Merge branch 'bnewbold-es-refactor' into 'master'

web ES search refactor

See merge request webgroup/fatcat!68

1 files changed, 349 insertions, 270 deletions

diff --git a/python/fatcat_web/search.py b/python/fatcat_web/search.py
index 4a87c735..55caa9c5 100644
--- a/python/fatcat_web/search.py
+++ b/python/fatcat_web/search.py
@@ -2,118 +2,257 @@
 """
 Helpers for doing elasticsearch queries (used in the web interface; not part of
 the formal API)
-
-TODO: ELASTICSEARCH_*_INDEX should probably be factored out and just hard-coded
 """
 
+import sys
 import datetime
-import requests
-from flask import abort, flash
+from dataclasses import dataclass
+from typing import List, Optional, Any
+
+import elasticsearch
+from elasticsearch_dsl import Search, Q
+import elasticsearch_dsl.response
+
 from fatcat_web import app
 
-def do_search(index, request, limit=30, offset=0, deep_page_limit=2000):
+@dataclass
+class ReleaseQuery:
+    q: Optional[str] = None
+    limit: Optional[int] = None
+    offset: Optional[int] = None
+    fulltext_only: bool = False
+    container_id: Optional[str] = None
+
+    @classmethod
+    def from_args(cls, args) -> 'ReleaseQuery':
+
+        query_str = args.get('q') or '*'
+
+        container_id = args.get('container_id')
+        # TODO: as filter, not in query string
+        if container_id:
+            query_str += ' container_id:"{}"'.format(container_id)
+
+        # TODO: where are container_issnl queries actually used?
+        issnl = args.get('container_issnl')
+        if issnl and query_str:
+            query_str += ' container_issnl:"{}"'.format(issnl)
+
+        offset = args.get('offset', '0')
+        offset = max(0, int(offset)) if offset.isnumeric() else 0
+
+        return ReleaseQuery(
+            q=query_str,
+            offset=offset,
+            fulltext_only=bool(args.get('fulltext_only')),
+            container_id=container_id,
+        )
+
+@dataclass
+class GenericQuery:
+    q: Optional[str] = None
+    limit: Optional[int] = None
+    offset: Optional[int] = None
+
+    @classmethod
+    def from_args(cls, args) -> 'GenericQuery':
+        query_str = args.get('q')
+        if not query_str:
+            query_str = '*'
+        offset = args.get('offset', '0')
+        offset = max(0, int(offset)) if offset.isnumeric() else 0
+
+        return GenericQuery(
+            q=query_str,
+            offset=offset,
+        )
+
+@dataclass
+class SearchHits:
+    count_returned: int
+    count_found: int
+    offset: int
+    limit: int
+    deep_page_limit: int
+    query_time_ms: int
+    results: List[Any]
+
+
+def results_to_dict(response: elasticsearch_dsl.response.Response) -> List[dict]:
+    """
+    Takes a response returns all the hits as JSON objects.
 
-    # Sanity checks
-    if limit > 100:
-        limit = 100
-    if offset < 0:
-        offset = 0
-    if offset > deep_page_limit:
-        # Avoid deep paging problem.
-        offset = deep_page_limit
+    Also handles surrogate strings that elasticsearch returns sometimes,
+    probably due to mangled data processing in some pipeline. "Crimes against
+    Unicode"; production workaround
+    """
+
+    results = []
+    for h in response:
+        r = h._d_
+        # print(h.meta._d_)
+        results.append(r)
 
-    request["size"] = int(limit)
-    request["from"] = int(offset)
-    # print(request)
-    resp = requests.get("%s/%s/_search" %
-            (app.config['ELASTICSEARCH_BACKEND'], index),
-        json=request)
-
-    if resp.status_code == 400:
-        print("elasticsearch 400: " + str(resp.content))
-        flash("Search query failed to parse; you might need to use quotes.<p><code>{}</code>".format(resp.content))
-        abort(resp.status_code)
-    elif resp.status_code != 200:
-        print("elasticsearch non-200 status code: " + str(resp.status_code))
-        print(resp.content)
-        abort(resp.status_code)
-
-    content = resp.json()
-    results = [h['_source'] for h in content['hits']['hits']]
     for h in results:
-        # Handle surrogate strings that elasticsearch returns sometimes,
-        # probably due to mangled data processing in some pipeline.
-        # "Crimes against Unicode"; production workaround
         for key in h:
             if type(h[key]) is str:
-                h[key] = h[key].encode('utf8', 'ignore').decode('utf8')
+                h[key] = h[key].encode("utf8", "ignore").decode("utf8")
+    return results
 
