New check_value= for test() to ignore 'y' (#7844)

Author: MichaelChirico

R/test.data.table.R

@@ -377,8 +377,8 @@ gc_mem = function() {
 # defers parsing to runtime, allowing the encoding check to run first and avoid source() warnings.
 utf8_check = function(test_str) identical(test_str, enc2native(test_str))
 
-test = function(num, x, y=TRUE,
-                error=NULL, warning=NULL, message=NULL, output=NULL, notOutput=NULL, ignore.warning=NULL,
+test = function(num, x, y=TRUE, ...,
+                error=NULL, warning=NULL, message=NULL, output=NULL, notOutput=NULL, ignore.warning=NULL, check_value=TRUE,
                 options=NULL, env=NULL,
                 context=NULL, requires_utf8=FALSE, optimize=NULL) {
   # if optimization is provided, test across multiple optimization levels
@@ -442,7 +442,7 @@ test = function(num, x, y=TRUE,
   }
   # Usage:
   # i) tests that x equals y when both x and y are supplied, the most common usage
-  # ii) tests that x is TRUE when y isn't supplied
+  # ii) tests that x is TRUE when y isn't supplied, or ignores y entirely when check_value=FALSE
   # iii) if error is supplied, y should be missing and x is tested to result in an error message matching the pattern
   # iv) if warning is supplied, y is checked to equal x, and x should result in a warning message matching the pattern
   # v) if output is supplied, x is evaluated and printed and the output is checked to match the pattern
@@ -636,7 +636,7 @@ test = function(num, x, y=TRUE,
       # nocov end
     }
   }
-  if (!fail && !length(error) && (!length(output) || !missing(y))) {   # TODO test y when output=, too
+  if (check_value && !fail && !length(error) && (!length(output) || !missing(y))) {   # TODO test y when output=, too
     capture.output(y <- try(y, silent=TRUE)) # y might produce verbose output, just toss it
     if (inherits(x, c("Date", "POSIXct"))) storage.mode(x) <- "numeric"
     if (inherits(y, c("Date", "POSIXct"))) storage.mode(y) <- "numeric"

inst/tests/other.Rraw

@@ -225,8 +225,8 @@ if (loaded[["parallel"]]) {
 }
 
 # example(":=", local=TRUE) triggered cedta==FALSE and then error, #2972
-test(14.1, {example(':=', package='data.table', local=TRUE, echo=FALSE); TRUE})
-test(14.2, {example('CJ', package='data.table', local=TRUE, echo=FALSE); TRUE})
+test(14.1, example(':=', package='data.table', local=TRUE, echo=FALSE), check_value=FALSE)
+test(14.2, example('CJ', package='data.table', local=TRUE, echo=FALSE), check_value=FALSE)
 
 if (loaded[["sf"]]) {  #2273
   DT = as.data.table(st_read(system.file("shape/nc.shp", package = "sf"), quiet=TRUE))

inst/tests/tests.Rraw

@@ -11280,10 +11280,10 @@ attr(x, 'tzone') = NULL
 test(1765.1, print(IDateTime(x)), output=".*idate.*itime.*1: 2017-03-1[67]")
 
 # test test's ignore.warning
-test(1765.2, {warning("foo"); 4L}, 4L, ignore.warning="foo")
-test(1765.3, {warning("foo"); 4L}, 4L, ignore.warning="Foo", warning="foo")
-test(1765.4, {warning("foobar1"); warning("foobar2"); warning("FOO"); 4L}, 4L, ignore.warning="bar", warning="FOO")
-test(1765.5, {warning("foobar1"); warning("foobar2"); warning("FOO"); 4L}, 4L, ignore.warning="2", warning=c("foobar1","FOO"))
+test(1765.2, warning("foo"), ignore.warning="foo", check_value=FALSE)
+test(1765.3, warning("foo"), ignore.warning="Foo", warning="foo", check_value=FALSE)
+test(1765.4, {warning("foobar1"); warning("foobar2"); warning("FOO")}, ignore.warning="bar", warning="FOO", check_value=FALSE)
+test(1765.5, {warning("foobar1"); warning("foobar2"); warning("FOO")}, ignore.warning="2", warning=c("foobar1","FOO"), check_value=FALSE)
 
 # print(null.data.table()) should not output NULL as well, #1852
 test(1766, data.table(NULL), notOutput = "NULL$")
@@ -17219,7 +17219,7 @@ test(2186, DT[, if (TRUE) .(a=1L) else .(a=1L, b=2L)], DT,
 
 # col.names='none' should apply when wrapping too, #4270
 DT = setDT(replicate(getOption('width'), 1, simplify = FALSE))
-test(2187, {print(DT, col.names='none'); TRUE}, notOutput="V")
+test(2187, print(DT, col.names='none'), notOutput="V", check_value=FALSE)
 
 # fifelse now supports vector na arguments and coerces NA to other types, PR#4289
 test(2188.01, fifelse(c(TRUE, FALSE, TRUE, NA), 1L, 2L, 1.0), c(1, 2, 1, 1))
@@ -20657,7 +20657,7 @@ test(2304.099, copy(DT)[2L, let(L1=list(NULL), L2=list(NULL))], ans)
 test(2304.100, set(copy(DT), i=2L, j=c("L1", "L2"), value=list(list(NULL), list(NULL))), ans)
 
