list(FILTER): Add PREDICATE mode

Help/command/list.rst

@@ -19,7 +19,7 @@ Synopsis
 
   `Modification`_
     list(`APPEND`_ <list> [<element>...])
-    list(`FILTER`_ <list> {INCLUDE | EXCLUDE} REGEX <regex>)
+    list(`FILTER`_ <list> <INCLUDE|EXCLUDE> <MODE>)
     list(`INSERT`_ <list> <index> [<element>...])
     list(`POP_BACK`_ <list> [<out-var>...])
     list(`POP_FRONT`_ <list> [<out-var>...])
@@ -120,15 +120,54 @@ Modification
   that empty list.
 
 .. signature::
-  list(FILTER <list> <INCLUDE|EXCLUDE> REGEX <regular_expression>)
+  list(FILTER <list> <INCLUDE|EXCLUDE> <MODE>)
 
   .. versionadded:: 3.6
 
   Includes or removes items from the list that match the mode's pattern.
-  In ``REGEX`` mode, items will be matched against the given regular expression.
 
-  For more information on regular expressions look under
-  :ref:`string(REGEX) <Regex Specification>`.
+  ``<MODE>`` must be one of the following:
+
+    ``REGEX``
+      Items will be matched against the given regular expression.
+
+      .. code-block:: cmake
+
+        list(FILTER <list> <INCLUDE|EXCLUDE> REGEX <regular_expression>)
+
+      For more information on regular expressions look under
+      :ref:`string(REGEX) <Regex Specification>`.
+
+    ``PREDICATE``
+      Specify a user-defined callable as a predicate.
+
+      .. code-block:: cmake
+
+        list(FILTER <list> <INCLUDE|EXCLUDE> PREDICATE <function>)
+
+      .. versionadded:: 4.4
+
+      ``<function>`` is a user-defined :command:`function` that acts as a
+      unary predicate.  The callable must accept exactly two parameters: the
+      input value and the name of an output variable.  The callable must set the
+      output variable to a boolean value in the calling scope.
+      The output variable is interpreted using standard CMake boolean evaluation.
+      If the callable does not set the output variable, it is an error.
+
+      Example:
+
+      .. code-block:: cmake
+
+        function(file_exists path result)
+          if(EXISTS "${path}")
+            set(${result} TRUE PARENT_SCOPE)
+          else()
+            set(${result} FALSE PARENT_SCOPE)
+          endif()
+        endfunction()
+
+        set(candidate_files main.c missing.c utils.c)
+        list(FILTER candidate_files INCLUDE PREDICATE file_exists)
 
 .. signature::
   list(INSERT <list> <element_index> <element> [<element> ...])

Help/release/dev/list-PREDICATE.rst

@@ -0,0 +1,10 @@
+list-PREDICATE
+--------------
+
+* The :command:`list(TRANSFORM)` command gained a new ``PREDICATE`` selector
+  that invokes a user-defined callable to decide which elements are
+  transformed.
+
+* The :command:`list(FILTER)` command gained a new ``PREDICATE`` mode
+  that invokes a user-defined callable to decide which elements are
+  included or excluded, complementing the existing ``REGEX`` mode.

Source/cmList.cxx

@@ -82,15 +82,26 @@ private:
   bool const IncludeMatches;
 };
 
+// Hash of call site (FilePath:Line) for unique variable names across recursive
+// calls.
+std::string OutputVarFor(cm::string_view prefix, cmMakefile& makefile)
+{
+  cmListFileContext context = makefile.GetBacktrace().Top();
+  std::size_t hash =
+    std::hash<std::string>{}(cmStrCat(context.FilePath, ":", context.Line));
+  return cmStrCat(prefix, hash, "_");
+}
+
 class PredicateEvaluator
 {
 public:
-  PredicateEvaluator(std::string const& functionName, cmMakefile& makefile,
+  PredicateEvaluator(std::string functionName, cmMakefile& makefile,
                      std::string errorPrefix = "sub-command TRANSFORM, "
                                                "selector PREDICATE")
-    : FunctionName(functionName)
+    : FunctionName(std::move(functionName))
     , Makefile(&makefile)
     , ErrorPrefix(std::move(errorPrefix))
+    , OutputVar(OutputVarFor("_cmake_predicate_out_", makefile))
   {
     if (!makefile.GetState()->GetCommand(this->FunctionName)) {
       throw cmList::transform_error(cmStrCat(this->ErrorPrefix,
@@ -101,13 +112,13 @@ public:
 
   bool operator()(std::string const& value)
   {
-    std::string const outputVar = "_list_predicate_out_";
-    this->Makefile->RemoveDefinition(outputVar);
+    this->Makefile->RemoveDefinition(this->OutputVar);
 
     cmListFileContext context = this->Makefile->GetBacktrace().Top();
     std::vector<cmListFileArgument> funcArgs;
     funcArgs.emplace_back(value, cmListFileArgument::Quoted, context.Line);
-    funcArgs.emplace_back(outputVar, cmListFileArgument::Quoted, context.Line);
+    funcArgs.emplace_back(this->OutputVar, cmListFileArgument::Quoted,
+                          context.Line);
     cmListFileFunction func{ this->FunctionName, context.Line, context.Line,
                              std::move(funcArgs) };
 
