encyclopedia.py 50.8 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
# Copyright 2018 Markus Scheidgen
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an"AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

15
"""
16
The encyclopedia API of the nomad@FAIRDI APIs.
17
"""
18
import re
19
import math
20
import numpy as np
21

22
23
from flask_restplus import Resource, abort, fields, marshal
from flask import request
24
from elasticsearch_dsl import Search, Q, A
25
from elasticsearch_dsl.utils import AttrDict
26

27
from nomad import config, files
28
from nomad.units import ureg
Lauri Himanen's avatar
Lauri Himanen committed
29
from nomad.atomutils import get_hill_decomposition
30
from nomad.datamodel.datamodel import EntryArchive
31
from .api import api
32
from .common import enable_gzip
33

34
ns = api.namespace("encyclopedia", description="Access encyclopedia metadata.")
35
36
re_formula = re.compile(r"([A-Z][a-z]?)(\d*)")

37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
material_prop_map = {
    # General
    "material_id": "encyclopedia.material.material_id",
    "formula": "encyclopedia.material.formula",
    "formula_reduced": "encyclopedia.material.formula_reduced",
    "system_type": "encyclopedia.material.material_type",
    # Bulk only
    "has_free_wyckoff_parameters": "encyclopedia.material.bulk.has_free_wyckoff_parameters",
    "strukturbericht_designation": "encyclopedia.material.bulk.strukturbericht_designation",
    "material_name": "encyclopedia.material.material_name",
    "bravais_lattice": "encyclopedia.material.bulk.bravais_lattice",
    "crystal_system": "encyclopedia.material.bulk.crystal_system",
    "point_group": "encyclopedia.material.bulk.point_group",
    "space_group_number": "encyclopedia.material.bulk.space_group_number",
    "space_group_international_short_symbol": "encyclopedia.material.bulk.space_group_international_short_symbol",
    "structure_prototype": "encyclopedia.material.bulk.structure_prototype",
    "structure_type": "encyclopedia.material.bulk.structure_type",
}
55
56


57
58
59
60
61
62
63
64
65
66
67
68
def rgetattr(obj, attr_name):
    """Used to perform attribute access based on a (possibly nested) attribute
    name given as string.
    """
    try:
        for attr in attr_name.split("."):
            obj = obj[attr]
    except KeyError:
        return None
    return obj


69
def get_es_doc_values(es_doc, mapping, keys=None):
70
71
    """Used to form a material definition for "materials/<material_id>" from
    the given ElasticSearch root document.
72
    """
73
74
75
    if keys is None:
        keys = mapping.keys()

76
    result = {}
77
    for key in keys:
78
        es_key = mapping[key]
79
        value = rgetattr(es_doc, es_key)
80
        result[key] = value
81
82
83
84

    return result


85
86
87
88
89
90
91
92
93
94
95
def get_enc_filter():
    """Returns a shared term filter that will leave out unpublished, embargoed
    or invalid entries.
    """
    return [
        Q("term", published=True),
        Q("term", with_embargo=False),
        Q("term", encyclopedia__status="success"),
    ]


96
material_query = api.parser()
97
98
99
100
101
102
103
104
material_query.add_argument(
    "property",
    type=str,
    choices=tuple(material_prop_map.keys()),
    help="Optional single property to retrieve for the given material. If not specified, all properties will be returned.",
    location="args"
)
material_result = api.model("material_result", {
105
106
    # General
    "material_id": fields.String,
107
108
    "formula": fields.String,
    "formula_reduced": fields.String,
109
    "system_type": fields.String,
110
    "n_matches": fields.Integer,
111
    # Bulk only
112
    "has_free_wyckoff_parameters": fields.Boolean,
113
    "strukturbericht_designation": fields.String,
114
    "material_name": fields.String,
115
116
    "bravais_lattice": fields.String,
    "crystal_system": fields.String,
117
    "point_group": fields.String,
118
119
120
    "space_group_number": fields.Integer,
    "space_group_international_short_symbol": fields.String,
    "structure_prototype": fields.String,
121
122
    "structure_type": fields.String,
})
123
124