 # the integer overflow in #6729 is only noticeable with UBSan
-if (test_R.utils) test(2305, { fread(testDir("issue_6729.txt.bz2")); TRUE })
+if (test_R.utils) test(2305, fread(testDir("issue_6729.txt.bz2")), check_value=FALSE)
 
 if (exists("sort_by", "package:base")) {
   # sort_by.data.table
@@ -20691,7 +20691,7 @@ DT <- data.table(a = 1:2, b = 2:1)
 setindex(DT, b)
 # make sure that print(DT) doesn't warn due to the header missing index column types, #6806
 # can't use output= here because the print() call is outside withCallingHandlers(...)
-test(2307, { capture.output(print(DT, class = TRUE, show.indices = TRUE)); TRUE })
+test(2307, capture.output(print(DT, class = TRUE, show.indices = TRUE)), check_value=FALSE)
 
 # fread with colClasses and keepLeadingZeros=TRUE #6851
 dt = data.table(date=as.IDate(c(NA, "2014-12-05")))
@@ -21667,9 +21667,9 @@ test(2370.4, yearmon("2016-08-03 01:02:03.45", format="character"), "2016M08")
 test(2370.5, yearmon(NA, format="character"), NA_character_)
 
 # multiple expected/observed warnings in test() are printed on aligned lines, #7092
-test(2371.1, test(0, {warning("a"); 2L}, 2L, warning=c("a", "b")), FALSE,
+test(2371.1, test(0, warning("a"), warning=c("a", "b"), check_value=FALSE), FALSE,
              output="Test 0 produced 1 warnings but expected 2\nExpected: a\n          b\nObserved: a")
-test(2372.2, test(0, {warning("a"); warning("b"); 2L}, 2L, warning="a"), FALSE,
+test(2372.2, test(0, {warning("a"); warning("b")}, warning="a", check_value=FALSE), FALSE,
              output="Test 0 produced 2 warnings but expected 1\nExpected: a\nObserved: a\n          b")
 
 # group-by on empty table works

man/test.Rd

@@ -5,22 +5,25 @@
   An internal testing function used in \code{data.table} test scripts that are run by \code{\link{test.data.table}}.
 }
 \usage{
-test(num, x, y = TRUE,
+test(num, x, y = TRUE, ...,
      error = NULL, warning = NULL, message = NULL,
      output = NULL, notOutput = NULL, ignore.warning = NULL,
+     check_value = TRUE,
      options = NULL, env = NULL, context = NULL,
      requires_utf8 = FALSE, optimize = NULL)
 }
 \arguments{
 \item{num}{ A unique identifier for a test, helpful in identifying the source of failure when testing is not working. Currently, we use a manually-incremented system with tests formatted as \code{n.m}, where essentially \code{n} indexes an issue and \code{m} indexes aspects of that issue. For the most part, your new PR should only have one value of \code{n} (scroll to the end of \code{inst/tests/tests.Rraw} to see the next available ID) and then index the tests within your PR by increasing \code{m}. Note -- \code{n.m} is interpreted as a number, so \code{123.4} and \code{123.40} are actually the same -- please \code{0}-pad as appropriate. Test identifiers are checked to be in increasing order at runtime to prevent duplicates being possible. }
 \item{x}{ An input expression to be evaluated. }
 \item{y}{ Pre-defined value to compare to \code{x}, by default \code{TRUE}. }
+\item{...}{ Ignored; used to prevent providing subsequent arguments positionally. }
 \item{error}{ When you are testing behaviour of code that you expect to fail with an error, supply the expected error message to this argument. It is interpreted as a regular expression, so you can be abbreviated, but try to include the key portion of the error so as not to accidentally include a different error message. }
 \item{warning}{ Same as \code{error}, in the case that you expect your code to issue a warning. Note that since the code evaluates successfully, you should still supply \code{y}. }
 \item{message}{ Same as \code{warning} but expects \code{message} exception. }
 \item{output}{ If you are testing the printing/console output behaviour; e.g. with \code{verbose=TRUE} or \code{options(datatable.verbose=TRUE)}. Again, regex-compatible and case sensitive. }
 \item{notOutput}{ Or if you are testing that a feature does \emph{not} print particular console output. Case insensitive (unlike output) so that the test does not incorrectly pass just because the string is not found due to case. }
 \item{ignore.warning}{ A single character string. Any warnings emitted by \code{x} that contain this string are dropped. Remaining warnings are compared to the expected \code{warning} as normal. }
+\item{check_value}{ Logical, default \code{TRUE}. If \code{FALSE}, \code{y} is ignored entirely. Useful, for example, when writing a test that \code{x} doesn't error. }
 \item{options}{ A named list of options to set for the duration of the test. Any code evaluated during this call to \code{test()} (usually, \code{x}, or maybe \code{y}) will run with the named options set, and the original options will be restored on return. This is a named list since different options can have different types in general, but in typical usage, only one option is set at a time, in which case a named vector is also accepted. }
 \item{env}{ A named list of environment variables to set for the duration of the test, much like \code{options}. A list entry set to \code{NULL} will unset (i.e., \code{\link{Sys.unsetenv}}) the corresponding variable. }
 \item{context}{ String, default \code{NULL}. Used to provide context where this is useful, e.g. in a test run in a loop where we can't just search for the test number. }