-    return {"count_returned": len(results),
-            "count_found": content['hits']['total'],
-            "results": results,
-            "offset": offset,
-            "deep_page_limit": deep_page_limit}
+def wrap_es_execution(search: Search) -> Any:
+    """
+    Executes a Search object, and converts various ES error types into
+    something we can pretty print to the user.
+    """
+    try:
+        resp = search.execute()
+    except elasticsearch.exceptions.RequestError as e:
+        # this is a "user" error
+        print("elasticsearch 400: " + str(e.info), file=sys.stderr)
+        if e.info.get("error", {}).get("root_cause", {}):
+            raise ValueError(str(e.info["error"]["root_cause"][0].get("reason")))
+        else:
+            raise ValueError(str(e.info))
+    except elasticsearch.exceptions.TransportError as e:
+        # all other errors
+        print("elasticsearch non-200 status code: {}".format(e.info), file=sys.stderr)
+        raise IOError(str(e.info))
+    return resp
 
+def do_container_search(
+    query: GenericQuery, deep_page_limit: int = 2000
+) -> SearchHits:
 
-def do_release_search(q, limit=30, fulltext_only=True, offset=0):
+    search = Search(using=app.es_client, index=app.config['ELASTICSEARCH_CONTAINER_INDEX'])
 
-    #print("Search hit: " + q)
-    if limit > 100:
-        # Sanity check
-        limit = 100
+    search = search.query(
+        "query_string",
+        query=query.q,
+        default_operator="AND",
+        analyze_wildcard=True,
+        allow_leading_wildcard=False,
+        lenient=True,
+        fields=["biblio"],
+    )
 
-    # Convert raw DOIs to DOI queries
-    if len(q.split()) == 1 and q.startswith("10.") and q.count("/") >= 1:
-        q = 'doi:"{}"'.format(q)
+    # Sanity checks
+    limit = min((int(query.limit or 25), 100))
+    offset = max((int(query.offset or 0), 0))
+    if offset > deep_page_limit:
+        # Avoid deep paging problem.
+        offset = deep_page_limit
 
-    if fulltext_only:
-        q += " in_web:true"
+    search = search[offset : (offset + limit)]
+
+    resp = wrap_es_execution(search)
+    results = results_to_dict(resp)
+
+    return SearchHits(
+        count_returned=len(results),
+        count_found=int(resp.hits.total),
+        offset=offset,
+        limit=limit,
+        deep_page_limit=deep_page_limit,
+        query_time_ms=int(resp.took),
+        results=results,
+    )
+
+def do_release_search(
+    query: ReleaseQuery, deep_page_limit: int = 2000
+) -> SearchHits:
+
+    search = Search(using=app.es_client, index=app.config['ELASTICSEARCH_RELEASE_INDEX'])
+
+    # availability filters
+    if query.fulltext_only:
+        search = search.filter("term", in_ia=True)
+
+    # Below, we combine several queries to improve scoring.
+
+    # this query use the fancy built-in query string parser
+    basic_biblio = Q(
+        "query_string",
+        query=query.q,
+        default_operator="AND",
+        analyze_wildcard=True,
+        allow_leading_wildcard=False,
+        lenient=True,
+        fields=[
+            "title^2",
+            "biblio",
+        ],
+    )
+    has_fulltext = Q("term", in_ia=True)
+    poor_metadata = Q(
+        "bool",
+        should=[
+            # if these fields aren't set, metadata is poor. The more that do
+            # not exist, the stronger the signal.
+            Q("bool", must_not=Q("exists", field="title")),
+            Q("bool", must_not=Q("exists", field="release_year")),
+            Q("bool", must_not=Q("exists", field="release_type")),
+            Q("bool", must_not=Q("exists", field="release_stage")),
+        ],
+    )
 
-    search_request = {
-        "query": {
-            "query_string": {
-                "query": q,
-                "default_operator": "AND",
-                "analyze_wildcard": True,
-                "lenient": True,
-                "fields": ["biblio"],
-            },
-        },
-    }
+    search = search.query(
+        "boosting",
+        positive=Q("bool", must=basic_biblio, should=[has_fulltext],),
+        negative=poor_metadata,
+        negative_boost=0.5,
+    )
+
+    # Sanity checks
+    limit = min((int(query.limit or 25), 100))
+    offset = max((int(query.offset or 0), 0))
+    if offset > deep_page_limit:
+        # Avoid deep paging problem.
+        offset = deep_page_limit
 
