{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "42382be0",
   "metadata": {},
   "source": [
    "# Callbacks: composition over inheritance"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fe1fa5ac",
   "metadata": {},
   "source": [
    "Reading:\n",
    "\n",
    "- [Python Design Patterns: The Composition Over Inheritance Principle](https://python-patterns.guide/gang-of-four/composition-over-inheritance/) by Brandon Rhodes\n",
    "- [Django Views the Right Way: Helpers vs mixins](https://spookylukey.github.io/django-views-the-right-way/common-context-data.html?highlight=mixin#discussion-helpers-vs-mixins) by Luke Plant"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "30ce2cf3",
   "metadata": {},
   "source": [
    "### Multiple inheritance (\"mixins\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "bd704487",
   "metadata": {},
   "outputs": [],
   "source": [
    "class Module:\n",
    "    def forward(self, x): return x ** 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6aaa3405",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "100"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Module().forward(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "182a19e6",
   "metadata": {},
   "outputs": [],
   "source": [
    "class Module:\n",
    "    def forward(self, x):\n",
    "        self.x = x\n",
    "        self.before_forward()\n",
    "        self.y = x ** 2\n",
    "        self.after_forward()\n",
    "        return self.y\n",
    "\n",
    "    def before_forward(self): pass\n",
    "    def after_forward(self): pass"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7ee2dabb",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "100"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Module().forward(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "08d10922",
   "metadata": {},
   "outputs": [],
   "source": [
    "class LoggingMixin:\n",
    "    def before_forward(self):\n",
    "        print(f'{self.x=}')\n",
    "        super().before_forward()\n",
    "\n",
    "    def after_forward(self):\n",
    "        print(f'{self.y=}')\n",
    "        super().after_forward()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0146c7d3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "self.x=10\n",
      "self.y=100\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "100"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "class MyModule(LoggingMixin, Module): pass\n",
    "MyModule().forward(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "de7c3626",
   "metadata": {},
   "outputs": [],
   "source": [
    "from torch import tensor\n",
    "\n",
    "class TensorMixin(Module):\n",
    "    def before_forward(self):\n",
    "        self.x = tensor(self.x)\n",
    "        super().before_forward()\n",
    "\n",
    "    def after_forward(self):\n",
    "        self.y = tensor(self.y)\n",
    "        super().after_forward()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4e028850",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "self.x=tensor(10)\n",
      "self.y=tensor(100)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "tensor(100)"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "class MyModule(TensorMixin, LoggingMixin, Module): pass\n",
    "MyModule().forward(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "61366f1c",
   "metadata": {},
   "source": [
    "### Callbacks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7be468a8",
   "metadata": {},
   "outputs": [],
   "source": [
    "class Module:\n",
    "    def __init__(self, cbs):\n",
    "        self.cbs = cbs\n",
    "        for cb in cbs: cb.mod = self\n",
    "\n",
    "    def forward(self, x):\n",
    "        self.x = x\n",
    "        self.callback('before_forward')\n",
    "        self.y = x ** 2\n",
    "        self.callback('after_forward')\n",
    "        return self.y\n",
    "\n",
    "    def callback(self, nm):\n",
    "        for cb in self.cbs: getattr(cb, nm, lambda o: None)()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9ff00775",
   "metadata": {},
   "outputs": [],
   "source": [
    "class LoggingCB:\n",
    "    def before_forward(self): print(f'{self.mod.x=}')\n",
    "    def after_forward(self): print(f'{self.mod.y=}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "484f1530",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "self.mod.x=10\n",
      "self.mod.y=100\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "100"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Module([LoggingCB()]).forward(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "246b5686",
   "metadata": {},
   "outputs": [],
   "source": [
    "class TensorCB:\n",
    "    def before_forward(self): self.mod.x = tensor(self.mod.x)\n",
    "    def after_forward(self): self.mod.y = tensor(self.mod.y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b9935e6a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "self.mod.x=tensor(10)\n",
      "self.mod.y=tensor(100)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "tensor(100)"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Module([TensorCB(), LoggingCB()]).forward(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "064b4254",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}