aboutsummaryrefslogtreecommitdiffstats
path: root/fatcat_scholar/search.py
blob: 121cb69f9ed94e2f81e81dd66b19c4af652086f1 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
"""
Helpers to make elasticsearch queries.
"""

import copy
import logging
import datetime
from gettext import gettext
from typing import List, Optional, Any

import sentry_sdk
import elasticsearch
from elasticsearch_dsl import Search, Q
from elasticsearch_dsl.response import Response
import fatcat_openapi_client

# pytype: disable=import-error
from pydantic import BaseModel

# pytype: enable=import-error

from fatcat_scholar.config import settings
from fatcat_scholar.identifiers import *
from fatcat_scholar.schema import ScholarDoc
from fatcat_scholar.query_parse import sniff_citation_query, pre_parse_query
from fatcat_scholar.query_citation import try_fuzzy_match

# i18n note: the use of gettext below doesn't actually do the translation here,
# it just ensures that the strings are caught by babel for translation later


class FulltextQuery(BaseModel):
    q: Optional[str] = None
    parsed_q: Optional[str] = None
    limit: Optional[int] = None
    offset: Optional[int] = None
    filter_time: Optional[str] = None
    filter_type: Optional[str] = None
    filter_availability: Optional[str] = None
    sort_order: Optional[str] = None
    collapse_key: Optional[str] = None
    debug: Optional[bool] = False
    time_options: Any = {
        "label": gettext("Release Date"),
        "slug": "filter_time",
        "default": "all_time",
        "list": [
            {"label": gettext("All Time"), "slug": "all_time"},
            {"label": gettext("Past Week"), "slug": "past_week"},
            {"label": gettext("Past Year"), "slug": "past_year"},
            {"label": gettext("Since 2000"), "slug": "since_2000"},
            {"label": gettext("Before 1925"), "slug": "before_1925"},
        ],
    }
    type_options: Any = {
        "label": gettext("Resource Type"),
        "slug": "filter_type",
        "default": "papers",
        "list": [
            {"label": gettext("Papers"), "slug": "papers"},
            {"label": gettext("Reports"), "slug": "reports"},
            {"label": gettext("Datasets"), "slug": "datasets"},
            {"label": gettext("Everything"), "slug": "everything"},
        ],
    }
    availability_options: Any = {
        "label": gettext("Availability"),
        "slug": "filter_availability",
        "default": "fulltext",
        "list": [
            {"label": gettext("Fulltext"), "slug": "fulltext"},
            {"label": gettext("Microfilm"), "slug": "microfilm"},
            {"label": gettext("Open Access"), "slug": "oa"},
            {"label": gettext("Metadata"), "slug": "everything"},
        ],
    }
    sort_options: Any = {
        "label": gettext("Sort Order"),
        "slug": "sort_order",
        "default": "relevancy",
        "list": [
            {"label": gettext("Relevancy"), "slug": "relevancy"},
            {"label": gettext("Recent First"), "slug": "time_desc"},
            {"label": gettext("Oldest First"), "slug": "time_asc"},
        ],
    }


class FulltextHits(BaseModel):
    query_type: str
    count_returned: int
    count_found: int
    offset: int
    limit: int
    deep_page_limit: int
    query_time_ms: int
    query_wall_time_ms: int
    results: List[Any]


# global sync client connection
es_client = elasticsearch.Elasticsearch(settings.ELASTICSEARCH_QUERY_BASE, timeout=25.0)


def transform_es_results(resp: Response) -> List[dict]:
    # convert from ES objects to python dicts
    results = []
    for h in resp:
        r = h._d_
        # print(h.meta._d_)
        r["_highlights"] = []
        if "highlight" in dir(h.meta):
            highlights = h.meta.highlight._d_
            for k in highlights:
                r["_highlights"] += highlights[k]
        r["_collapsed"] = []
        r["_collapsed_count"] = 0
        if "inner_hits" in dir(h.meta):
            if isinstance(h.meta.inner_hits.more_pages.hits.total, int):
                r["_collapsed_count"] = h.meta.inner_hits.more_pages.hits.total - 1
            else:
                r["_collapsed_count"] = (
                    h.meta.inner_hits.more_pages.hits.total["value"] - 1
                )
            for k in h.meta.inner_hits.more_pages:
                if k["key"] != r["key"]:
                    r["_collapsed"].append(k)
        results.append(r)

    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")
        # ensure collapse_key is a single value, not an array
        if type(h["collapse_key"]) == list:
            h["collapse_key"] = h["collapse_key"][0]
        # add ScholarDoc object as a helper (eg, to call python helpers)
        try:
            h["_obj"] = ScholarDoc.parse_obj(h)
        except Exception:
            pass
    return results


def apply_filters(search: Search, query: FulltextQuery) -> Search:
    """
    Applies query filters to ES Search object based on query
    """
    # type filters
    if query.filter_type == "papers" or query.filter_type is None:
        search = search.filter(
            "terms", type=["article-journal", "paper-conference", "chapter", "article"]
        )
    elif query.filter_type == "reports":
        search = search.filter("terms", type=["report", "standard",])
    elif query.filter_type == "datasets":
        search = search.filter("terms", type=["dataset", "software",])
    elif query.filter_type == "everything":
        pass
    else:
        raise ValueError(
            f"Unknown 'filter_type' parameter value: '{query.filter_type}'"
        )

