{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "86b10564",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "def print_info(a):\n",
    "    \"\"\" Print the content of an array, and its metadata. \"\"\"\n",
    "    \n",
    "    txt = f\"\"\"\n",
    "dtype\\t{a.dtype}\n",
    "ndim\\t{a.ndim}\n",
    "shape\\t{a.shape}\n",
    "strides\\t{a.strides}\n",
    "    \"\"\"\n",
    "\n",
    "    print(a)\n",
    "    print(txt)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "53bd92f9",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 0  1  2  3]\n",
      " [ 4  5  6  7]\n",
      " [ 8  9 10 11]]\n",
      "\n",
      "dtype\tint64\n",
      "ndim\t2\n",
      "shape\t(3, 4)\n",
      "strides\t(32, 8)\n",
      "    \n"
     ]
    }
   ],
   "source": [
    "x = np.arange(12).reshape(3, 4).copy()\n",
    "print_info(x)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d2ee43d7",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Views"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f4838e77",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "Operations that only require changing the metadata always do so, and return a **view**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "f1b82845",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 1  3]\n",
      " [ 9 11]]\n",
      "\n",
      "dtype\tint64\n",
      "ndim\t2\n",
      "shape\t(2, 2)\n",
      "strides\t(64, 16)\n",
      "    \n"
     ]
    }
   ],
   "source": [
    "# slice\n",
    "y = x[0::2, 1::2]\n",
    "print_info(y)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3199b45b",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "A view shares the same memory block as the original array. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "28ea1c71",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 0  1  2  3  4  5  6  7  8  9 10 11]]\n",
      "\n",
      "dtype\tint64\n",
      "ndim\t2\n",
      "shape\t(1, 12)\n",
      "strides\t(96, 8)\n",
      "    \n"
     ]
    }
   ],
   "source": [
    "z = x.reshape(1, 12)\n",
    "print_info(z)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d88fbf5d",
   "metadata": {},
   "source": [
    "CAREFUL: Modifying the view **changes the original array** and all other views of that array as well!"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7f35dcc3",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "##### in place operations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "4ff516cc",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 1  3]\n",
      " [ 9 11]]\n"
     ]
    }
   ],
   "source": [
    "print(y) # a view of x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "46822b5a",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[101 103]\n",
      " [109 111]]\n",
      "\n",
      "dtype\tint64\n",
      "ndim\t2\n",
      "shape\t(2, 2)\n",
      "strides\t(64, 16)\n",
      "    \n"
     ]
    }
   ],
   "source": [
    "y += 100 \n",
    "print_info(y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "ad9a7950",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[  0 101   2 103]\n",
      " [  4   5   6   7]\n",
      " [  8 109  10 111]]\n",
      "\n",
      "dtype\tint64\n",
      "ndim\t2\n",
      "shape\t(3, 4)\n",
      "strides\t(32, 8)\n",
      "    \n",
      "[[  0 101   2 103   4   5   6   7   8 109  10 111]]\n",
      "\n",
      "dtype\tint64\n",
      "ndim\t2\n",
      "shape\t(1, 12)\n",
      "strides\t(96, 8)\n",
      "    \n"
     ]
    }
   ],
   "source": [
    "print_info(x)\n",
    "print_info(z)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4fc789c1",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "Functions that take an array as an input should **avoid modifying it in place!**\n",
    "\n",
    "Always make a copy or be super extra clear in the docstring."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "aa25ac4b",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [],
   "source": [
    "def robust_log(x, cte = 3):\n",
    "    \"\"\" \n",
    "    Returns the log of an array, deals with values that are equal to 0.\n",
    "\n",
    "    `x` is expected to have non-negative values.\n",
    "    \"\"\"\n",
    "    x[x == 0] += cte\n",
    "    return np.log(x)\n",
    "    \n",
    "# this is not being very clear"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "471d9d6b",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [],
   "source": [
    "a = np.array([[96, 0.01], [0, 1]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "6c05d356",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0. 1.]\n",
