Issue 4125 - [emscripten] Convert filter classes to C++

compiled/bindings.h

+#ifndef ADBLOCK_PLUS_TOOLS_H
+#define ADBLOCK_PLUS_TOOLS_H
+
+#include <cstdio>
+#include <cstdlib>
+#include <cstring>
+#include <exception>
+#include <map>
+#include <string>
+#include <type_traits>
+#include <utility>
+#include <vector>
+
+#include <emscripten.h>
+
+#include "String.h"
+#include "intrusive_ptr.h"
+
+typedef void* TYPEID;
+
+enum class TypeCategory
+{
+  UNKNOWN,
+  VOID,
+  INT,
+  DEPENDENT_STRING,
+  OWNED_STRING,
+  STRING_REF,
+  CLASS_PTR
+};
+
+template<typename T>
+struct TypeInfo
+{
+  static char c;
+  constexpr operator TYPEID() const
+  {
+    return &c;
+  }
+
+  constexpr operator TypeCategory() const
+  {
+    if (std::is_void<T>())
+      return TypeCategory::VOID;
+
+    if (std::is_integral<T>() || std::is_enum<T>())
+      return TypeCategory::INT;
+
+    if (std::is_same<DependentString,T>() || std::is_same<const DependentString,T>())
+      return TypeCategory::DEPENDENT_STRING;
+
+    if (std::is_same<OwnedString,T>() || std::is_same<const OwnedString,T>())
+      return TypeCategory::OWNED_STRING;
+
+    if (std::is_same<String&,T>() || std::is_same<const String&,T>() ||
+      std::is_same<DependentString&,T>())
+    {
+      return TypeCategory::STRING_REF;
+    }
+
+    if (std::is_pointer<T>() && std::is_class<typename std::remove_pointer<T>::type>())
+      return TypeCategory::CLASS_PTR;
+
+    return TypeCategory::UNKNOWN;
+  }
+
+  TYPEID pointer_type() const
+  {
+    if (std::is_pointer<T>())
+      return TypeInfo<typename std::remove_pointer<T>::type>();
+    else
+      return nullptr;
+  }
+};
+
+template<typename T>
+char TypeInfo<T>::c;
+
+struct FunctionInfo
+{
+  TypeCategory returnType;
+  TYPEID pointerType;
+  std::vector<TypeCategory> args;
+  bool instance_function;
+  int effectiveArgs;
+  TypeCategory effectiveReturnType;
+  char name[1024];
+
+  FunctionInfo()
+  {
+    name[0] = 0;
+  }
+
+  FunctionInfo(TypeCategory returnType, TYPEID pointerType,
+      std::initializer_list<TypeCategory> argTypes, bool instance_function,
+      void* function)
+      : returnType(returnType), pointerType(pointerType),
+        instance_function(instance_function)
+  {
+    name[0] = 0;
+    if (!*reinterpret_cast<int*>(function))
+      return;
+
+    for (auto it = argTypes.begin(); it != argTypes.end(); ++it)
+    {
+      if (*it != TypeCategory::INT && *it != TypeCategory::STRING_REF &&
+          *it != TypeCategory::CLASS_PTR)
+      {
+        throw std::runtime_error("Unexpected function argument type");
+      }
+      args.push_back(*it);
+    }
+
+    if (returnType != TypeCategory::VOID && returnType != TypeCategory::INT &&
+        returnType != TypeCategory::DEPENDENT_STRING &&
+        returnType != TypeCategory::OWNED_STRING &&
+        returnType != TypeCategory::CLASS_PTR)
+    {
+      throw std::runtime_error("Unexpected function return type");
+    }
+
+    effectiveArgs = args.size();
+    effectiveReturnType = returnType;
+    if (instance_function)
+      effectiveArgs++;
+
+    if (returnType == TypeCategory::DEPENDENT_STRING ||
+        returnType == TypeCategory::OWNED_STRING)
+    {
+      effectiveArgs++;
+      effectiveReturnType = TypeCategory::VOID;
+    }
+
+    get_function_name(function, effectiveArgs,
+        effectiveReturnType == TypeCategory::VOID);
+  }
+
+  template<typename ReturnType, typename... Args>
+  FunctionInfo(ReturnType (*function)(Args...))
+      : FunctionInfo(TypeInfo<ReturnType>(),
+        TypeInfo<ReturnType>().pointer_type(), { TypeInfo<Args>()... }, false,
+        &function)
+  {
+  }
+
+  template<typename ClassType, typename ReturnType, typename... Args>
+  FunctionInfo(ReturnType (ClassType::*function)(Args...))
