Issue 5131 - [emscripten] Clean separation of bindings code and runtime code

compiled/bindings/generator.h

 /*
  * This file is part of Adblock Plus <https://adblockplus.org/>,
  * Copyright (C) 2006-2017 eyeo GmbH
  *
  * Adblock Plus is free software: you can redistribute it and/or modify
  * it under the terms of the GNU General Public License version 3 as
  * published by the Free Software Foundation.
  *
  * Adblock Plus is distributed in the hope that it will be useful,
  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  * GNU General Public License for more details.
  *
  * You should have received a copy of the GNU General Public License
  * along with Adblock Plus.  If not, see <http://www.gnu.org/licenses/>.
  */
 
 #pragma once
 
 #include <cstddef>
 #include <cstdint>
-#include <cstdio>
 #include <cstdlib>
-#include <exception>
 #include <functional>
-#include <map>
 #include <string>
 #include <type_traits>
 #include <utility>
 #include <vector>
 
 #include <emscripten.h>
 
-#include "String.h"
-#include "intrusive_ptr.h"
+#include "../String.h"
+#include "../intrusive_ptr.h"
 
 namespace bindings_internal
 {
   typedef void* TYPEID;
 
   enum class TypeCategory
   {
     UNKNOWN,
     VOID,
     INT,
@@ skipping 69 matching lines @@
   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();
 
     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';
-
-      // The function parameter is a pointer to the function pointer.
-      // Emscripten's "function pointers" are actually integers indicating the
-      // position in the call table. 0 represents nullptr.
-      if (!*reinterpret_cast<int*>(function))
-        return;
-
-      std::string signature;
-      if (returnType == TypeCategory::DEPENDENT_STRING ||
-          returnType == TypeCategory::OWNED_STRING)
-      {
-        // Objects aren't really returned but passed as parameter. Note that
-        // this pointer might come before it but we don't care because both
-        // are integers (pointers) as far as Emscripten is concerned.
-        signature += "vi";
-      }
-      else if (returnType == TypeCategory::VOID)
-        signature += 'v';
-      else if (returnType == TypeCategory::FLOAT)
-        signature += 'd';
-      else if (returnType == TypeCategory::INT ||
-          returnType == TypeCategory::INT64 ||
-          returnType == TypeCategory::STRING_REF ||
-          returnType == TypeCategory::CLASS_PTR)
-      {
-        signature += 'i';
-      }
-      else
-        throw std::runtime_error("Unexpected function return type");
-
-      if (instance_function)
-      {
-        // this pointer is an implicit parameter
-        signature += 'i';
-      }
-
-      for (const auto& item : argTypes)
-      {
-        if (item == TypeCategory::INT || item == TypeCategory::STRING_REF ||
-            item == TypeCategory::CLASS_PTR)
-        {
-          signature += 'i';
-        }
-        else if (item == TypeCategory::INT64)
-          signature += "ii";
-        else if (item == TypeCategory::FLOAT)
-          signature += 'd';
-        else
-          throw std::runtime_error("Unexpected function argument type");
-        args.push_back(item);
-      }
-
-      get_function_name(function, signature.c_str());
-    }
+        void* function);
 
     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';
-    }
+    bool empty() const;
 
-    void get_function_name(void* ptr, const char* signature)
-    {
-      // 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 a full stack so that we will cause it to abort. Sometimes the
-      // function we are calling will also be missing from the build. The result
-      // is the same: abort() is called which in turn calls stackTrace(). By
-      // replacing stackTrace() we get access to the call stack and search it
-      // for the name of our function.
-
-      EM_ASM_ARGS({
-        var signature = AsciiToString($2);
-        var args = [];
-        for (var i = 1; i < signature.length; i++)
-          args.push(0);
-
-        var oldPrint = Module.print;
-        var oldPrintErr = Module.printErr;
-        var oldStackTrace = stackTrace;
-        var sp = Runtime.stackSave();
-        Module.print = function(){};
-        Module.printErr = function(){};
-        stackTrace = 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;
-        };
-
-        Runtime.stackRestore(STACK_MAX);
-
-        try
-        {
-          Runtime.dynCall(signature, HEAP32[$1 >> 2], args);
-        }
-        catch(e)
-        {
-          Module.stringToAscii(e, $0);
-        }
-        finally
-        {
-          Runtime.stackRestore(sp);
-          Module.print = oldPrint;
-          Module.printErr = oldPrintErr;
-          stackTrace = oldStackTrace;
-        }
-      }, name, ptr, signature);
-    }
+    void get_function_name(void* ptr, const char* signature);
   };
 
   class NoBaseClass
   {
   };
 
   struct PropertyInfo
   {
     std::string name;
     FunctionInfo getter;
     FunctionInfo setter;
     std::string jsValue;
   };
 
   struct MethodInfo
   {
     std::string name;
     FunctionInfo call;
   };
 
   struct DifferentiatorInfo
   {
     size_t offset;
     std::vector<std::pair<int, std::string>> mapping;
   };
 
   struct ClassInfo
   {
-    ClassInfo* baseClass;
+    TYPEID id;
+    TYPEID baseClass;
     std::string name;
     std::vector<PropertyInfo> properties;
     std::vector<MethodInfo> methods;
     DifferentiatorInfo subclass_differentiator;
     ptrdiff_t ref_counted_offset;
   };
 
   typedef std::function<void()>  CustomGenerator;
 