@@ -119,7 +130,7 @@ public:
                  "\" failed during execution."));
     }
 
-    cmValue result = this->Makefile->GetDefinition(outputVar);
+    cmValue result = this->Makefile->GetDefinition(this->OutputVar);
     if (!result) {
       throw cmList::transform_error(
         cmStrCat(this->ErrorPrefix, ": function \"", this->FunctionName,
@@ -127,7 +138,7 @@ public:
     }
 
     bool boolResult = cmIsOn(*result);
-    this->Makefile->RemoveDefinition(outputVar);
+    this->Makefile->RemoveDefinition(this->OutputVar);
     return boolResult;
   }
 
@@ -135,6 +146,26 @@ private:
   std::string FunctionName;
   cmMakefile* Makefile = nullptr;
   std::string ErrorPrefix;
+  std::string OutputVar;
+};
+
+class MatchesPredicate
+{
+public:
+  MatchesPredicate(PredicateEvaluator& evaluator, cmList::FilterMode mode)
+    : Evaluator(evaluator)
+    , IncludeMatches(mode == cmList::FilterMode::INCLUDE)
+  {
+  }
+
+  bool operator()(std::string const& target)
+  {
+    return this->Evaluator(target) ^ this->IncludeMatches;
+  }
+
+private:
+  PredicateEvaluator& Evaluator;
+  bool IncludeMatches;
 };
 }
 
@@ -154,6 +185,23 @@ cmList& cmList::filter(cm::string_view pattern, FilterMode mode)
   return *this;
 }
 
+cmList& cmList::filter(std::string const& functionName, FilterMode mode,
+                       cmMakefile& makefile)
+{
+  try {
+    PredicateEvaluator evaluator(functionName, makefile,
+                                 "sub-command FILTER, mode PREDICATE");
+
+    auto it = std::remove_if(this->Values.begin(), this->Values.end(),
+                             MatchesPredicate{ evaluator, mode });
+    this->Values.erase(it, this->Values.end());
+  } catch (transform_error& e) {
+    throw std::invalid_argument(e.what());
+  }
+
+  return *this;
+}
+
 namespace {
 class StringSorter
 {
@@ -655,6 +703,7 @@ public:
     TransformAction::Initialize(selector);
     this->FunctionName = functionName;
     this->Makefile = &makefile;
+    this->OutputVar = OutputVarFor("_cmake_transform_apply_out_", makefile);
 
     // Validate: command must exist
     if (!makefile.GetState()->GetCommand(this->FunctionName)) {
@@ -678,17 +727,15 @@ public:
       return s;
     }
 
-    // Use a unique output variable name to avoid collisions
-    std::string const outputVar = "_list_transform_apply_out_";
-
     // Unset the output variable before calling
-    this->Makefile->RemoveDefinition(outputVar);
+    this->Makefile->RemoveDefinition(this->OutputVar);
 
     // Build the function call: functionName(s, outputVar)
     cmListFileContext context = this->Makefile->GetBacktrace().Top();
     std::vector<cmListFileArgument> funcArgs;
     funcArgs.emplace_back(s, cmListFileArgument::Quoted, context.Line);
-    funcArgs.emplace_back(outputVar, cmListFileArgument::Quoted, context.Line);
+    funcArgs.emplace_back(this->OutputVar, cmListFileArgument::Quoted,
+                          context.Line);
     cmListFileFunction func{ this->FunctionName, context.Line, context.Line,
                              std::move(funcArgs) };
 
@@ -701,7 +748,7 @@ public:
     }
 
     // Read back the output variable
-    cmValue result = this->Makefile->GetDefinition(outputVar);
+    cmValue result = this->Makefile->GetDefinition(this->OutputVar);
     if (!result) {
       throw transform_error(
         cmStrCat("sub-command TRANSFORM, action APPLY: function \"",
@@ -713,7 +760,7 @@ public:
     std::string output = *result;
 