+      : FunctionInfo(TypeInfo<ReturnType>(),
+        TypeInfo<ReturnType>().pointer_type(), { TypeInfo<Args>()... }, true,
+        &function)
+  {
+  }
+
+  template<typename ClassType, typename ReturnType, typename... Args>
+  FunctionInfo(ReturnType (ClassType::*function)(Args...) const)
+      : FunctionInfo(TypeInfo<ReturnType>(),
+        TypeInfo<ReturnType>().pointer_type(), { TypeInfo<Args>()... }, true,
+        &function)
+  {
+  }
+
+  bool empty() const
+  {
+    return name[0] == 0;
+  }
+
+  void get_function_name(void* ptr, int numArgs, bool voidResult)
+  {
+    // This is a hack, C++ won't let us get the mangled function name. JavaScript
+    // is more dynamic so we pass the pointer to our function there. With that
+    // and the function signature we can call the function - with intentionally
+    // invalid parameters so that we will cause a segmentation fault (pointers
+    // cannot be negative). Sometimes the function we are calling will also be
+    // missing from the build. The result is the same: abort() is called. By
+    // replacing abort() we get access to the call stack and search it for the
+    // name of our function.
+
+    EM_ASM_ARGS({
+      var signature = $3 ? "v" : "i";
+      var args = [];
+      for (var i = 0; i < $2; i++)
+      {
+        signature += "i";
+        args.push(-10000);
+      }
+
+      var oldPrintErr = Module.printErr;
+      var oldAbort = abort;
+      Module.printErr = function(){};
+      abort = function()
+      {
+        var stack = [];
+        for (var f = arguments.callee.caller; f; f = f.caller)
+        {
+          if (f.name)
+          {
+            if (f.name.indexOf("dynCall") == 0)
+              break;
+            else
+              stack.push(f.name);
+          }
+        }
+
+        result = stack[stack.length - 1];
+        if (result && result.indexOf("__wrapper") >= 0)
+          result = stack[stack.length - 2];
+        throw result;
+      };
+
+      try
+      {
+        Runtime.dynCall(signature, HEAP32[$1 >> 2], args);
+      }
+      catch(e)
+      {
+        Module.stringToAscii(e, $0);
+      }
+      finally
+      {
+        Module.printErr = oldPrintErr;
+        abort = oldAbort;
+      }
+    }, name, ptr, numArgs, voidResult);
+  }
+};
+
+class NoBaseClass
+{
+};
+
+struct PropertyInfo
+{
+  std::string name;
+  FunctionInfo getter;
+  FunctionInfo setter;
+  std::string jsValue;
+};
+
+struct MethodInfo
+{
+  std::string name;
+  FunctionInfo call;
+};
+
+struct DifferentiatorInfo
+{
+  FunctionInfo call;
+  std::vector<std::pair<int,std::string>> mapping;
+};
+
+struct ClassInfo
+{
+  ClassInfo* baseClass;
+  std::string name;
+  std::vector<PropertyInfo> properties;
+  std::vector<MethodInfo> methods;
+  std::vector<FunctionInfo> initializers;
+  DifferentiatorInfo subclass_differentiator;
+};
+
+std::map<TYPEID,ClassInfo> classes;
+
+void register_class(const char* name, TYPEID classID, TYPEID baseClassID)
+{
+  auto it = classes.find(classID);
+  if (it != classes.end())
+    throw std::runtime_error(std::string("Duplicate definition for class ") + name);
+
+  ClassInfo* baseClass = nullptr;
+  if (baseClassID != TypeInfo<NoBaseClass>())
+  {
+    it = classes.find(baseClassID);
+    if (it == classes.end())
+      throw std::runtime_error(std::string("Unknown base class defined for class ") + name);
+    baseClass = &(it->second);
+  }
+
+  ClassInfo classInfo;
+  classInfo.baseClass = baseClass;
+  classInfo.name = name;
+  classes[classID] = classInfo;
+}
+
+void register_property(TYPEID classID, const char* name,
+    const FunctionInfo& getter, const FunctionInfo& setter,
+    const char* jsValue = "")
+{
+  auto it = classes.find(classID);
+  if (it == classes.end())
+    throw std::runtime_error(std::string("Property defined on unknown class: ") + name);
+