-    resp = do_search(app.config['ELASTICSEARCH_RELEASE_INDEX'], search_request, offset=offset)
-    for h in resp['results']:
+    search = search[offset : (offset + limit)]
+
+    resp = wrap_es_execution(search)
+    results = results_to_dict(resp)
+
+    for h in results:
         # Ensure 'contrib_names' is a list, not a single string
         if type(h['contrib_names']) is not list:
             h['contrib_names'] = [h['contrib_names'], ]
         h['contrib_names'] = [name.encode('utf8', 'ignore').decode('utf8') for name in h['contrib_names']]
-    resp["query"] = { "q": q }
-    resp["limit"] = limit
-    return resp
 
+    return SearchHits(
+        count_returned=len(results),
+        count_found=int(resp.hits.total),
+        offset=offset,
+        limit=limit,
+        deep_page_limit=deep_page_limit,
+        query_time_ms=int(resp.took),
+        results=results,
+    )
 
-def do_container_search(q, limit=30, offset=0):
+def get_elastic_container_random_releases(ident, limit=5):
+    """
+    Returns a list of releases from the container.
+    """
 
-    # Convert raw ISSN-L to ISSN-L query
-    if len(q.split()) == 1 and len(q) == 9 and q[0:4].isdigit() and q[4] == '-':
-        q = 'issnl:"{}"'.format(q)
+    assert limit > 0 and limit <= 100
 
-    search_request = {
-        "query": {
-            "query_string": {
-                "query": q,
-                "default_operator": "AND",
-                "analyze_wildcard": True,
-                "lenient": True,
-                "fields": ["biblio"],
-            },
-        },
-    }
+    search = Search(using=app.es_client, index=app.config['ELASTICSEARCH_RELEASE_INDEX'])
+    search = search.query(
+        'bool',
+        must=[
+            Q('term', container_id=ident),
+            Q('range', release_year={ "lte": datetime.datetime.today().year }),
+        ]
+    )
+    search = search.sort('-in_web', '-release_date')
+    search = search[:int(limit)]
 
-    resp = do_search(app.config['ELASTICSEARCH_CONTAINER_INDEX'], search_request, limit=limit, offset=offset)
-    resp["query"] = { "q": q }
-    resp["limit"] = limit
-    return resp
+    search = search.params(request_cache=True)
+    resp = wrap_es_execution(search)
+    results = results_to_dict(resp)
+
+    return results
 
 def get_elastic_entity_stats():
     """
@@ -127,85 +266,73 @@ def get_elastic_entity_stats():
 
     stats = {}
 
-    # 2. releases
-    #  - total count
-    #  - total citation records
-    #  - total (paper, chapter, proceeding)
-    #  - " with fulltext on web
-    #  - " open access
-    #  - " not in KBART, in IA
-    #
-    # Can do the above with two queries:
-    #  - all releases, aggregate count and sum(ref_count)
-    #  - in-scope works, aggregate count by (fulltext, OA, kbart/ia)
-
-    # 2a. release totals
-    query = {
-        "size": 0,
-        "aggs": {
-            "release_ref_count": { "sum": { "field": "ref_count" } }
-        }
-    }
-    resp = requests.get(
-        "{}/fatcat_release/_search".format(app.config['ELASTICSEARCH_BACKEND']),
-        json=query,
-        params=dict(request_cache="true"))
-    # TODO: abort()
-    resp.raise_for_status()
-    resp = resp.json()
+    # release totals
+    search = Search(using=app.es_client, index=app.config['ELASTICSEARCH_RELEASE_INDEX'])
+    search.aggs.bucket(
+        'release_ref_count',
+        'sum',
+        field='ref_count',
+    )
+    search = search[:0]  # pylint: disable=unsubscriptable-object
+
+    search = search.params(request_cache=True)
+    resp = wrap_es_execution(search)
+
     stats['release'] = {
-        "total": resp['hits']['total'],
-        "refs_total": int(resp['aggregations']['release_ref_count']['value']),
+        "total": int(resp.hits.total),
+        "refs_total": int(resp.aggregations.release_ref_count.value),
     }
 
