train data eda notebook

deepseek-coder-1.3b-instruct/train_data_eda.ipynb

@@ -2,12 +2,12 @@
  "cells": [
   {
    "cell_type": "code",
-   "execution_count": 6,
+   "execution_count": 18,
    "metadata": {},
    "outputs": [],
    "source": [
     "import pandas as pd\n",
-    "\n",
+    "import re\n",
     "df = pd.read_csv(\"../train-data/sql_train.tsv\", sep=\"\\t\")"
    ]
   },
@@ -31,36 +31,157 @@
     "df.columns"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## By character count"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 14,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "less_than_90 = short_queries = df[df['sql_query'].str.len() < 90]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 17,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "short_queries.to_csv(\"../train-data/less_than_90.tsv\", sep=\"\\t\", index=False)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## From to Where"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 25,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df['after_from'] = df['sql_query'].str.extract(r'FROM\\s+(\\w+)', flags=re.IGNORECASE)[0]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 26,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "0              team\n",
+       "1              game\n",
+       "2              game\n",
+       "3              game\n",
+       "4              game\n",
+       "           ...     \n",
+       "1039           game\n",
+       "1040           game\n",
+       "1041    other_stats\n",
+       "1042    other_stats\n",
+       "1043           game\n",
+       "Name: after_from, Length: 1044, dtype: object"
+      ]
+     },
+     "execution_count": 26,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "df['after_from']"
+   ]
+  },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 27,
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/plain": [
-       "count                                                  1044\n",
-       "unique                                                 1043\n",
-       "top       SELECT ROUND(AVG(pts_home),2) AS avg_home_poin...\n",
-       "freq                                                      2\n",
-       "Name: sql_query, dtype: object"
+       "array(['team', 'game', 'other_stats'], dtype=object)"
       ]
      },
-     "execution_count": 8,
+     "execution_count": 27,
      "metadata": {},
      "output_type": "execute_result"
     }
    ],
    "source": [
-    "df['sql_query'].str.len().describe()"
+    "df['after_from'].dropna().unique()\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 28,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df_game = df[df['after_from'] == 'game']\n",
+    "df_game.to_csv(\"../train-data/queries_from_game.tsv\", sep=\"\\t\", index=False)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 29,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df_game = df[df['after_from'] == 'team']\n",
+    "df_game.to_csv(\"../train-data/queries_from_team.tsv\", sep=\"\\t\", index=False)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 30,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df_game = df[df['after_from'] == 'other_stats']\n",
+    "df_game.to_csv(\"../train-data/queries_from_other_stats.tsv\", sep=\"\\t\", index=False)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Contain Join"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 31,
    "metadata": {},
    "outputs": [],
-   "source": []
+   "source": [
+    "# Queries that contain the word JOIN (case-insensitive)\n",
+    "df_with_join = df[df['sql_query'].str.contains(r'\\bJOIN\\b', case=False, na=False)]\n",
+    "df_with_join.to_csv(\"../train-data/with_join.tsv\", sep=\"\\t\", index=False)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 32,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "\n",
+    "# Queries that do NOT contain the word JOIN\n",
+    "df_without_join = df[~df['sql_query'].str.contains(r'\\bJOIN\\b', case=False, na=False)]\n",
+    "df_without_join.to_csv(\"../train-data/without_join.tsv\", sep=\"\\t\", index=False)"
+   ]
   }
  ],
  "metadata": {