125
@ns.route("/materials/<string:material_id>")
126
class EncMaterialResource(Resource):
127
128
129
    @api.response(404, "The material does not exist")
    @api.response(200, "Metadata send", fields.Raw)
    @api.doc("material/<material_id>")
130
    @api.expect(material_query)
131
    @api.marshal_with(material_result, skip_none=True)
132
    def get(self, material_id):
133
134
        """Used to retrieve basic information related to the specified
        material.
135
        """
136
137
138
139
        # Parse request arguments
        args = material_query.parse_args()
        prop = args.get("property", None)
        if prop is not None:
140
141
            keys = [prop]
            es_keys = [material_prop_map[prop]]
142
143
        else:
            keys = list(material_prop_map.keys())
144
            es_keys = list(material_prop_map.values())
145

146
147
148
149
150
        # Find the first public entry with this material id and take
        # information from there. In principle all other entries should have
        # the same information.
        s = Search(index=config.elastic.index_name)
        query = Q(
151
            "bool",
152
            filter=get_enc_filter() + [
153
                Q("term", encyclopedia__material__material_id=material_id),
154
155
156
            ]
        )
        s = s.query(query)
157

158
        # Only one representative entry is returned by collapsing the results.
159
160
        s = s.extra(**{
            "_source": {"includes": es_keys},
161
            "size": 1,
162
163
            "collapse": {"field": "encyclopedia.material.material_id"},
        })
164
165
        response = s.execute()

166
        # No such material
167
        if len(response) == 0:
168
            abort(404, message="There is no material {}".format(material_id))
169

170
        # Add values from ES entry
171
        entry = response[0]
172
        result = get_es_doc_values(entry, material_prop_map, keys)
173

174
175
176
        return result, 200


177
range_query = api.model("range_query", {
178
179
180
    "max": fields.Float,
    "min": fields.Float,
})
181
182
183
materials_after = api.model("materials_after", {
    "materials": fields.String,
})
184
185
materials_query = api.model("materials_input", {
    "search_by": fields.Nested(api.model("search_query", {
186
187
        "exclusive": fields.Boolean(default=False),
        "formula": fields.String,
Lauri Himanen's avatar
Lauri Himanen committed
188
        "element": fields.String,
189
        "page": fields.Integer(default=1),
190
        "after": fields.Nested(materials_after, allow_null=True),
191
192
193
        "per_page": fields.Integer(default=25),
        "pagination": fields.Boolean,
    })),
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
    "material_name": fields.List(fields.String),
    "structure_type": fields.List(fields.String),
    "space_group_number": fields.List(fields.Integer),
    "system_type": fields.List(fields.String),
    "crystal_system": fields.List(fields.String),
    "band_gap": fields.Nested(range_query, description="Band gap range in eV."),
    "band_gap_direct": fields.Boolean,
    "has_band_structure": fields.Boolean,
    "has_dos": fields.Boolean,
    "has_fermi_surface": fields.Boolean,
    "has_thermal_properties": fields.Boolean,
    "functional_type": fields.List(fields.String),
    "basis_set_type": fields.List(fields.String),
    "code_name": fields.List(fields.String),
    "mass_density": fields.Nested(range_query, description="Mass density range in kg / m ** 3."),
209
})
210
211
212
213
214
pages_result = api.model("page_info", {
    "per_page": fields.Integer,
    "total": fields.Integer,
    "page": fields.Integer,
    "pages": fields.Integer,
215
    "after": fields.Nested(materials_after),
216
217
})

218
219
materials_result = api.model("materials_result", {
    "total_results": fields.Integer(allow_null=False),
220
221
    "results": fields.List(fields.Nested(material_result, skip_none=True)),
    "pages": fields.Nested(pages_result, skip_none=True),
222
    "es_query": fields.String(allow_null=False),
223
224
225
})