-    # 2b. paper counts
-    query = {
-        "size": 0,
-        "query": {
-            "terms": { "release_type": [
-                # "chapter", "thesis",
-                "article-journal", "paper-conference",
-            ] } },
-        "aggs": { "paper_like": { "filters": { "filters": {
-                "in_web": { "term": { "in_web": "true" } },
-                "is_oa": { "term": { "is_oa": "true" } },
-                "in_kbart": { "term": { "in_kbart": "true" } },
-                "in_web_not_kbart": { "bool": { "filter": [
-                        { "term": { "in_web": "true" } },
-                        { "term": { "in_kbart": "false" } }
-                ]}}
-        }}}}
-    }
-    resp = requests.get(
-        "{}/fatcat_release/_search".format(app.config['ELASTICSEARCH_BACKEND']),
-        json=query,
-        params=dict(request_cache="true"))
-    # TODO: abort()
-    resp.raise_for_status()
-    resp = resp.json()
-    buckets = resp['aggregations']['paper_like']['buckets']
+    # paper counts
+    search = Search(using=app.es_client, index=app.config['ELASTICSEARCH_RELEASE_INDEX'])
+    search = search.query(
+        'terms',
+        release_type=[
+            "article-journal",
+            "paper-conference",
+            # "chapter",
+            # "thesis",
+        ],
+    )
+    search.aggs.bucket(
+        'paper_like',
+        'filters',
+        filters={
+            "in_web": { "term": { "in_web": "true" } },
+            "is_oa": { "term": { "is_oa": "true" } },
+            "in_kbart": { "term": { "in_kbart": "true" } },
+            "in_web_not_kbart": { "bool": { "filter": [
+                { "term": { "in_web": "true" } },
+                { "term": { "in_kbart": "false" } },
+            ]}},
+        }
+    )
+    search = search[:0]
+
+    search = search.params(request_cache=True)
+    resp = wrap_es_execution(search)
+    buckets = resp.aggregations.paper_like.buckets
     stats['papers'] = {
-        'total': resp['hits']['total'],
-        'in_web': buckets['in_web']['doc_count'],
-        'is_oa': buckets['is_oa']['doc_count'],
-        'in_kbart': buckets['in_kbart']['doc_count'],
-        'in_web_not_kbart': buckets['in_web_not_kbart']['doc_count'],
+        'total': resp.hits.total,
+        'in_web': buckets.in_web.doc_count,
+        'is_oa': buckets.is_oa.doc_count,
+        'in_kbart': buckets.in_kbart.doc_count,
+        'in_web_not_kbart': buckets.in_web_not_kbart.doc_count,
     }
 
-    # 3. containers
-    #   => total count
-    query = {
-        "size": 0,
-    }
-    resp = requests.get(
-        "{}/fatcat_container/_search".format(app.config['ELASTICSEARCH_BACKEND']),
-        json=query,
-        params=dict(request_cache="true"))
-    # TODO: abort()
-    resp.raise_for_status()
-    resp = resp.json()
+    # container counts
+    search = Search(using=app.es_client, index=app.config['ELASTICSEARCH_CONTAINER_INDEX'])
+    search.aggs.bucket(
+        'release_ref_count',
+        'sum',
+        field='ref_count',
+    )
+    search = search[:0]  # pylint: disable=unsubscriptable-object
+
+    search = search.params(request_cache=True)
+    resp = wrap_es_execution(search)
     stats['container'] = {
-        "total": resp['hits']['total'],
+        "total": resp.hits.total,
     }
 
     return stats
@@ -221,30 +348,36 @@ def get_elastic_container_stats(ident, issnl=None):
         preserved
     """
 
-    query = {
-        "size": 0,
-        "query": {
-            "term": { "container_id": ident }
+    search = Search(using=app.es_client, index=app.config['ELASTICSEARCH_RELEASE_INDEX'])
+    search = search.query(
+        'term',
+        container_id=ident,
+    )
+    search.aggs.bucket(
+        'container_stats',
+        'filters',
+        filters={
+            "in_web": {
+                "term": { "in_web": True },
+            },
+            "in_kbart": {
+                "term": { "in_kbart": True },
+            },
+            "is_preserved": {
+                "term": { "is_preserved": True },
+            },
         },
-        "aggs": { "container_stats": { "filters": { "filters": {
-                "in_web": { "term": { "in_web": "true" } },
-                "in_kbart": { "term": { "in_kbart": "true" } },
-                "is_preserved": { "term": { "is_preserved": "true" } },
-        }}}}
-    }
-    resp = requests.get(
-        "{}/fatcat_release/_search".format(app.config['ELASTICSEARCH_BACKEND']),
-        json=query,
-        params=dict(request_cache="true"))
-    # TODO: abort()
-    #print(resp.json())
-    resp.raise_for_status()
-    resp = resp.json()
-    buckets = resp['aggregations']['container_stats']['buckets']
+    )
+    search = search[:0]
+
+    search = search.params(request_cache=True)
+    resp = wrap_es_execution(search)
+
+    buckets = resp.aggregations.container_stats.buckets
     stats = {
         'ident': ident,
         'issnl': issnl,
-        'total': resp['hits']['total'],
+        'total': resp.hits.total,
         'in_web': buckets['in_web']['doc_count'],
         'in_kbart': buckets['in_kbart']['doc_count'],
         'is_preserved': buckets['is_preserved']['doc_count'],
@@ -252,48 +385,6 @@ def get_elastic_container_stats(ident, issnl=None):
 
