Merge pull request #50 from qqueue/elasticsearchin

elasticsearch-based search (WIP)

import_to_es.py

@@ -0,0 +1,94 @@
+#!/usr/bin/env python
+"""
+Bulk load torents from mysql into elasticsearch `nyaav2` index,
+which is assumed to already exist.
+This is a one-shot deal, so you'd either need to complement it
+with a cron job or some binlog-reading thing (TODO)
+"""
+from nyaa.models import Torrent
+from elasticsearch import Elasticsearch
+from elasticsearch import helpers
+import progressbar
+import sys
+
+bar = progressbar.ProgressBar(
+        max_value=Torrent.query.count(),
+        widgets=[
+            progressbar.SimpleProgress(),
+            ' [', progressbar.Timer(), '] ',
+            progressbar.Bar(),
+            ' (', progressbar.ETA(), ') ',
+            ])
+
+es = Elasticsearch()
+
+# turn into thing that elasticsearch indexes. We flatten in
+# the stats (seeders/leechers) so we can order by them in es naturally.
+# we _don't_ dereference uploader_id to the user's display name however,
+# instead doing that at query time. I _think_ this is right because
+# we don't want to reindex all the user's torrents just because they
+# changed their name, and we don't really want to FTS search on the user anyway.
+# Maybe it's more convenient to derefence though.
+def mk_es(t):
+    return {
+        "_id": t.id,
+        "_type": "torrent",
+        "_index": "nyaav2",
+        "_source": {
+            # we're also indexing the id as a number so you can
+            # order by it. seems like this is just equivalent to
+            # order by created_time, but oh well
+            "id": t.id,
+            "display_name": t.display_name,
+            "created_time": t.created_time,
+            "updated_time": t.updated_time,
+            "description": t.description,
+            # not analyzed but included so we can render magnet links
+            # without querying sql again.
+            "info_hash": t.info_hash.hex(),
+            "filesize": t.filesize,
+            "uploader_id": t.uploader_id,
+            "main_category_id": t.main_category_id,
+            "sub_category_id": t.sub_category_id,
+            # XXX all the bitflags are numbers
+            "anonymous": bool(t.anonymous),
+            "trusted": bool(t.trusted),
+            "remake": bool(t.remake),
+            "complete": bool(t.complete),
+            # TODO instead of indexing and filtering later
+            # could delete from es entirely. Probably won't matter
+            # for at least a few months.
+            "hidden": bool(t.hidden),
+            "deleted": bool(t.deleted),
+            "has_torrent": t.has_torrent,
+            # XXX last_updated isn't initialized
+            "stats_last_updated": t.stats.last_updated or t.created_time,
+            "download_count": t.stats.download_count,
+            "leech_count": t.stats.leech_count,
+            "seed_count": t.stats.seed_count,
+        }
+    }
+
+# page through an sqlalchemy query, like the per_fetch but
+# doesn't break the eager joins its doing against the stats table.
+# annoying that this isn't built in somehow.
+def page_query(query, limit=sys.maxsize, batch_size=10000):
+    start = 0
+    while True:
+        # XXX very inelegant way to do this, i'm confus
+        stop = min(limit, start + batch_size)
+        if stop == start:
+            break
+        things = query.slice(start, stop)
+        if not things:
+            break
+        had_things = False
+        for thing in things:
+            had_things = True
+            yield(thing)
+        if not had_things or stop == limit:
+            break
+        bar.update(start)
+        start = min(limit, start + batch_size)
+
+helpers.bulk(es, (mk_es(t) for t in page_query(Torrent.query)), chunk_size=10000)

nyaa/routes.py

@@ -27,6 +27,11 @@ from email.mime.multipart import MIMEMultipart
 from email.mime.text import MIMEText
 from email.utils import formatdate
 
+from elasticsearch import Elasticsearch
+from elasticsearch_dsl import Search, Q
+
+es_client = Elasticsearch()
+
 DEBUG_API = False
 
 
@@ -67,6 +72,16 @@ def search(term='', user=None, sort='id', order='desc', category='0_0', quality_
     sort_ = sort.lower()
     if sort_ not in sort_keys:
         flask.abort(400)
+
+    # XXX gross why are all the names subtly different
+    es_sort = ({
+        'id': 'id',
+        'size': 'filesize',
+        'name': 'display_name',
+        'seeders': 'seed_count',
+        'leechers': 'leech_count',
+        'downloads': 'download_count'
+    })[sort]
     sort = sort_keys[sort]
 
     order_keys = {
@@ -78,6 +93,10 @@ def search(term='', user=None, sort='id', order='desc', category='0_0', quality_
     if order_ not in order_keys:
         flask.abort(400)
 