-  std::map<TYPEID, ClassInfo> classes;
-  std::vector<CustomGenerator> customGenerators;
-
   void register_class(const char* name, TYPEID classID, TYPEID baseClassID,
-                      ptrdiff_t ref_counted_offset)
-  {
-    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;
-    classInfo.subclass_differentiator.offset = SIZE_MAX;
-    classInfo.ref_counted_offset = ref_counted_offset;
-    classes[classID] = classInfo;
-  }
+                      ptrdiff_t ref_counted_offset);
 
   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);
-  }
+      const char* jsValue = "");
 
   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);
-  }
+      const FunctionInfo& call);
 
   void register_differentiator(TYPEID classID, size_t offset,
-      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.offset != SIZE_MAX)
-      throw std::runtime_error("More than one subclass differentiator defined for class " + it->second.name);
-
-    DifferentiatorInfo differentiatorInfo;
-    differentiatorInfo.offset = offset;
-    differentiatorInfo.mapping = mapping;
-    it->second.subclass_differentiator = differentiatorInfo;
-  }
+      std::vector<std::pair<int, std::string>>& mapping);
 
   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::vector<std::string>& params);
 
-    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 += ")";
+  const std::string wrapCall(const FunctionInfo& call);
 
-    if (call.returnType == TypeCategory::VOID)
-      return "  " + call_str + ";\n";
-    else if (call.returnType == TypeCategory::INT ||
-          call.returnType == TypeCategory::FLOAT)
-    {
-      return "  var result = " + call_str + ";\n";
-    }
-    else if (call.returnType == TypeCategory::INT64)
-    {
-      // Emscripten saves the high bits of a 64-bit return value in a special
-      // variable called tempRet0. We cannot use bit operators to combine the
-      // values because JavaScript operates on 32-bit integers.
-      return "  var result = (" + call_str + " >>> 0)" +
-             "             + (Runtime.getTempRet0() >>> 0) * 0x100000000;\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 = readString(string);\n";
-      if (call.returnType == TypeCategory::OWNED_STRING)
-        result += "  Module._DestroyString(string);\n";
-      return result;
-    }
-    else if (call.returnType == TypeCategory::STRING_REF)
-    {
-      return "  var result = readString(" + call_str + ");\n";
-    }
-    else if (call.returnType == TypeCategory::CLASS_PTR)
-    {
-      std::string result;
-      result += "  var result = " + call_str + ";\n";
-      result += "  if (result)\n";
-      result += "  {\n";
+  std::string generatePropertyDescriptor(const PropertyInfo& property);
 
-      auto it = classes.find(call.pointerType);
-      if (it == classes.end())
-        throw std::runtime_error("Function " + std::string(call.name) + " returns pointer to unknown class");
+  void printHelpers();
 