    # time filters
    if query.filter_time == "past_week":
        date_today = datetime.date.today()
        week_ago_date = str(date_today - datetime.timedelta(days=7))
        tomorrow_date = str(date_today + datetime.timedelta(days=1))
        search = search.filter("range", date=dict(gte=week_ago_date, lte=tomorrow_date))
    elif query.filter_time == "past_year":
        # (date in the past year) or (year is this year)
        # the later to catch papers which don't have release_date defined
        date_today = datetime.date.today()
        this_year = date_today.year
        tomorrow_date = str(date_today + datetime.timedelta(days=1))
        year_ago_date = str(date_today - datetime.timedelta(days=365))
        search = search.filter(
            Q("range", date=dict(gte=year_ago_date, lte=tomorrow_date))
            | Q("term", year=this_year)
        )
    elif query.filter_time == "since_2000":
        this_year = datetime.date.today().year
        search = search.filter("range", year=dict(gte=2000, lte=this_year))
    elif query.filter_time == "before_1925":
        search = search.filter("range", year=dict(lt=1925))
    elif query.filter_time == "all_time" or query.filter_time is None:
        pass
    else:
        raise ValueError(
            f"Unknown 'filter_time' parameter value: '{query.filter_time}'"
        )

    # availability filters
    if query.filter_availability == "oa":
        search = search.filter("term", tags="oa")
    elif query.filter_availability == "everything":
        pass
    elif query.filter_availability == "fulltext" or query.filter_availability is None:
        search = search.filter(
            "terms", **{"access.access_type": ["wayback", "ia_file", "ia_sim"]}
        )
    elif query.filter_availability == "microfilm":
        search = search.filter("term", **{"access.access_type": "ia_sim"})
    else:
        raise ValueError(
            f"Unknown 'filter_availability' parameter value: '{query.filter_availability}'"
        )

    return search


def process_query(query: FulltextQuery) -> FulltextHits:

    if not query.q:
        return do_fulltext_search(query)

    # try handling raw identifier queries
    if len(query.q.strip().split()) == 1 and not '"' in query.q:
        doi = clean_doi(query.q)
        if doi:
            return do_lookup_query(f'doi:"{doi}"')
        pmcid = clean_pmcid(query.q)
        if pmcid:
            return do_lookup_query(f'pmcid:"{pmcid}"')
        if query.q.strip().startswith("key:"):
            return do_lookup_query(query.q)

    # if this is a citation string, do a fuzzy lookup
    if settings.ENABLE_CITATION_QUERY and sniff_citation_query(query.q):
        api_conf = fatcat_openapi_client.Configuration()
        api_conf.host = settings.FATCAT_API_HOST
        api_client = fatcat_openapi_client.DefaultApi(
            fatcat_openapi_client.ApiClient(api_conf)
        )
        fatcat_es_client = elasticsearch.Elasticsearch("https://search.fatcat.wiki")
        key: Optional[str] = None
        # "best effort" fuzzy match lookup (but aggressively skip on any exception)
        try:
            key = try_fuzzy_match(
                query.q,
                grobid_host=settings.GROBID_HOST,
                es_client=fatcat_es_client,
                fatcat_api_client=api_client,
            )
        except Exception as e:
            logging.warn(f"citation fuzzy failure: {e}")
            sentry_sdk.set_level("warning")
            sentry_sdk.capture_exception(e)
            pass
        if key:
            result = do_lookup_query(f"key:{key}")
            if result:
                result.query_type = "citation"
                return result

    # fall through to regular query, with pre-parsing
    query = copy.copy(query)
    if query.q:
        query.parsed_q = pre_parse_query(query.q)

    return do_fulltext_search(query)


def do_lookup_query(lookup: str) -> FulltextHits:
    logging.info(f"lookup query: {lookup}")
    query = FulltextQuery(
        q=lookup,
        filter_type="everything",
        filter_availability="everything",
        filter_time="all_time",
    )
    result = do_fulltext_search(query)
    result.query_type = "lookup"
    return result


def do_fulltext_search(
    query: FulltextQuery, deep_page_limit: int = 2000
) -> FulltextHits:

    search = Search(using=es_client, index=settings.ELASTICSEARCH_QUERY_FULLTEXT_INDEX)

    if query.collapse_key:
        search = search.filter("term", collapse_key=query.collapse_key)
    else:
        search = search.extra(
            collapse={
                "field": "collapse_key",
                "inner_hits": {"name": "more_pages", "size": 0,},
            }
        )