     // Clean up
-    this->Makefile->RemoveDefinition(outputVar);
+    this->Makefile->RemoveDefinition(this->OutputVar);
 
     return output;
   }
@@ -721,6 +768,7 @@ public:
 private:
   std::string FunctionName;
   cmMakefile* Makefile = nullptr;
+  std::string OutputVar;
 };
 
 // Descriptor of action
@@ -980,16 +1028,19 @@ cmList& cmList::transform(TransformAction action, std::string const& arg,
                           cmMakefile& makefile,
                           std::unique_ptr<TransformSelector> selector)
 {
-  auto descriptor = TransformConfigure(action, selector, 1);
+  // Validate action and arity via the static registry.
+  TransformConfigure(action, selector, 1);
 
-  auto* applyAction =
-    static_cast<TransformActionApply*>(descriptor->Transform.get());
-  applyAction->Initialize(static_cast<::TransformSelector*>(selector.get()),
-                          arg, makefile);
+  // Create a local instance rather than reusing the singleton from
+  // Descriptors.  A user function invoked by APPLY may itself call
+  // list(TRANSFORM ... APPLY ...), which would clobber a shared instance.
+  TransformActionApply applyAction;
+  applyAction.Initialize(static_cast<::TransformSelector*>(selector.get()),
+                         arg, makefile);
 
   static_cast<::TransformSelector&>(*selector).Transform(
-    this->Values, [&descriptor](std::string const& s) -> std::string {
-      return (*descriptor->Transform)(s);
+    this->Values, [&applyAction](std::string const& s) -> std::string {
+      return applyAction(s);
     });
 
   return *this;

Source/cmList.h

@@ -821,8 +821,11 @@ public:
     EXCLUDE
   };
   // Includes or removes items from the list
-  // Throw std::invalid_argument if regular expression is invalid
+  // Throw std::invalid_argument if regular expression is invalid or predicate
+  // function is unknown / does not set its output variable
   cmList& filter(cm::string_view regex, FilterMode mode);
+  cmList& filter(std::string const& functionName, FilterMode mode,
+                 cmMakefile& makefile);
 
   cmList& reverse()
   {
@@ -912,6 +915,9 @@ public:
     template <typename Type>
     static std::unique_ptr<TransformSelector> New(std::string&&);
 
+    // NewPREDICATE is public (unlike NewAT/NewFOR/NewREGEX) because it takes
+    // a cmMakefile& parameter that cannot be dispatched through the existing
+    // New<Type>() templates.
     static std::unique_ptr<TransformSelector> NewPREDICATE(
       std::string const& functionName, cmMakefile& makefile);
 

Source/cmListCommand.cxx

@@ -968,25 +968,46 @@ bool HandleFilterCommand(std::vector<std::string> const& args,
   }
 
   std::string const& mode = args[3];
-  if (mode != "REGEX") {
-    status.SetError("sub-command FILTER does not recognize mode " + mode);
-    return false;
-  }
-  if (args.size() != 5) {
-    status.SetError("sub-command FILTER, mode REGEX "
-                    "requires five arguments.");
-    return false;
-  }
-  std::string const& pattern = args[4];
+  if (mode == "REGEX") {
+    if (args.size() != 5) {
+      status.SetError("sub-command FILTER, mode REGEX "
+                      "requires five arguments.");
+      return false;
+    }
+    std::string const& pattern = args[4];
 
-  try {
-    status.GetMakefile().AddDefinition(
-      listName, list->filter(pattern, filterMode).to_string());
-    return true;
-  } catch (std::invalid_argument& e) {
-    status.SetError(e.what());
-    return false;
+    try {
+      status.GetMakefile().AddDefinition(
+        listName, list->filter(pattern, filterMode).to_string());
+      return true;
+    } catch (std::invalid_argument& e) {
+      status.SetError(e.what());
+      return false;
+    }
   }
+
+  if (mode == "PREDICATE") {
+    if (args.size() != 5) {
+      status.SetError("sub-command FILTER, mode PREDICATE "
+                      "requires five arguments.");
+      return false;
+    }
+    std::string const& functionName = args[4];
+
+    try {
+      status.GetMakefile().AddDefinition(
+        listName,
+        list->filter(functionName, filterMode, status.GetMakefile())
+          .to_string());
+      return true;
+    } catch (std::invalid_argument& e) {
+      status.SetError(e.what());
+      return false;
+    }
+  }
+
+  status.SetError("sub-command FILTER does not recognize mode " + mode);
+  return false;
 }
 } // namespace
 

