{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import gzip\n",
    "import random\n",
    "from pprint import pprint\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Reading the dataset ...\n",
      "Adding http://www.jstor.org/stable/10.1086/491498 to sample\n",
      "Adding http://www.jstor.org/stable/10.1086/432295 to sample\n",
      "Adding http://www.jstor.org/stable/10.1086/379413 to sample\n",
      "Adding http://www.jstor.org/stable/228664 to sample\n",
      "Adding http://www.jstor.org/stable/236768 to sample\n",
      "Adding http://www.jstor.org/stable/227706 to sample\n",
      "Adding http://www.jstor.org/stable/231357 to sample\n",
      "Adding http://www.jstor.org/stable/3080697 to sample\n",
      "Adding http://www.jstor.org/stable/229231 to sample\n",
      "Adding http://www.jstor.org/stable/230556 to sample\n",
      "Adding http://www.jstor.org/stable/10.1086/670902 to sample\n",
      "Adding http://www.jstor.org/stable/228263 to sample\n",
      "Adding http://www.jstor.org/stable/229843 to sample\n",
      "Adding http://www.jstor.org/stable/10.1086/678012 to sample\n",
      "Adding http://www.jstor.org/stable/230061 to sample\n",
      "Adding http://www.jstor.org/stable/10.1086/376025 to sample\n",
      "Adding http://www.jstor.org/stable/10.1086/653929 to sample\n",
      "Adding http://www.jstor.org/stable/226119 to sample\n",
      "Adding http://www.jstor.org/stable/10.1086/491505 to sample\n",
      "Adding http://www.jstor.org/stable/235887 to sample\n",
      "Adding http://www.jstor.org/stable/10.1086/682793 to sample\n",
      "Adding http://www.jstor.org/stable/227572 to sample\n",
      "Adding http://www.jstor.org/stable/10.1086/386402 to sample\n",
      "Adding http://www.jstor.org/stable/223695 to sample\n",
      "Adding http://www.jstor.org/stable/235969 to sample\n",
      "Dataset reading complete. 25 total documents.\n"
     ]
    }
   ],
   "source": [
    "sample_doc_numbers = random.sample(range(0, 19000), 25)\n",
    "sample_docs = []\n",
    "\n",
    "print(\"Reading the dataset ...\")\n",
    "\n",
    "with gzip.open(\"./datasets/dset1.jsonl.gz\", \"rb\") as inf:\n",
    "    for row_num, row in enumerate(inf):\n",
    "        doc = json.loads(row)\n",
    "        if row_num not in sample_doc_numbers:\n",
    "            continue\n",
    "        print(f\"Adding {doc['id']} to sample\")\n",
    "        sample_docs.append(doc)\n",
    "\n",
    "print(f\"Dataset reading complete. {len(sample_docs)} total documents.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {},
   "outputs": [],
   "source": [
    "doc1 = sample_docs[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "metadata": {},
   "outputs": [],
   "source": [
    "to_delete = [\"unigramCount\", \"bigramCount\", \"trigramCount\", \"fullText\"]\n",
    "for k in to_delete:\n",
    "    if k in doc1.keys():\n",
    "        del doc1[k]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'creator': ['Joan‐Pau Rubiés'],\n",
      " 'datePublished': '2005-06-01',\n",
      " 'docType': 'article',\n",
      " 'id': 'http://www.jstor.org/stable/10.1086/491498',\n",
      " 'identifier': [{'name': 'issn', 'value': '00211753'},\n",
      "                {'name': 'oclc', 'value': '49976319'},\n",
      "                {'name': 'lccn', 'value': '2002-227035'},\n",
      "                {'name': 'local_uuid',\n",
      "                 'value': 'd22c16bb-d068-3bdf-9962-8d0db608891e'},\n",
      "                {'name': 'local_doi', 'value': '10.1086/491498'},\n",
      "                {'name': 'journal_id', 'value': 'isis'}],\n",
      " 'isPartOf': 'Isis',\n",
      " 'issueNumber': '2',\n",
      " 'language': ['eng'],\n",
      " 'outputFormat': ['unigram', 'bigram', 'trigram'],\n",
      " 'pageCount': 2,\n",
      " 'pageEnd': '276',\n",
      " 'pageStart': '275',\n",
      " 'pagination': 'pp. 275-276',\n",
      " 'provider': 'jstor',\n",
      " 'publicationYear': 2005,\n",
      " 'publisher': 'The University of Chicago Press',\n",
      " 'sourceCategory': ['History of Science & Technology', 'History'],\n",
      " 'title': 'Review Article',\n",
      " 'url': 'http://www.jstor.org/stable/10.1086/491498',\n",
      " 'volumeNumber': '96',\n",
      " 'wordCount': 1051}\n"
     ]
    }
   ],
   "source": [
    "pprint(doc1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {},
   "outputs": [],
   "source": [
    "fields_to_keep = [\n",
    "    \"id\",\n",
    "    \"title\",\n",
    "    \"isPartOf\",\n",
    "    \"publicationYear\",\n",
    "    \"creator\",\n",
    "    \"wordCount\",\n",
    "    \"provider\",\n",
    "    \"url\"\n",
    "]\n",
    "filtered_sample_docs = []\n",
    "for doc in sample_docs:\n",
    "    new_doc = {}\n",
    "    for f in fields_to_keep:\n",
    "        value = doc.get(f)\n",
    "        new_doc[f] = value\n",
    "    filtered_sample_docs.append(new_doc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'creator': ['Joan‐Pau Rubiés'],\n",
      " 'id': 'http://www.jstor.org/stable/10.1086/491498',\n",
      " 'isPartOf': 'Isis',\n",
      " 'provider': 'jstor',\n",
      " 'publicationYear': 2005,\n",
      " 'title': 'Review Article',\n",
      " 'url': 'http://www.jstor.org/stable/10.1086/491498',\n",
      " 'wordCount': 1051}\n"
     ]
    }
   ],
   "source": [
    "pprint(filtered_sample_docs[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(\"datasets/filtered_dset1.json\", \"w\") as of:\n",
    "    json.dump(filtered_sample_docs, of)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.6.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}