226
@ns.route("/materials")
227
class EncMaterialsResource(Resource):
228
229
230
    @api.response(404, "No materials found")
    @api.response(400, "Bad request")
    @api.response(200, "Metadata send", fields.Raw)
231
    @api.expect(materials_query, validate=False)
232
    @api.marshal_with(materials_result, skip_none=True)
233
    @api.doc("materials")
234
235
236
237
238
239
240
241
242
    def post(self):
        """Used to query a list of materials with the given search options.
        """
        # Get query parameters as json
        try:
            data = marshal(request.get_json(), materials_query)
        except Exception as e:
            abort(400, message=str(e))

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
        # The queries that correspond to AND queries typically need to access
        # multiple calculations at once to find the material ids that
        # correspond to the query. To implement this behaviour we need to run
        # an initial aggregation that checks that the requested properties are
        # present for a material. This is a a very crude solution that does not
        # scale to complex queries, but I'm not sure we can do much better
        # until we have a separate index for materials.
        property_map = {
            "has_thermal_properties": "encyclopedia.properties.thermodynamical_properties",
            "has_band_structure": "encyclopedia.properties.electronic_band_structure",
            "has_dos": "encyclopedia.properties.electronic_dos",
            "has_fermi_surface": "encyclopedia.properties.fermi_surface",
        }
        requested_properties = []
        # The size is set very large because all the results need to be
        # returned. We cannot get the results in a paginated way with composite
        # aggregation, because pipeline aggregations are not compatible with
        # them.
        agg_parent = A("terms", field="encyclopedia.material.material_id", size=5000000)
        for key, value in property_map.items():
            if data[key] is True:
                agg = A("filter", exists={"field": value})
                agg_parent.bucket(key, agg)
                requested_properties.append(key)
        if len(requested_properties) > 1:
            bool_query = Q(
                "bool",
                filter=get_enc_filter(),
            )
            s = Search(index=config.elastic.index_name)
            s = s.query(bool_query)
            s.aggs.bucket("materials", agg_parent)
            buckets_path = {x: "{}._count".format(x) for x in requested_properties}
            script = " && ".join(["params.{} > 0".format(x) for x in requested_properties])
            agg_parent.pipeline("selector", A(
                "bucket_selector",
                buckets_path=buckets_path,
                script=script,
            ))
            s = s.extra(**{
                "size": 0,
            })
            response = s.execute()
            material_ids = [x["key"] for x in response.aggs.materials.buckets]
            if len(material_ids) == 0:
                abort(404, message="No materials found for the given search criteria or pagination.")

        # After finding the material ids that fill the AND conditions, continue
        # with a simple OR query.
292
        filters = get_enc_filter()
293
294
295
296
        must_nots = []
        musts = []

        def add_terms_filter(source, target, query_type="terms"):
297
            if data[source] is not None:
298
299
                filters.append(Q(query_type, **{target: data[source]}))

300
301
        if len(requested_properties) > 1:
            filters.append(Q("terms", encyclopedia__material__material_id=material_ids))
302
303
        add_terms_filter("material_name", "encyclopedia.material.material_name")
        add_terms_filter("structure_type", "encyclopedia.material.bulk.structure_type")
304
        add_terms_filter("space_group_number", "encyclopedia.material.bulk.space_group_number")
305
306
307
308
309
310
311
        add_terms_filter("system_type", "encyclopedia.material.material_type")
        add_terms_filter("crystal_system", "encyclopedia.material.bulk.crystal_system")
        add_terms_filter("band_gap_direct", "encyclopedia.properties.band_gap_direct", query_type="term")
        add_terms_filter("functional_type", "encyclopedia.method.functional_type")
        add_terms_filter("basis_set_type", "dft.basis_set")
        add_terms_filter("code_name", "dft.code_name")

312
313
        # Add exists filters if only one property was requested. The initial
        # aggregation will handlei multiple simultaneous properties.
314
315
316
317
318
319
320
321
        def add_exists_filter(source, target):
            param = data[source]
            if param is not None:
                query = Q("exists", field=target)
                if param is True:
                    filters.append(query)
                elif param is False:
                    must_nots.append(query)
322
323
324
        if len(requested_properties) == 1:
            prop_name = requested_properties[0]
            add_exists_filter(prop_name, property_map[prop_name])
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347