      "\n",
      "dtype\tfloat64\n",
      "ndim\t1\n",
      "shape\t(2,)\n",
      "strides\t(8,)\n",
      "    \n"
     ]
    }
   ],
   "source": [
    "# a view of `a`\n",
    "b = a[1, :]\n",
    "print_info(b)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "9d96fb61",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 4.56434819, -4.60517019],\n",
       "       [ 1.09861229,  0.        ]])"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "robust_log(a)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "35d0327d",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([3., 1.])"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.set_printoptions(suppress=True)\n",
    "b"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fa8cf77a",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "Better to make a copy!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "c5359eac",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [],
   "source": [
    "def robust_log(x, cte = 3):\n",
    "    \"\"\" Returns the log of an array, deals with values that are 0.\n",
    "\n",
    "    `x` is expected to have non-negative values.\n",
    "    \"\"\"\n",
    "    x = x.copy()\n",
    "    x[x == 0] += cte\n",
    "    return np.log(x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "0bf9b2d5",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0. 1.]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[ 4.56434819, -4.60517019],\n",
       "       [ 1.09861229,  0.        ]])"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a = np.array([[96, 0.01], [0, 1]])\n",
    "b = a[1, :]\n",
    "print(b)\n",
    "robust_log(a)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "895209ce",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0., 1.])"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "b"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d664b462",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Copies\n",
    "\n",
    "- Operations that cannot be executed by changing the metadata create a new memory block, and return a **copy**\n",
    "- can be forced by method .copy()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "716aec53",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Choosing row, columns, or individual elements of an array by giving explicitly their indices (a.k.a \"fancy indexing\") it's an operation that in general cannot be executed by changing the metadata alone.\n",
    "\n",
    "Therefore, **fancy indexing always returns a copy**."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "fbcf3100",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 0  1  2  3]\n",
      " [ 4  5  6  7]\n",
      " [ 8  9 10 11]]\n",
      "\n",
      "dtype\tint64\n",
      "ndim\t2\n",
      "shape\t(3, 4)\n",
      "strides\t(32, 8)\n",
      "    \n"
     ]
    }
   ],
   "source": [
    "x = np.arange(12).reshape(3, 4).copy()\n",
    "print_info(x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "6c50e46e",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 1  0 11]\n",
      "\n",
      "dtype\tint64\n",
      "ndim\t1\n",
      "shape\t(3,)\n",
      "strides\t(8,)\n",
      "    \n"
     ]
    }
   ],
   "source": [
    "#print(x)\n",
    "z = x[[0, 0, 2], [1, 0, 3]]\n",
    "# what's z equal to?\n",
    "\n",
    "print_info(z)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "9d65a5c3",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1001 1000 1011]\n",
      "\n",
      "dtype\tint64\n",
      "ndim\t1\n",
      "shape\t(3,)\n",
      "strides\t(8,)\n",
      "    \n",
      "[[ 0  1  2  3]\n",
      " [ 4  5  6  7]\n",
      " [ 8  9 10 11]]\n",
      "\n",
      "dtype\tint64\n",
      "ndim\t2\n",
      "shape\t(3, 4)\n",
      "strides\t(32, 8)\n",
      "    \n"
     ]
    }
   ],
   "source": [
    "z += 1000\n",
    "print_info(z)\n",
    "\n",
    "# the original array is unchanged => not a view!\n",
    "print_info(x)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "25aa99a4",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "**Views** are created, when you use other strides to read your data. Slicing and regular indexing allows that, as you know how many byte steps you need to take to get the data.\n",
    "\n",
    "**Fancy indexing** does not allow that, because the data you are asking **cannot** be obtained by just changing the strides. Thus, numpy needs to create a **copy** of it in memory."
   ]
  }
 ],
 "metadata": {
  "celltoolbar": "Slideshow",
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}