Tests/RunCMake/list/FILTER-PREDICATE-Exclude-stderr.txt

@@ -0,0 +1,2 @@
+^mylist was: FILTER_THIS_BIT;DO_NOT_FILTER_THIS;thisisanitem;FILTER_THIS_THING
+mylist is: DO_NOT_FILTER_THIS;thisisanitem$

Tests/RunCMake/list/FILTER-PREDICATE-Exclude.cmake

@@ -0,0 +1,50 @@
+# Predicate: returns TRUE for items starting with "FILTER_THIS_"
+function(starts_with_filter input result)
+  if(input MATCHES "^FILTER_THIS_")
+    set(${result} TRUE PARENT_SCOPE)
+  else()
+    set(${result} FALSE PARENT_SCOPE)
+  endif()
+endfunction()
+
+set(mylist FILTER_THIS_BIT DO_NOT_FILTER_THIS thisisanitem FILTER_THIS_THING)
+message("mylist was: ${mylist}")
+list(FILTER mylist EXCLUDE PREDICATE starts_with_filter)
+message("mylist is: ${mylist}")
+
+# EXCLUDE with macro predicate
+macro(is_short input result)
+  string(LENGTH "${input}" _len)
+  if(_len LESS 6)
+    set(${result} TRUE)
+  else()
+    set(${result} FALSE)
+  endif()
+endmacro()
+
+set(mylist ab cdefgh ij klmnop qr)
+list(FILTER mylist EXCLUDE PREDICATE is_short)
+if(NOT mylist STREQUAL "cdefgh;klmnop")
+  message(FATAL_ERROR "FILTER(EXCLUDE PREDICATE macro) is \"${mylist}\", expected \"cdefgh;klmnop\"")
+endif()
+
+# EXCLUDE on empty list
+set(empty_list "")
+list(FILTER empty_list EXCLUDE PREDICATE starts_with_filter)
+if(NOT empty_list STREQUAL "")
+  message(FATAL_ERROR "FILTER(EXCLUDE PREDICATE empty) is \"${empty_list}\", expected \"\"")
+endif()
+
+# EXCLUDE where nothing matches (all elements kept)
+set(mylist alpha bravo charlie)
+list(FILTER mylist EXCLUDE PREDICATE starts_with_filter)
+if(NOT mylist STREQUAL "alpha;bravo;charlie")
+  message(FATAL_ERROR "FILTER(EXCLUDE PREDICATE no-match) is \"${mylist}\", expected \"alpha;bravo;charlie\"")
+endif()
+
+# EXCLUDE where everything matches (all elements removed)
+set(mylist FILTER_THIS_A FILTER_THIS_B FILTER_THIS_C)
+list(FILTER mylist EXCLUDE PREDICATE starts_with_filter)
+if(NOT mylist STREQUAL "")
+  message(FATAL_ERROR "FILTER(EXCLUDE PREDICATE all-match) is \"${mylist}\", expected \"\"")
+endif()

Tests/RunCMake/list/FILTER-PREDICATE-Include-stderr.txt

@@ -0,0 +1,2 @@
+^mylist was: FILTER_THIS_BIT;DO_NOT_FILTER_THIS;thisisanitem;FILTER_THIS_THING
+mylist is: FILTER_THIS_BIT;FILTER_THIS_THING$

Tests/RunCMake/list/FILTER-PREDICATE-Include.cmake

@@ -0,0 +1,51 @@
+# Predicate: returns TRUE for items starting with "FILTER_THIS_"
+function(starts_with_filter input result)
+  if(input MATCHES "^FILTER_THIS_")
+    set(${result} TRUE PARENT_SCOPE)
+  else()
+    set(${result} FALSE PARENT_SCOPE)
+  endif()
+endfunction()
+
+# Predicate as macro
+macro(is_short input result)
+  string(LENGTH "${input}" _len)
+  if(_len LESS 6)
+    set(${result} TRUE)
+  else()
+    set(${result} FALSE)
+  endif()
+endmacro()
+
+set(mylist FILTER_THIS_BIT DO_NOT_FILTER_THIS thisisanitem FILTER_THIS_THING)
+message("mylist was: ${mylist}")
+list(FILTER mylist INCLUDE PREDICATE starts_with_filter)
+message("mylist is: ${mylist}")
+
+# INCLUDE with macro predicate
+set(mylist ab cdefgh ij klmnop qr)
+list(FILTER mylist INCLUDE PREDICATE is_short)
+if(NOT mylist STREQUAL "ab;ij;qr")
+  message(FATAL_ERROR "FILTER(INCLUDE PREDICATE macro) is \"${mylist}\", expected \"ab;ij;qr\"")
+endif()
+
+# INCLUDE on empty list
+set(empty_list "")
+list(FILTER empty_list INCLUDE PREDICATE starts_with_filter)
+if(NOT empty_list STREQUAL "")
+  message(FATAL_ERROR "FILTER(INCLUDE PREDICATE empty) is \"${empty_list}\", expected \"\"")
+endif()
+
+# INCLUDE where nothing matches (all elements removed)
+set(mylist alpha bravo charlie)
+list(FILTER mylist INCLUDE PREDICATE starts_with_filter)
+if(NOT mylist STREQUAL "")
+  message(FATAL_ERROR "FILTER(INCLUDE PREDICATE no-match) is \"${mylist}\", expected \"\"")
+endif()
+
+# INCLUDE where everything matches
+set(mylist FILTER_THIS_A FILTER_THIS_B FILTER_THIS_C)
+list(FILTER mylist INCLUDE PREDICATE starts_with_filter)
+if(NOT mylist STREQUAL "FILTER_THIS_A;FILTER_THIS_B;FILTER_THIS_C")
+  message(FATAL_ERROR "FILTER(INCLUDE PREDICATE all-match) is \"${mylist}\", expected \"FILTER_THIS_A;FILTER_THIS_B;FILTER_THIS_C\"")
+endif()