+  PropertyInfo propertyInfo;
+  propertyInfo.name = name;
+  propertyInfo.getter = getter;
+  propertyInfo.setter = setter;
+  propertyInfo.jsValue = jsValue;
+  it->second.properties.push_back(propertyInfo);
+}
+
+void register_method(TYPEID classID, const char* name,
+    const FunctionInfo& call)
+{
+  auto it = classes.find(classID);
+  if (it == classes.end())
+    throw std::runtime_error(std::string("Method defined on unknown class: ") + name);
+
+  MethodInfo methodInfo;
+  methodInfo.name = name;
+  methodInfo.call = call;
+  it->second.methods.push_back(methodInfo);
+}
+
+void register_initializer(TYPEID classID, const FunctionInfo& call)
+{
+  auto it = classes.find(classID);
+  if (it == classes.end())
+    throw std::runtime_error("Initializer defined on unknown class");
+
+  it->second.initializers.push_back(call);
+}
+
+void register_differentiator(TYPEID classID, const FunctionInfo& call,
+    std::vector<std::pair<int,std::string>>& mapping)
+{
+  auto it = classes.find(classID);
+  if (it == classes.end())
+    throw std::runtime_error("Subclass differentiator defined on unknown class");
+
+  if (!it->second.subclass_differentiator.call.empty())
+    throw std::runtime_error("More than one subclass differentiator defined for class " + it->second.name);
+
+  DifferentiatorInfo differentiatorInfo;
+  differentiatorInfo.call = call;
+  differentiatorInfo.mapping = mapping;
+  it->second.subclass_differentiator = differentiatorInfo;
+}
+
+const std::string generateCall(const FunctionInfo& call,
+    std::vector<std::string>& params)
+{
+  if (call.returnType == TypeCategory::DEPENDENT_STRING ||
+      call.returnType == TypeCategory::OWNED_STRING)
+  {
+    params.insert(params.begin(), "string");
+  }
+
+  std::string call_str(call.name);
+  call_str += "(";
+  for (int i = 0; i < params.size(); i++)
+  {
+    if (i > 0)
+      call_str += ", ";
+    call_str += params[i];
+  }
+  call_str += ")";
+
+  if (call.returnType == TypeCategory::VOID)
+    return "  " + call_str + ";\n";
+  else if (call.returnType == TypeCategory::INT)
+    return "  var result = " + call_str + ";\n";
+  else if (call.returnType == TypeCategory::DEPENDENT_STRING ||
+      call.returnType == TypeCategory::OWNED_STRING)
+  {
+    std::string result;
+    result += "  var string = createString();\n";
+    result += "  " + call_str + ";\n";
+    result += "  var result = getStringData(string);\n";
+    if (call.returnType == TypeCategory::OWNED_STRING)
+      result += "  Module._DestroyString(string);\n";
+    return result;
+  }
+  else if (call.returnType == TypeCategory::CLASS_PTR)
+  {
+    std::string result;
+    result += "  var result = " + call_str + ";\n";
+    result += "  if (result)\n";
+    result += "  {\n";
+    result += "    Module._AddRef(result);\n";
+
+    auto it = classes.find(call.pointerType);
+    if (it == classes.end())
+      throw std::runtime_error("Function " + std::string(call.name) + " returns pointer to unknown class");
+
+    const ClassInfo& cls = it->second;
+    if (cls.subclass_differentiator.call.empty())
+      result += "    result = " + cls.name + "(result);\n";
+    else
+    {
+      result += "    var type = " +
+          std::string(cls.subclass_differentiator.call.name) + "(result);\n";
+      result += "    if (type in " + cls.name + "_mapping)\n";
+      result += "      result = new (exports[" + cls.name + "_mapping[type]])(result);\n";
+      result += "    else\n";
+      result += "      throw new Error('Unexpected " + cls.name + " type: ' + type);\n";
+    }
+
+    result += "  }\n";
+    return result;
+  }
+  else
+    throw std::runtime_error("Unexpected return type for " + std::string(call.name));
+}
+
+const std::string wrapCall(const FunctionInfo& call)
+{
+  char buffer[20];
+  bool hasStringArgs = false;
+  std::vector<std::string> params;
