sort_type.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from utils import set_cffi_tmpdir, disable_stop_on_error\n",
    "set_cffi_tmpdir()\n",
    "disable_stop_on_error()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 让类型对象支持排序"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Python 3中类型不能排序，下面的代码抛出`TypeError`异常："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "ename": "TypeError",
     "evalue": "unorderable types: type() < type()",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-2-d0fcb0aa548d>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m     12\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     13\u001b[0m \u001b[0mclasses\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;33m[\u001b[0m\u001b[0mZ\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mB\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mA\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mC\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 14\u001b[1;33m \u001b[0msorted\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mclasses\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;31mTypeError\u001b[0m: unorderable types: type() < type()"
     ]
    }
   ],
   "source": [
    "class A:\n",
    "    pass\n",
    "\n",
    "class B:\n",
    "    pass\n",
    "\n",
    "class C:\n",
    "    pass\n",
    "\n",
    "class Z:\n",
    "    pass\n",
    "\n",
    "classes = [Z, B, A, C]\n",
    "sorted(classes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "我们知道对象的比较行为由其类型中的魔法方法`__lt__（）`决定。而在`type`对象的字典中没有`__lt__`："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "False"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\"__lt__\" in type.__dict__"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "下面我们定义`__lt__()`函数并将其添加进`type.__dict__`，但是抛出异常。这是由于Python对类型对象的字典进行了写保护："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "ename": "TypeError",
     "evalue": "'mappingproxy' object does not support item assignment",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-10-356183d7f355>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m      2\u001b[0m     \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m__name__\u001b[0m \u001b[1;33m<\u001b[0m \u001b[0mother\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m__name__\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      3\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 4\u001b[1;33m \u001b[0mtype\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m__dict__\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m\"__lt__\"\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0m__lt__\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;31mTypeError\u001b[0m: 'mappingproxy' object does not support item assignment"
     ]
    }
   ],
   "source": [
    "def __lt__(self, other):\n",
    "    return self.__name__ < other.__name__\n",
    "\n",
    "type.__dict__[\"__lt__\"] = __lt__"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`type.__dict__`是一个`mappingproxy`对象，它对字典对象进行只读包装："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "mappingproxy"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type(type.__dict__)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "为了往`type`的字典中添加内容，必须获得该字典的引用。可以使用`gc.get_referents()`实现。`get_referents(obj)`可以获得`obj`对象引用的所有其它对象。因此可以使用它获取`mappingproxy`对象内部包装的字典对象。下面将`__lt__()`函数添加进行`type`的字典中："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import gc\n",
    "type_dict = gc.get_referents(type.__dict__)[0]\n",
    "type_dict[\"__lt__\"] = __lt__\n",
    "\"__lt__\" in type.__dict__"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "然而即使在字典中添加了`__lt__`函数，仍然无法对类型对象进行排序。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "ename": "TypeError",
     "evalue": "unorderable types: type() < type()",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-18-563726727430>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[0msorted\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mclasses\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;31mTypeError\u001b[0m: unorderable types: type() < type()"
     ]
    }
   ],
   "source": [
    "sorted(classes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "为了让类型对象支持比较，还需要修改`type`对象对应的C结构体`PyTypeObject`的`tp_richcompare`字段。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  4           0 LOAD_FAST                0 (a)\n",
      "              3 LOAD_FAST                1 (b)\n",
      "              6 COMPARE_OP               0 (<)\n",
      "              9 RETURN_VALUE\n"
     ]
    }
   ],
   "source": [
    "import dis\n",
    "\n",
    "def f(a, b):\n",
    "    return a < b\n",
    "\n",
    "dis.dis(f)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "一个自定义的类要支持比较，必须满足以下两个条件：\n",
    "\n",
    "1. 类的`__dict__`中包含名为`__lt__`的函数。\n",
    "2. 类对应的结构体的字段`tp_richcompare`保存`slot_tp_richcompare()`函数的地址，该函数在`typeobject.c`文件中定义。\n",
    "\n",
    "前面我们已经在类的字典中添加了名为`__lt__`的函数，下面看看如何修改`tp_richcompare`字段。首先需要获取该字段的偏移地址："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "200"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import cffi\n",
    "\n",
    "ffi = cffi.FFI()\n",
    "ffi.cdef(\"\"\"\n",
    "ssize_t tp_richcompare;\n",
    "\"\"\")\n",
    "\n",
    "lib = ffi.verify(\"\"\"\n",
    "ssize_t tp_richcompare = offsetof(PyTypeObject, tp_richcompare);\n",
    "\"\"\")\n",
    "\n",
    "lib.tp_richcompare"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "由于`slot_tp_richcompare()`为`static`函数，无法直接获取其地址。因此这里定义一个`Dummy`类，其`tp_richcompare`字段中保存的就是`slot_tp_richcompare()`的地址："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class Dummy:\n",
    "    def __lt__(self, other):\n",
    "        return id(self) < id(other)\n",
    "    \n",
    "class Dummy2(Dummy):\n",
    "    pass\n",
    "\n",
    "class Dummy3:\n",
    "    def __lt__(self, other):\n",
    "        return id(self) < id(other)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "下面获取函数地址，并与`Dummy2`和`Dummy3`的字段中的值比较，它们都是同一个地址："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "addr_slot_tp_richcompare = int(ffi.cast(\"ssize_t *\", id(Dummy) + lib.tp_richcompare)[0])\n",
    "\n",
    "assert addr_slot_tp_richcompare == int(ffi.cast(\"ssize_t *\", id(Dummy2) + lib.tp_richcompare)[0])\n",
    "assert addr_slot_tp_richcompare == int(ffi.cast(\"ssize_t *\", id(Dummy3) + lib.tp_richcompare)[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "下面将`addr_slot_tp_richcompare`写入到`type`对应的结构体的`tp_richcompare`字段中："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "ffi.cast(\"ssize_t *\", id(type) + lib.tp_richcompare)[0] = addr_slot_tp_richcompare"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "于是就可以对类进行排序了："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[__main__.A, __main__.B, __main__.C, __main__.Z]"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sorted(classes)"
   ]
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [default]",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.2"
  },
  "toc": {
   "toc_cell": false,
   "toc_number_sections": true,
   "toc_threshold": 4,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}