Tests/RunCMake/list/FILTER-PREDICATE-NoOutput-result.txt

@@ -0,0 +1 @@
+1

Tests/RunCMake/list/FILTER-PREDICATE-NoOutput-stderr.txt

@@ -0,0 +1,5 @@
+^CMake Error at FILTER-PREDICATE-NoOutput\.cmake:6 \(list\):
+  list sub-command FILTER, mode PREDICATE: function "bad_predicate" did not
+  set the output variable\.
+Call Stack \(most recent call first\):
+  CMakeLists\.txt:3 \(include\)$

Tests/RunCMake/list/FILTER-PREDICATE-NoOutput.cmake

@@ -0,0 +1,6 @@
+function(bad_predicate in out)
+  # Deliberately does NOT set ${out}
+endfunction()
+
+set(mylist alpha bravo charlie)
+list(FILTER mylist INCLUDE PREDICATE bad_predicate)

Tests/RunCMake/list/FILTER-PREDICATE-TooManyArguments-result.txt

@@ -0,0 +1 @@
+1

Tests/RunCMake/list/FILTER-PREDICATE-TooManyArguments-stderr.txt

@@ -0,0 +1,4 @@
+^CMake Error at FILTER-PREDICATE-TooManyArguments\.cmake:6 \(list\):
+  list sub-command FILTER, mode PREDICATE requires five arguments\.
+Call Stack \(most recent call first\):
+  CMakeLists\.txt:3 \(include\)$

Tests/RunCMake/list/FILTER-PREDICATE-TooManyArguments.cmake

@@ -0,0 +1,6 @@
+function(my_predicate in out)
+  set(${out} TRUE PARENT_SCOPE)
+endfunction()
+
+set(mylist alpha bravo charlie)
+list(FILTER mylist INCLUDE PREDICATE my_predicate extra_arg)

Tests/RunCMake/list/FILTER-PREDICATE-UnknownFunction-result.txt

@@ -0,0 +1 @@
+1

Tests/RunCMake/list/FILTER-PREDICATE-UnknownFunction-stderr.txt

@@ -0,0 +1,5 @@
+^CMake Error at FILTER-PREDICATE-UnknownFunction\.cmake:2 \(list\):
+  list sub-command FILTER, mode PREDICATE: unknown function
+  "no_such_function"\.
+Call Stack \(most recent call first\):
+  CMakeLists\.txt:3 \(include\)$

Tests/RunCMake/list/FILTER-PREDICATE-UnknownFunction.cmake

@@ -0,0 +1,2 @@
+set(mylist alpha bravo charlie)
+list(FILTER mylist INCLUDE PREDICATE no_such_function)

Tests/RunCMake/list/RunCMakeTest.cmake

@@ -37,6 +37,11 @@ run_cmake(FILTER-REGEX-InvalidMode)
 run_cmake(FILTER-REGEX-InvalidOperator)
 run_cmake(FILTER-REGEX-Valid0)
 run_cmake(FILTER-REGEX-Valid1)
+run_cmake(FILTER-PREDICATE-UnknownFunction)
+run_cmake(FILTER-PREDICATE-NoOutput)
+run_cmake(FILTER-PREDICATE-TooManyArguments)
+run_cmake(FILTER-PREDICATE-Include)
+run_cmake(FILTER-PREDICATE-Exclude)
 
 run_cmake(JOIN-NoArguments)
 run_cmake(JOIN-NoVariable)