+  std::string prefix = "function(";
+  for (int i = 0; i < call.args.size(); i++)
+  {
+    sprintf(buffer, "arg%i", i);
+    if (i > 0)
+      prefix += ", ";
+    prefix += buffer;
+
+
+    if (call.args[i] == TypeCategory::STRING_REF)
+    {
+      hasStringArgs = true;
+      params.push_back(std::string("createString(") + buffer + ")");
+    }
+    else
+      params.push_back(buffer);
+  }
+  prefix += ")\n{\n";
+
+  std::string suffix = "}";
+  if (call.returnType != TypeCategory::VOID)
+    suffix = "  return result;\n" + suffix;
+
+  if (call.returnType == TypeCategory::DEPENDENT_STRING ||
+      call.returnType == TypeCategory::OWNED_STRING || hasStringArgs)
+  {
+    prefix += "  var sp = Runtime.stackSave();\n";
+    suffix = "  Runtime.stackRestore(sp);\n" + suffix;
+  }
+
+  if (call.instance_function)
+    params.insert(params.begin(), "this._pointer");
+
+  return prefix + generateCall(call, params) + suffix;
+}
+
+const std::string generatePropertyDescriptor(const PropertyInfo& property)
+{
+  if (!property.jsValue.empty())
+    return "value: " + property.jsValue;
+
+  std::string result("get: " + wrapCall(property.getter));
+  if (!property.setter.empty())
+    result += ", set: " + wrapCall(property.setter);
+  return result;
+}
+
+void printHelpers()
+{
+  printf("var sizeofString = %i\n", sizeof(String));
+
+  puts(R"(
+function copyString(str, buffer)
+{
+  var length = str.length;
+  for (var i = 0, pointer = (buffer >> 1); i < length; i++, pointer++)
+    HEAP16[pointer] = str.charCodeAt(i);
+  return length;
+}
+
+function createString(str)
+{
+  var length = 0;
+  var buffer = 0;
+  if (str)
+  {
+    buffer = Runtime.stackAlloc(str.length * 2);
+    length = copyString(str, buffer);
+  }
+
+  var result = Module.Runtime.stackAlloc(sizeofString);
+  Module._InitString(result, buffer, length);
+  return result;
+}
+
+function getStringData(str)
+{
+  var length = Module._GetStringLength(str);
+  var pointer = Module._GetStringData(str) >> 1;
+  return String.fromCharCode.apply(String, HEAP16.slice(pointer, pointer + length));
+}
+
+function createClass(superclass)
+{
+  var result = function(pointer)
+  {
+    this._pointer = pointer;
+  };
+  if (superclass)
+    result.prototype = Object.create(superclass.prototype);
+  result.prototype.delete = function()
+  {
+    Module._ReleaseRef(this._pointer);
+  };
+  return result;
+})");
+}
+
+void printClass(ClassInfo& cls)
+{
+  DifferentiatorInfo differentiator = cls.subclass_differentiator;
+  if (!differentiator.call.empty())
+  {
+    printf("var %s_mapping = \n", cls.name.c_str());
+    puts("{");
+    for (auto it = differentiator.mapping.begin(); it != differentiator.mapping.end(); ++it)
+      printf("  %i: '%s',\n", it->first, it->second.c_str());
+    puts("};");
+  }
+
+  printf("exports.%s = createClass(%s);\n", cls.name.c_str(),
+      (cls.baseClass ? ("exports." + cls.baseClass->name).c_str() : ""));
+
+  for (auto it = cls.properties.begin(); it != cls.properties.end(); ++it)
+  {
+    printf("Object.defineProperty(exports.%s.prototype, '%s', {%s});\n",
+        cls.name.c_str(), it->name.c_str(),
+        generatePropertyDescriptor(*it).c_str());
+  }
+
+  for (auto it = cls.methods.begin(); it != cls.methods.end(); ++it)
+  {
+    std::string obj("exports." + cls.name);
+    if (it->call.instance_function)
+      obj += ".prototype";
+    printf("%s.%s = %s;\n", obj.c_str(), it->name.c_str(),
+        wrapCall(it->call).c_str());
+  }
+
+  for (auto it = cls.initializers.begin(); it != cls.initializers.end(); ++it)
+    printf("%s()\n", it->name);
+}
+
+void printBindings()
+{
+  printHelpers();
+
+  for (auto it = classes.begin(); it != classes.end(); ++it)
+    printClass(it->second);
+}
+
+#if defined(PRINT_BINDINGS)
+  // Bindings generation step: collect bindings information and print
+  // corresponding JS code.
