add spark udf

Author: khuyentran1401

Chapter5/spark.ipynb

@@ -1612,6 +1612,181 @@
     "\n",
     "new_df.show()"
    ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "afb930f8",
+   "metadata": {},
+   "source": [
+    "### Vectorized Operations in PySpark: pandas_udf vs Standard UDF"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "619e824b",
+   "metadata": {
+    "tags": [
+     "hide-cell"
+    ]
+   },
+   "outputs": [],
+   "source": [
+    "!pip install -U pyspark"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "40c80193",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from pyspark.sql import SparkSession\n",
+    "\n",
+    "# Create SparkSession\n",
+    "spark = SparkSession.builder.getOrCreate()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "94329cf7",
+   "metadata": {},
+   "source": [
+    "Standard UDF functions process data row-by-row, resulting in Python function call overhead. \n",
+    "\n",
+    "In contrast, pandas_udf utilizes Pandas' vectorized operations to process entire columns in a single operation, significantly improving performance."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "a4633f44",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "                                                                                \r"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "+----+\n",
+      "|val1|\n",
+      "+----+\n",
+      "| 1.0|\n",
+      "| 2.0|\n",
+      "| 3.0|\n",
+      "| 4.0|\n",
+      "+----+\n",
+      "\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Sample DataFrame\n",
+    "data = [(1.0,), (2.0,), (3.0,), (4.0,)]\n",
+    "df = spark.createDataFrame(data, [\"val1\"])\n",
+    "\n",
+    "df.show()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "id": "fcf0cdf9",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "                                                                                \r"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "+----+----+\n",
+      "|val1|val2|\n",
+      "+----+----+\n",
+      "| 1.0| 2.0|\n",
+      "| 2.0| 3.0|\n",
+      "| 3.0| 4.0|\n",
+      "| 4.0| 5.0|\n",
+      "+----+----+\n",
+      "\n"
+     ]
+    }
+   ],
+   "source": [
+    "from pyspark.sql.functions import udf\n",
+    "\n",
+    "# Standard UDF\n",
+    "@udf('double')\n",
+    "def plus_one(val):\n",
+    "    return val + 1\n",
+    "\n",
+    "# Apply the Standard UDF\n",
+    "df.withColumn('val2', plus_one(df.val1)).show()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "e1ec8b2b",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "                                                                                \r"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "+----+----+\n",
+      "|val1|val2|\n",
+      "+----+----+\n",
+      "| 1.0| 2.0|\n",
+      "| 2.0| 3.0|\n",
+      "| 3.0| 4.0|\n",
+      "| 4.0| 5.0|\n",
+      "+----+----+\n",
+      "\n"
+     ]
+    }
+   ],
+   "source": [
+    "from pyspark.sql.functions import pandas_udf\n",
+    "import pandas as pd\n",
+    "\n",
+    "\n",
+    "# Pandas UDF\n",
+    "@pandas_udf(\"double\")\n",
+    "def pandas_plus_one(val: pd.Series) -> pd.Series:\n",
+    "    return val + 1\n",
+    "\n",
+    "\n",
+    "# Apply the Pandas UDF\n",
+    "df.withColumn(\"val2\", pandas_plus_one(df.val1)).show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "1ab94b5b",
+   "metadata": {},
+   "source": [
+    "[Learn more about pandas_udf](https://bit.ly/4aRBNTX)."
+   ]
   }
  ],
  "metadata": {