+    # funky, es sort is default asc, prefixed by '-' if desc
+    if "desc" == order:
+        es_sort = "-" + es_sort
+
     filter_keys = {
         '0': None,
         '1': (models.TorrentFlags.REMAKE, False),
@@ -126,28 +145,37 @@ def search(term='', user=None, sort='id', order='desc', category='0_0', quality_
     if flask.g.user:
         same_user = flask.g.user.id == user
 
+    s = Search(using=es_client, index='nyaav2')
     if term:
         query = db.session.query(models.TorrentNameSearch)
+        s = s.query("query_string", default_field="display_name", default_operator="AND", query=term)
     else:
         query = models.Torrent.query
 
     # Filter by user
     if user:
+        s = s.filter("term", uploader_id=user)
+
         query = query.filter(models.Torrent.uploader_id == user)
         # If admin, show everything
         if not admin:
             # If user is not logged in or the accessed feed doesn't belong to user,
             # hide anonymous torrents belonging to the queried user
             if not same_user:
+                # TODO adapt to es syntax
                 query = query.filter(models.Torrent.flags.op('&')(
                     int(models.TorrentFlags.ANONYMOUS | models.TorrentFlags.DELETED)).is_(False))
 
     if main_category:
+        s = s.filter("term", main_category_id=main_cat_id)
         query = query.filter(models.Torrent.main_category_id == main_cat_id)
     elif sub_category:
+        s = s.filter("term", main_category_id=main_cat_id)
+        s = s.filter("term", sub_category_id=sub_cat_id)
         query = query.filter((models.Torrent.main_category_id == main_cat_id) &
                              (models.Torrent.sub_category_id == sub_cat_id))
 
+    # TODO i dunno what this means in es
     if filter_tuple:
         query = query.filter(models.Torrent.flags.op('&')(int(filter_tuple[0])).is_(filter_tuple[1]))
 
@@ -157,6 +185,7 @@ def search(term='', user=None, sort='id', order='desc', category='0_0', quality_
             int(models.TorrentFlags.HIDDEN | models.TorrentFlags.DELETED)).is_(False))
 
     if term:
+        # note already handled in es
         for item in shlex.split(term, posix=False):
             if len(item) >= 2:
                 query = query.filter(FullTextSearch(
@@ -166,14 +195,25 @@ def search(term='', user=None, sort='id', order='desc', category='0_0', quality_
     if sort.class_ != models.Torrent:
         query = query.join(sort.class_)
 
+    s = s.sort(es_sort)
     query = query.order_by(getattr(sort, order)())
 
+    per = app.config['RESULTS_PER_PAGE']
     if rss:
-        query = query.limit(app.config['RESULTS_PER_PAGE'])
+        pass
+        #query = query.limit(app.config['RESULTS_PER_PAGE'])
     else:
-        query = query.paginate_faste(page, per_page=app.config['RESULTS_PER_PAGE'], step=5)
+        # page is 1-based?
+        s = s[(page-1)*per:page*per]
+        #query = query.paginate_faste(page, per_page=app.config['RESULTS_PER_PAGE'], step=5)
 
-    return query
+    s = s.highlight_options(tags_schema='styled')
+    s = s.highlight("display_name")
+
+    #return query
+    from pprint import pprint
+    print(json.dumps(s.to_dict()))
+    return s.execute()
 
 
 @app.errorhandler(404)
@@ -445,6 +485,7 @@ def activate_user(payload):
 
     user.status = models.UserStatusType.ACTIVE
 
+
     db.session.add(user)
     db.session.commit()
 

nyaa/static/css/main.css

@@ -97,4 +97,14 @@ table.torrent-list thead th.sorting_desc:after {
 	    margin-left: 20px;
 	    margin-bottom: 10px;
 	}
-}
+}
+
+/* elasticsearch term highlight */
+.hlt1 {
+    font-style: normal;
+    display: inline-block;
+    padding: 0 3px;
+    border-radius: 3px;
+    border: 1px solid rgba(100, 56, 0, 0.8);
+    background: rgba(200,127,0,0.3);
+}