+  #define EMSCRIPTEN_BINDINGS \
+      static struct BindingsInitializer {\
+          BindingsInitializer();\
+          BindingsInitializer(bool dummy)\
+          {\
+            try\
+            {\
+              BindingsInitializer();\
+              printBindings();\
+            }\
+            catch (const std::exception& e)\
+            {\
+              EM_ASM_ARGS(\
+                console.error("Error occurred generating JavaScript bindings: " +\
+                    Module.AsciiToString($0)), e.what()\
+              );\
+              abort();\
+            }\
+          }\
+      } BindingsInitializer_instance(true);\
+      BindingsInitializer::BindingsInitializer()
+#else
+  // Actual compilation step: ignore bindings information but define some
+  // exported helper functions necessary for the bindings.
+  #define EMSCRIPTEN_BINDINGS \
+      extern "C"\
+      {\
+        void EMSCRIPTEN_KEEPALIVE InitString(DependentString* str,\
+            String::value_type* data, String::size_type len)\
+        {\
+          /* String is already allocated on stack, we merely need to call*/\
+          /* constructor.*/\
+          new (str) DependentString(data, len);\
+        }\
+        void EMSCRIPTEN_KEEPALIVE DestroyString(OwnedString* str)\
+        {\
+          /* Stack memory will be freed automatically, we need to call*/\
+          /* destructor explicitly however.*/\
+          str->~OwnedString();\
+        }\
+        String::size_type EMSCRIPTEN_KEEPALIVE GetStringLength(\
+            const String& str)\
+        {\
+          return str.length();\
+        }\
+        const String::value_type* EMSCRIPTEN_KEEPALIVE GetStringData(\
+            const String& str)\
+        {\
+          return str.data();\
+        }\
+        void EMSCRIPTEN_KEEPALIVE AddRef(ref_counted* ptr)\
+        {\
+          ptr->AddRef();\
+        }\
+        void EMSCRIPTEN_KEEPALIVE ReleaseRef(ref_counted* ptr)\
+        {\
+          ptr->ReleaseRef();\
+        }\
+      }\
+      void BindingsInitializer_dummy()
+#endif
+
+template<typename ClassType, typename BaseClass = NoBaseClass,
+    typename std::enable_if<std::is_base_of<ref_counted, ClassType>::value>::type* = nullptr>
+class class_
+{
+public:
+  class_(const char* name)
+  {
+    register_class(name, TypeInfo<ClassType>(), TypeInfo<BaseClass>());
+  }
+
+  template<typename FieldType>
+  const class_& property(const char* name,
+      FieldType (ClassType::*getter)() const,
+      void (ClassType::*setter)(FieldType) = 0) const
+  {
+    register_property(TypeInfo<ClassType>(), name, getter, setter);
+    return *this;
+  }
+
+  const class_& class_property(const char* name,
+      const char* jsValue) const
+  {
+    register_property(TypeInfo<ClassType>(), name, FunctionInfo(),
+        FunctionInfo(), jsValue);
+    return *this;
+  }
+
+  template<typename ReturnType, typename... Args>
+  const class_& function(const char* name, ReturnType (ClassType::*method)(Args...)) const
+  {
+    register_method(TypeInfo<ClassType>(), name, method);
+    return *this;
+  }
+
+  template<typename ReturnType, typename... Args>
+  const class_& function(const char* name, ReturnType (ClassType::*method)(Args...) const) const
+  {
+    register_method(TypeInfo<ClassType>(), name, method);
+    return *this;
+  }
+
+  template<typename ReturnType, typename... Args>
+  const class_& class_function(const char* name, ReturnType (*method)(Args...)) const
+  {
+    register_method(TypeInfo<ClassType>(), name, method);
+    return *this;
+  }
+
+  const class_& class_initializer(void (*function)()) const
+  {
+    register_initializer(TypeInfo<ClassType>(), function);
+    return *this;
+  }
+
+  template<typename ReturnType,
+      typename std::enable_if<std::is_convertible<ReturnType, int>::value>::type* = nullptr>
+  const class_& subclass_differentiator(ReturnType (*function)(ClassType*),
+      std::initializer_list<std::pair<ReturnType,const char*>> list) const
+  {
+    std::vector<std::pair<int,std::string>> mapping;
+    for (auto it = list.begin(); it != list.end(); ++it)
+      mapping.push_back(std::pair<int,std::string>(it->first, it->second));
+
+    register_differentiator(TypeInfo<ClassType>(), function, mapping);
+    return *this;
+  }
+};
+
+#endif