-      const ClassInfo& cls = it->second;
-      auto offset = cls.subclass_differentiator.offset;
-      if (offset == SIZE_MAX)
-        result += "    result = exports." + cls.name + "(result);\n";
-      else
-        result += "    result = exports." + cls.name + ".fromPointer(result);\n";
-
-      result += "  }\n";
-      result += "  else\n";
-      result += "    result = null;\n";
-      return result;
-    }
-    else
-      throw std::runtime_error("Unexpected return type for " + std::string(call.name));
-  }
-
-  const std::string wrapCall(const FunctionInfo& call)
-  {
-    bool hasStringArgs = false;
-    std::vector<std::string> params;
-    std::string prefix = "function(";
-    for (int i = 0; i < call.args.size(); i++)
-    {
-      std::string argName("arg" + std::to_string(i));
-      if (i > 0)
-        prefix += ", ";
-      prefix += argName;
-
-      if (call.args[i] == TypeCategory::STRING_REF)
-      {
-        hasStringArgs = true;
-        params.push_back(std::string("createString(") + argName + ")");
-      }
-      else if (call.args[i] == TypeCategory::CLASS_PTR)
-        params.push_back(argName + "._pointer");
-      else if (call.args[i] == TypeCategory::INT64)
-      {
-        // 64-bit integers are passed as two integer parameters
-        params.push_back(argName + " >>> 0");
-        params.push_back(argName + " / 0x100000000 >>> 0");
-      }
-      else
-        params.push_back(argName);
-    }
-    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;
-  }
-
-  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 = Runtime.stackAlloc(sizeofString);
-        Module._InitString(result, buffer, length);
-        return result;
-      }
-
-      function readString(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, ref_counted_offset)
-      {
-        var result = function(pointer)
-        {
-          this._pointer = pointer;
-        };
-        if (superclass)
-          result.prototype = Object.create(superclass.prototype);
-        result.prototype.delete = function()
-        {
-          Module._ReleaseRef(this._pointer + ref_counted_offset);
-        };
-        return result;
-      }
-    )");
-  }
-
-  void printClass(const ClassInfo& cls)
-  {
-    printf("exports.%s = createClass(%s, %i);\n", cls.name.c_str(),
-        (cls.baseClass ? ("exports." + cls.baseClass->name).c_str() : "null"),
-        cls.ref_counted_offset);
-
-    DifferentiatorInfo differentiator = cls.subclass_differentiator;
-    if (differentiator.offset != SIZE_MAX)
-    {
-      printf("var %s_mapping = \n", cls.name.c_str());
-      puts("{");
-      for (const auto& item : differentiator.mapping)
-        printf("  %i: '%s',\n", item.first, item.second.c_str());
-      puts("};");
-
-      printf("exports.%s.fromPointer = function(ptr)\n", cls.name.c_str());
-      puts("{");
-      printf("  var type = HEAP32[ptr + %i >> 2];\n", differentiator.offset);
-      printf("  if (type in %s_mapping)\n", cls.name.c_str());
-      printf("    return new (exports[%s_mapping[type]])(ptr);\n", cls.name.c_str());
-      printf("  throw new Error('Unexpected %s type: ' + type);\n", cls.name.c_str());
-      puts("};");
-    }
-    else
-    {
-      printf("exports.%s.fromPointer = function(ptr)\n", cls.name.c_str());
-      puts("{");
-      printf("  return new exports.%s(ptr);\n", cls.name.c_str());
-      puts("};");
-    }
-
-    for (const auto& item : cls.properties)
-    {
-      printf("Object.defineProperty(exports.%s.prototype, '%s', {%s});\n",
-          cls.name.c_str(), item.name.c_str(),
-          generatePropertyDescriptor(item).c_str());
-    }
-
-    for (const auto& item : cls.methods)
-    {
-      std::string obj("exports." + cls.name);
-      if (item.call.instance_function)
-        obj += ".prototype";
-      printf("%s.%s = %s;\n", obj.c_str(), item.name.c_str(),
-          wrapCall(item.call).c_str());
-    }
-  }
-
-  void printBindings()
-  {
-    printHelpers();
-
-    for (const auto& item : classes)
-      printClass(item.second);
-
-    for (const auto& item : customGenerators)
-      item();
-  }
+  void printClass(const ClassInfo& cls);
 }
 
-#if defined(PRINT_BINDINGS)
-  // Bindings generation step: collect bindings information and print
-  // corresponding JS code.
-  #define EMSCRIPTEN_BINDINGS \
-      void InitializeBindings();\
-      int main()\
-      {\
-        try\
-        {\
-          InitializeBindings();\
-          bindings_internal::printBindings();\
-        }\
-        catch (const std::exception& e)\
-        {\
-          EM_ASM_ARGS(\
-            console.error("Error occurred generating JavaScript bindings: " +\
-                Module.AsciiToString($0)), e.what()\
-          );\
-          abort();\
-        }\
-        return 0;\
-      }\
-      void InitializeBindings()
-#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 ReleaseRef(ref_counted* ptr)\
-        {\
-          ptr->ReleaseRef();\
-        }\
-      }\
-      void BindingsInitializer_dummy()
-#endif
-
 template<typename ClassType,
     typename BaseClass = bindings_internal::NoBaseClass,
     typename std::enable_if<std::is_base_of<ref_counted, ClassType>::value>::type* = nullptr>
 class class_
 {
 public:
   class_(const char* name)

[Wladimir Palant, 2017/04/12 14:17:00]

I would have preferred to move the implementations here into generator.cpp as
well but that appears to be still impossible with the current version of C++
standard.

   {
     ClassType* ptr = reinterpret_cast<ClassType*>(0x10000000);
     ptrdiff_t ref_counted_offset =
         reinterpret_cast<char*>(static_cast<ref_counted*>(ptr)) -
         reinterpret_cast<char*>(ptr);
     bindings_internal::register_class(name,
         bindings_internal::TypeInfo<ClassType>(),
         bindings_internal::TypeInfo<BaseClass>(),
         ref_counted_offset
       );
@@ skipping 54 matching lines @@
     std::vector<std::pair<int, std::string>> mapping;
     for (const auto& item : list)
       mapping.emplace_back(item.first, item.second);
 
     bindings_internal::register_differentiator(
         bindings_internal::TypeInfo<ClassType>(), offset, mapping);
     return *this;
   }
 };
 
-void custom_generator(bindings_internal::CustomGenerator generator)
-{
-  bindings_internal::customGenerators.push_back(generator);
-}
+void custom_generator(bindings_internal::CustomGenerator generator);
+
+void printBindings();