Unified callback system.

base/callback.h.pump

Index: base/callback.h.pump
diff --git a/base/callback.h.pump b/base/callback.h.pump
new file mode 100644
index 0000000000000000000000000000000000000000..9fc4b0b0ce9743e862defe7314fb6188919e4334
--- /dev/null
+++ b/base/callback.h.pump
@@ -0,0 +1,291 @@
+$$ This is a pump file for generating file templates.  Pump is a python
+$$ script that is part of the Google Test suite of utilities.  Description
+$$ can be found here:
+$$
+$$ http://code.google.com/p/googletest/wiki/PumpManual
+$$
+
+$var MAX_ARITY = 6
+
+// Copyright (c) 2011 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_CALLBACK_H_
+#define BASE_CALLBACK_H_
+#pragma once
+
+#include "base/callback_helpers.h"
+#include "base/callback_old.h"
+
+// New, super-duper, unified Callback system.  This will eventually replace
+// NewRunnableMethod, NewRunnableFunction, CreateFunctor, and CreateCallback
+// systems currently in the Chromium code base.
+//
+// WHAT IS THIS:
+//
+// The templated Callback class is a generalized function object. Together
+// with the Bind() function in bind.h, they provide a type-safe method for
+// performing currying of arguments, and creating a "closure."
+//
+// In programing languages, a closure is a first-class function where all its
+// parameters have been bound (usually via currying).  Closures are well
+// suited for representing, and passing around a unit of delayed execution.
+// They are used in Chromium code to schedule tasks on different MessageLoops.
+//
+//
+// MEMORY MANAGEMENT AND PASSING
+//
+// The Callback objects themselves should be passed by const-reference, and
+// stored by copy. They internally store their state via a refcounted class
+// and thus do not need to be deleted.
+//
+// The reason to pass via a const-reference is to avoid unnecessary
+// AddRef/Release pairs to the internal state.
+//
+//
+// EXAMPLE USAGE:
+//
+// /* Binding a normal function. */
+// int Return5() { return 5; }
+// base::Callback<int(int)> func_cb = base::Bind(&Return5);
+// LOG(INFO) << func_cb.Run(5);  // Prints 5.
+//
+// void PrintHi() { LOG(INFO) << "hi."; }
+// base::Closure void_func_cb = base::Bind(&PrintHi);
+// LOG(INFO) << void_func_cb.Run();  // Prints: hi.
+//
+// /* Binding a class method. */
+// class Ref : public RefCountedThreadSafe<Ref> {
+//  public:
+//   int Foo() { return 3; }
+//   void PrintBye() { LOG(INFO) << "bye."; }
+// };
+// scoped_refptr<Ref> ref = new Ref();
+// base::Callback<int(void)> ref_cb = base::Bind(&Ref::Foo, ref.get());
+// LOG(INFO) << ref_cb.Run();  // Prints out 3.
+//
+// base::Closure void_ref_cb = base::Bind(&Ref::PrintBye, ref.get());
+// void_ref_cb.Run();  // Prints: bye.
+//
+// /* Binding a class method in a non-refcounted class.
+//  *
+//  * WARNING: You must be sure the referee outlives the callback!
+//  *          This is particularly important if you post a closure to a
+//  *          MessageLoop because then it becomes hard to know what the
+//  *          lifetime of the referee needs to be.
+//  */
+// class NoRef {
+//  public:
+//   int Foo() { return 4; }
+//   void PrintWhy() { LOG(INFO) << "why???"; }
+// };
+// NoRef no_ref;
+// base::Callback<int(void)> base::no_ref_cb =
+//     base::Bind(&NoRef::Foo, base::Unretained(&no_ref));
+// LOG(INFO) << ref_cb.Run();  // Prints out 4.
+//
+// base::Closure void_no_ref_cb =
+//     base::Bind(&NoRef::PrintWhy, base::Unretained(no_ref));
+// void_no_ref_cb.Run();  // Prints: why???