     return stats
 
-def get_elastic_container_random_releases(ident, limit=5):
-    """
-    Returns a list of releases from the container.
-    """
-
-    assert limit > 0 and limit <= 100
-
-    query = {
-        "size": int(limit),
-        "sort": [
-            { "in_web": {"order": "desc"} },
-            { "release_date": {"order": "desc"} },
-        ],
-        "query": {
-            "bool": {
-                "must": [
-                    { "term": { "container_id": ident } },
-                    { "range": { "release_year": { "lte": datetime.datetime.today().year } } },
-                ],
-            },
-        },
-    }
-    resp = requests.get(
-        "{}/fatcat_release/_search".format(app.config['ELASTICSEARCH_BACKEND']),
-        json=query,
-        params=dict(request_cache="true"))
-    # TODO: abort()
-    #print(resp.json())
-    resp.raise_for_status()
-    resp = resp.json()
-    #print(resp)
-    hits = [h['_source'] for h in resp['hits']['hits']]
-    for h in hits:
-        # Handle surrogate strings that elasticsearch returns sometimes,
-        # probably due to mangled data processing in some pipeline.
-        # "Crimes against Unicode"; production workaround
-        for key in h:
-            if type(h[key]) is str:
-                h[key] = h[key].encode('utf8', 'ignore').decode('utf8')
-
-    return hits
-
 def get_elastic_container_histogram(ident):
     """
     Fetches a stacked histogram
@@ -304,58 +395,46 @@ def get_elastic_container_histogram(ident):
         (year, in_ia, count)
     """
 
-    query = {
-        "aggs": {
-            "year_in_ia": {
-                "composite": {
-                    "size": 1000,
-                    "sources": [
-                        {"year": {
-                            "histogram": {
-                                "field": "release_year",
-                                "interval": 1,
-                        }}},
-                        {"in_ia": {
-                            "terms": {
-                                "field": "in_ia",
-                        }}},
-                    ],
+    search = Search(using=app.es_client, index=app.config['ELASTICSEARCH_RELEASE_INDEX'])
+    search = search.query(
+        'bool',
+        must=[
+            Q("range", release_year={
+                "gte": datetime.datetime.today().year - 499,
+                "lte": datetime.datetime.today().year,
+            }),
+        ],
+        filter=[
+            Q("bool", minimum_should_match=1, should=[
+                Q("match", container_id=ident),
+            ]),
+        ],
+    )
+    search.aggs.bucket(
+        'year_in_ia',
+        'composite',
+        size=1000,
+        sources=[
+            {"year": {
+                "histogram": {
+                    "field": "release_year",
+                    "interval": 1,
                 },
-            },
-        },
-        "size": 0,
-        "query": {
-            "bool": {
-                "must": [{
-                    "range": {
-                        "release_year": {
-                            "gte": datetime.datetime.today().year - 499,
-                            "lte": datetime.datetime.today().year,
-                        }
-                    }
-                }],
-                "filter": [{
-                    "bool": {
-                        "should": [{
-                            "match": {
-                                "container_id": ident
-                            }
-                        }],
-                        "minimum_should_match": 1,
-                    },
-                }],
-            }
-        }
-    }
-    resp = requests.get(
-        "{}/fatcat_release/_search".format(app.config['ELASTICSEARCH_BACKEND']),
-        json=query,
-        params=dict(request_cache="true"))
-    resp.raise_for_status()
-    # TODO: abort()
-    resp = resp.json()
-    #print(resp)
+            }},
+            {"in_ia": {
+                "terms": {
+                    "field": "in_ia",
+                },
+            }},
+        ],
+    )
+    search = search[:0]
+
+    search = search.params(request_cache='true')
+    resp = wrap_es_execution(search)
+
+    buckets = resp.aggregations.year_in_ia.buckets
     vals = [(h['key']['year'], h['key']['in_ia'], h['doc_count'])
-            for h in resp['aggregations']['year_in_ia']['buckets']]
+            for h in buckets]
     vals = sorted(vals)
     return vals