    # apply filters from query
    search = apply_filters(search, query)

    # we combined several queries to improve scoring.

    # this query use the fancy built-in query string parser
    basic_fulltext = Q(
        "query_string",
        query=query.parsed_q or query.q,
        default_operator="AND",
        analyze_wildcard=True,
        allow_leading_wildcard=False,
        lenient=True,
        quote_field_suffix=".exact",
        fields=["title^4", "biblio_all^3", "everything",],
    )
    has_fulltext = Q("terms", **{"access_type": ["ia_sim", "ia_file", "wayback"]})
    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="year")),
            Q("bool", must_not=Q("exists", field="type")),
            Q("bool", must_not=Q("exists", field="stage")),
            Q("bool", must_not=Q("exists", field="biblio.container_name")),
        ],
    )

    if query.filter_availability == "fulltext" or query.filter_availability is None:
        base_query = basic_fulltext
    else:
        base_query = Q("bool", must=basic_fulltext, should=[has_fulltext])

    if query.q == "*":
        search = search.query("match_all")
        search = search.sort("_doc")
    else:
        search = search.query(
            "boosting", positive=base_query, negative=poor_metadata, negative_boost=0.5,
        )

    # simplified version of basic_fulltext query, for highlighting
    highlight_query = Q(
        "query_string",
        query=query.parsed_q or query.q,
        default_operator="AND",
        lenient=True,
    )
    search = search.highlight(
        "abstracts.body",
        "fulltext.body",
        "fulltext.acknowledgement",
        "fulltext.annex",
        highlight_query=highlight_query.to_dict(),
        require_field_match=False,
        number_of_fragments=2,
        fragment_size=200,
        order="score",
        # TODO: this will fix highlight encoding, but requires ES 7.x
        # encoder="html",
    )

    # sort order
    if query.sort_order == "time_asc":
        search = search.sort("year", "date")
    elif query.sort_order == "time_desc":
        search = search.sort("-year", "-date")
    elif query.sort_order == "relevancy" or query.sort_order is None:
        pass
    else:
        raise ValueError(f"Unknown 'sort_order' parameter value: '{query.sort_order}'")

    # Sanity checks
    limit = min((int(query.limit or 15), 100))
    offset = max((int(query.offset or 0), 0))
    if offset > deep_page_limit:
        # Avoid deep paging problem.
        offset = deep_page_limit

    search = search.params(track_total_hits=True)
    search = search[offset : (offset + limit)]

    query_start = datetime.datetime.now()
    try:
        resp = search.execute()
    except elasticsearch.exceptions.RequestError as e_raw:
        # this is a "user" error
        e: Any = e_raw
        logging.warn("elasticsearch 400: " + str(e.info))
        if e.info.get("error", {}).get("root_cause", {}):
            raise ValueError(str(e.info["error"]["root_cause"][0].get("reason"))) from e
        else:
            raise ValueError(str(e.info)) from e
    except elasticsearch.exceptions.TransportError as e:
        # all other errors
        logging.warn(f"elasticsearch non-200 status code: {e.info}")
        raise IOError(str(e.info)) from e
    query_delta = datetime.datetime.now() - query_start

    # convert from API objects to dicts
    results = transform_es_results(resp)

    count_found: int = 0
    if isinstance(resp.hits.total, int):
        count_found = int(resp.hits.total)
    else:
        count_found = int(resp.hits.total["value"])
    count_returned = len(results)

    # if we grouped to less than a page of hits, update returned count
    if (not query.collapse_key) and offset == 0 and (count_returned < limit):
        count_found = count_returned

    return FulltextHits(
        query_type="fulltext",
        count_returned=count_returned,
        count_found=count_found,
        offset=offset,
        limit=limit,
        deep_page_limit=deep_page_limit,
        query_time_ms=int(resp.took),
        query_wall_time_ms=int(query_delta.total_seconds() * 1000),
        results=results,
    )


def es_scholar_index_alive() -> bool:
    """
    Checks if the configured back-end elasticsearch index exists and can
    service queries. Intended to be used in health checks, called every couple
    seconds.

    Note that the regular client.indices.exists(index) function call will
    return an error if the cluster leader can not be reached, even if the local
    node could service queries in a read-only manner.

    The client.count(body=None, index=index) API returns quickly enough (though
    might be slow during indexing?), and returns context about the number of
    shards queried, and thus seems like a good fit for this check.
    """
    try:
        resp = es_client.count(
            body=None, index=settings.ELASTICSEARCH_QUERY_FULLTEXT_INDEX
        )
    except elasticsearch.exceptions.RequestError as e_raw:
        if e_raw.status_code == 404:
            return False
        else:
            raise e_raw
    try:
        return bool(resp["_shards"]["successful"] == resp["_shards"]["total"])
    except KeyError:
        return False


def get_es_scholar_doc(key: str) -> Optional[dict]:
    """
    Fetch a single document from search index, by key. Returns None if not found.
    """
    try:
        resp = es_client.get(settings.ELASTICSEARCH_QUERY_FULLTEXT_INDEX, key)
    except elasticsearch.exceptions.NotFoundError:
        return None
    doc = resp["_source"]
    try:
        doc["_obj"] = ScholarDoc.parse_obj(doc)
    except Exception:
        pass
    return doc