+//
+// /* Binding a reference. */
+// int Identity(int n) { return n; }
+// int value = 1;
+// base::Callback<int(void)> bound_copy_cb = base::Bind(&Identity, value);
+// base::Callback<int(void)> bound_ref_cb =
+//     base::Bind(&Identity, base::ConstRef(value));
+// LOG(INFO) << bound_copy_cb.Run();  // Prints 1.
+// LOG(INFO) << bound_ref_cb.Run();  // Prints 1.
+// value = 2;
+// LOG(INFO) << bound_copy_cb.Run();  // Prints 1.
+// LOG(INFO) << bound_ref_cb.Run();  // Prints 2.
+//
+//
+// WHERE IS THIS DESIGN FROM:
+//
+// The design Callback and Bind is heavily influenced by C++'s
+// tr1::function/tr1::bind, and by the "Google Callback" system used inside
+// Google.
+//
+//
+// HOW THE IMPLEMENTATION WORKS:
+//
+// There are three main components to the system:
+//   1) The Callback classes.
+//   2) The Bind() functions.
+//   3) The arguments wrappers (eg., Unretained() and ConstRef()).
+//
+// The Callback classes represent a generic function pointer. Internally,
+// it stores a refcounted piece of state that represents the target function
+// and all its bound parameters.  Each Callback specialization has a templated
+// constructor that takes an InvokerStorageHolder<> object.  In the context of
+// the constructor, the static type of this InvokerStorageHolder<> object
+// uniquely identifies the function it is representing, all its bound
+// parameters, and a DoInvoke() that is capable of invoking the target.
+//
+// Callback's constructor is takes the InvokerStorageHolder<> that has the
+// full static type and erases the target function type, and the bound
+// parameters.  It does this by storing a pointer to the specific DoInvoke()
+// function, and upcasting the state of InvokerStorageHolder<> to a
+// InvokerStorageBase. This is safe as long as this InvokerStorageBase pointer
+// is only used with the stored DoInvoke() pointer.
+//
+// To create InvokerStorageHolder<> objects, we use the Bind() functions.
+// These functions, along with a set of internal templates, are reponsible for
+//
+//  - Unwrapping the function signature into return type, and parameters
+//  - Determining the number of parameters that are bound
+//  - Creating the storage for the bound parameters
+//  - Performing compile-time asserts to avoid error-prone behavior
+//  - Returning an InvokerStorageHolder<> with an DoInvoke() that has an arity
+//    matching the number of unbound parameters, and knows the correct
+//    refcounting semantics for the target object if we are binding a class
+//    method.
+//
+// The Bind functions do the above using type-inference, and template
+// specializations.
+//
+// By default Bind() will store copies of all bound parameters, and attempt
+// to refcount a target object if the function being bound is a class method.
+//
+// To change this behavior, we introduce a set of argument wrappers
+// (eg. Unretained(), and ConstRef()).  These are simple container templates
+// that are passed by value, and wrap a pointer to argument.  See the
+// file-level comment in base/bind_helpers.h for more info.
+//
+// These types are passed to the Unwrap() functions, and the MaybeRefcount()
+// functions respectively to modify the behavior of Bind().  The Unwrap()
+// and MaybeRefcount() functions change behavior by doing partial
+// specialization based on whether or not a parameter is a wrapper type.
+//
+// ConstRef() is similar to tr1::cref.  Unretained() is specific to Chromium.
+//
+//
+// WHY NOT TR1 FUNCTION/BIND?
+//
+// Direct use of tr1::function and tr1::bind was considered, but ultimately
+// rejected because of the number of copy constructors invocations involved
+// in the binding of arguments during construction, and the forwarding of
+// arguments during invocation.  These copies will no longer be an issue in
+// C++0x because C++0x will support rvalue reference allowing for the compiler
+// to avoid these copies.  However, waiting for C++0x is not an option.
+//
+// Measured with valgrind on gcc version 4.4.3 (Ubuntu 4.4.3-4ubuntu5), the
+// tr1::bind call itself will invoke a non-trivial copy constructor three times
+// for each bound parameter.  Also, each when passing a tr1::function, each
+// bound argument will be copied again.
+//
+// In addition to the copies taken at binding and invocation, copying a
+// tr1::function causes a copy to be made of all the bound parameters and
+// state.
