//bar", "./foo", "../foo")
-  y <- c(FALSE,  TRUE,   TRUE,      TRUE,       TRUE,    TRUE)
-  expect_identical(f(x), y)
-
-  x <- c("\\\\", "\\foo", "foo\\", "foo\\bar", ".\\foo", "..\\foo", ".",  "..", "../")
-  y <- c(FALSE,  FALSE,   TRUE,    TRUE,       TRUE,     TRUE,      TRUE, TRUE, TRUE)
-  expect_identical(f(x), y)
-})
-
-
 test_that("is_path", {
   f <- lintr:::is_path
 

tests/testthat/test-methods.R

@@ -1,23 +1,3 @@
-test_that("it returns the input if less than the max", {
-  expect_identical(lintr:::trim_output(character()), character())
-  expect_identical(lintr:::trim_output("test", max = 10L), "test")
-})
-
-test_that("it returns the input trimmed strictly to max if no lints found", {
-  expect_identical(lintr:::trim_output("testing a longer non_lint string", max = 7L), "testing")
-})
-
-test_that("it returns the input trimmed to the last full lint if one exists within the max", {
-  t1 <- readChar(test_path("lints"), file.size(test_path("lints")))
-  if (.Platform$OS.type == "windows") {
-    # Magic numbers expect newlines to be 1 character
-    t1 <- gsub("\r\n", "\n", t1, fixed = TRUE)
-  }
-  expect_identical(lintr:::trim_output(t1, max = 200L), substr(t1, 1L, 195L))
-  expect_identical(lintr:::trim_output(t1, max = 400L), substr(t1, 1L, 380L))
-  expect_identical(lintr:::trim_output(t1, max = 2000L), substr(t1, 1L, 1930L))
-})
-
 test_that("as.data.frame.lints", {
   l1 <- Lint(
     "dummy.R",