nyaa/templates/search_results.html

@@ -8,7 +8,7 @@
 	{{ caller() }}
 </th>
 {% endmacro %}
-{% if torrent_query.items %}
+{% if torrent_query.hits.total > 0 %}
 <div class="table-responsive">
 	<table class="table table-bordered table-hover table-striped torrent-list">
 		<thead>
@@ -45,26 +45,26 @@
 			</tr>
 		</thead>
 		<tbody>
-			{% for torrent in torrent_query.items %}
+			{% for torrent in torrent_query %}
 			<tr class="{% if torrent.deleted %}deleted{% elif torrent.hidden %}warning{% elif torrent.remake %}danger{% elif torrent.trusted %}success{% else %}default{% endif %}">
-				{% set cat_id = (torrent.main_category.id|string) + '_' + (torrent.sub_category.id|string) %}
+				{% set cat_id = (torrent.main_category_id|string) + '_' + (torrent.sub_category_id|string) %}
 				{% set icon_dir = config.SITE_FLAVOR %}
 				<td style="padding:0 4px;">
-				<a href="/?c={{ cat_id }}" title="{{ torrent.main_category.name }} - {{ torrent.sub_category.name }}">
+				<a href="/?c={{ cat_id }}" title="{{ torrent.main_category_id }} - {{ torrent.sub_category_id }}">
 					<img src="/static/img/icons/{{ icon_dir }}/{{ cat_id }}.png">
 				</a>
 				</td>
-				<td><a href="{{ url_for('view_torrent', torrent_id=torrent.id) }}">{{ torrent.display_name | escape }}</a></td>
+				<td><a href="{{ url_for('view_torrent', torrent_id=torrent.meta.id) }}">{{ torrent.meta.highlight.display_name[0] | safe }}</a></td>
 				<td style="white-space: nowrap;text-align: center;">
 					{% if torrent.has_torrent %}<a href="{{ url_for('download_torrent', torrent_id=torrent.id) }}"><i class="fa fa-fw fa-download"></i></a>{% endif %}
 					<a href="{{ torrent.magnet_uri }}"><i class="fa fa-fw fa-magnet"></i></a>
 				</td>
 				<td class="text-center">{{ torrent.filesize | filesizeformat(True) }}</td>
-				<td class="text-center" data-timestamp="{{ torrent.created_utc_timestamp|int }}">{{ torrent.created_time.strftime('%Y-%m-%d %H:%M') }}</td>
+                <td class="text-center" {#data-timestamp="{{ torrent.created_time|int }}"#}>{{ torrent.created_time }}</td>
 				{% if config.ENABLE_SHOW_STATS %}
-				<td class="text-center" style="color: green;">{{ torrent.stats.seed_count }}</td>
+				<td class="text-center" style="color: green;">{{ torrent.seed_count }}</td>
-				<td class="text-center" style="color: red;">{{ torrent.stats.leech_count }}</td>
+				<td class="text-center" style="color: red;">{{ torrent.leech_count }}</td>
-				<td class="text-center">{{ torrent.stats.download_count }}</td>
+				<td class="text-center">{{ torrent.download_count }}</td>
 				{% endif %}
 			</tr>
 			{% endfor %}
@@ -75,7 +75,9 @@
 <h3>No results found</h3>
 {% endif %}
 
+{#
 <center>
 	{% from "bootstrap/pagination.html" import render_pagination %}
 	{{ render_pagination(torrent_query) }}
 </center>
+#}

requirements.txt

@@ -24,11 +24,33 @@ pycodestyle==2.3.1
 pycparser==2.17
 pyparsing==2.2.0
 six==1.10.0
-SQLAlchemy>=1.1.9
+SQLAlchemy==1.1.9
 SQLAlchemy-FullText-Search==0.2.3
-SQLAlchemy-Utils>=0.32.14
+SQLAlchemy-Utils==0.32.14
 uWSGI==2.0.15
 visitor==0.1.3
 webassets==0.12.1
 Werkzeug==0.12.1
 WTForms==2.1
+## The following requirements were added by pip freeze:
+decorator==4.0.11
+elasticsearch==5.3.0
+elasticsearch-dsl==5.2.0
+ipython==6.0.0
+ipython-genutils==0.2.0
+jedi==0.10.2
+mysql-replication==0.13
+pexpect==4.2.1
+pickleshare==0.7.4
+pkg-resources==0.0.0
+progressbar2==3.20.0
+prompt-toolkit==1.0.14
+ptyprocess==0.5.1
+Pygments==2.2.0
+PyMySQL==0.7.11
+python-dateutil==2.6.0
+python-utils==2.1.0
+simplegeneric==0.8.1
+traitlets==4.3.2
+urllib3==1.21.1
+wcwidth==0.1.7