+//
+// Furthermore, in Chromium, it is desirable for the Callback to take a
+// reference on a target object when representing a class method call.  This
+// is not supported by tr1.
+//
+// Lastly, tr1::function and tr1::bind has a more general and flexible API.
+// This includes things like argument reordering by use of
+// tr1::bind::placeholder, support for non-const reference parameters, and some
+// limited amount of subtyping of the tr1::function object (eg.,
+// tr1::function<int(int)> is convertible to tr1::function<void(int)>).
+//
+// These are not features that are required in Chromium. Some of them, such as
+// allowing for reference parameters, and subtyping of functions, may actually
+// because a source of errors. Removing support for these features actually
+// allows for a simpler implementation, and a terser Currying API.
+//
+//
+// WHY NOT GOOGLE CALLBACKS?
+//
+// The Google callback system also does not support refcounting.  Furthermore,
+// its implementation has a number of strange edge cases with respect to type
+// conversion of its arguments.  In particular, the argument's constness must
+// at times match exactly the function signature, or the type-inference might
+// break.  Given the above, writing a custom solution was easier.
+//
+//
+// MISSING FUNCTIONALITY
+//  - Invoking the return of Bind.  Bind(&foo).Run() does not work;
+//  - Binding arrays to functions that take a non-const pointer.
+//    Example:
+//      void Foo(const char* ptr);
+//      void Bar(char* ptr);
+//      Bind(&Foo, "test");
+//      Bind(&Bar, "test");  // This fails because ptr is not const.
+
+namespace base {
+
+// First, we forward declare the Callback class template. This informs the
+// compiler that the template only has 1 type parameter which is the function
+// signature that the Callback is representing.
+//
+// After this, create template specializations for 0-$(MAX_ARITY) parameters. Note that
+// even though the template typelist grows, the specialization still
+// only has one type: the function signature.
+template <typename Sig>
+class Callback;
+
+
+$range ARITY 0..MAX_ARITY
+$for ARITY [[
+$range ARG 1..ARITY
+
+$if ARITY == 0 [[
+template <typename R>
+class Callback<R(void)> {
+]] $else [[
+template <typename R, $for ARG , [[typename A$(ARG)]]>
+class Callback<R($for ARG , [[A$(ARG)]])> {
+]]
+
+ public:
+  typedef R(*PolymorphicInvoke)(internal::InvokerStorageBase*[[]]
+$if ARITY != 0 [[, ]]
+$for ARG ,
+                                [[const A$(ARG)&]]);
+
+  Callback() : polymorphic_invoke_(NULL) { }
+
+  // We pass InvokerStorageHolder by const ref to avoid incurring an
+  // unnecessary AddRef/Unref pair even though we will modify the object.
+  // We cannot use a normal reference because the compiler will warn
+  // since this is often used on a return value, which is a temporary.
+  //
+  // Note that this constructor CANNOT be explicit, and that Bind() CANNOT
+  // return the exact Callback<> type.  See base/bind.h for details.
+  template <typename T>
+  Callback(const internal::InvokerStorageHolder<T>& invoker_holder)
+      : polymorphic_invoke_(&T::FunctionTraits::DoInvoke) {
+    invoker_storage_.swap(invoker_holder.invoker_storage_);
+  }
+
+
+$if ARITY == 0 [[
+  R Run(void) const {
+]] $else [[
+  R Run($for ARG ,
+        [[const A$(ARG)& a$(ARG)]]) const {
+]]
+
+    return polymorphic_invoke_(invoker_storage_.get()[[]]
+$if ARITY != 0 [[, ]]
+$for ARG ,
+                               [[a$(ARG)]]);
+  }
+
+ private:
+  scoped_refptr<internal::InvokerStorageBase> invoker_storage_;
+  PolymorphicInvoke polymorphic_invoke_;
+};
+
+
+]]  $$ for ARITY
+
+// Syntactic sugar to make Callbacks<void(void)> easier to declare since it
+// will be used in a lot of APIs with delayed execution.
+typedef Callback<void(void)> Closure;
+
+}  // namespace base
+
+#endif  // BASE_CALLBACK_H