        # Add range filters
        def add_range_filter(source, target, source_unit=None, target_unit=None):
            param = data[source]
            query_dict = {}
            if param["min"] is not None:
                if source_unit is None and target_unit is None:
                    gte = param["min"]
                else:
                    gte = (param["min"] * source_unit).to(target_unit).magnitude
                query_dict["gte"] = gte
            if param["max"] is not None:
                if source_unit is None and target_unit is None:
                    lte = param["max"]
                else:
                    lte = (param["max"] * source_unit).to(target_unit).magnitude
                query_dict["lte"] = lte
            if len(query_dict) != 0:
                query = Q("range", **{target: query_dict})
                filters.append(query)

        add_range_filter("band_gap", "encyclopedia.properties.band_gap", ureg.eV, ureg.J)
        add_range_filter("mass_density", "encyclopedia.properties.mass_density")
348

349
350
351
        # Create query for elements or formula
        search_by = data["search_by"]
        formula = search_by["formula"]
Lauri Himanen's avatar
Lauri Himanen committed
352
        elements = search_by["element"]
353
354
355
        exclusive = search_by["exclusive"]

        if formula is not None:
Lauri Himanen's avatar
Lauri Himanen committed
356
357
            # Here we determine a list of atom types. The types may occur
            # multiple times and at multiple places.
358
359
360
361
362
363
364
365
366
367
            element_list = []
            matches = re_formula.finditer(formula)
            for match in matches:
                groups = match.groups()
                symbol = groups[0]
                count = groups[1]
                if symbol != "":
                    if count == "":
                        element_list.append(symbol)
                    else:
Lauri Himanen's avatar
Lauri Himanen committed
368
369
370
371
372
                        element_list += [symbol] * int(count)

            # The given list of species is reformatted with the Hill system
            # into a query string. The counts are reduced by the greatest
            # common divisor.
373
            names, reduced_counts = get_hill_decomposition(element_list, reduced=True)
Lauri Himanen's avatar
Lauri Himanen committed
374
375
376
377
378
            query_string = []
            for name, count in zip(names, reduced_counts):
                if count == 1:
                    query_string.append(name)
                else:
379
                    query_string.append("{}{}".format(name, int(count)))
Lauri Himanen's avatar
Lauri Himanen committed
380
            query_string = " ".join(query_string)
381
382
383

            # With exclusive search we look for exact match
            if exclusive:
Lauri Himanen's avatar
Lauri Himanen committed
384
                filters.append(Q("term", **{"encyclopedia.material.species_and_counts.keyword": query_string}))
385
386
387
388
389
            # With non-exclusive search we look for match that includes at
            # least all parts of the formula, possibly even more.
            else:
                musts.append(Q(
                    "match",
Lauri Himanen's avatar
Lauri Himanen committed
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
                    encyclopedia__material__species_and_counts={"query": query_string, "operator": "and"}
                ))
        elif elements is not None:
            # The given list of species is reformatted with the Hill system into a query string
            species, _ = get_hill_decomposition(elements.split(","))
            query_string = " ".join(species)

            # With exclusive search we look for exact match
            if exclusive:
                filters.append(Q("term", **{"encyclopedia.material.species.keyword": query_string}))
            # With non-exclusive search we look for match that includes at
            # least all species, possibly even more.
            else:
                musts.append(Q(
                    "match",
                    encyclopedia__material__species={"query": query_string, "operator": "and"}
406
407
                ))

408
409
        page = search_by["page"]
        per_page = search_by["per_page"]
410
        after = search_by["after"]
411
        bool_query = Q(
412
            "bool",
413
414
415
416
            filter=filters,
            must_not=must_nots,
            must=musts,
        )
417

418
419
420
421
        # The top query filters out entries based on the user query
        s = Search(index=config.elastic.index_name)
        s = s.query(bool_query)

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
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
        # 1: The paginated approach: No way to know the amount of materials,
        # but can return aggregation results in a quick fashion including
        # the number of calculation entries per material.
        mode = "collapse"
        if mode == "aggregation":
            # The materials are grouped by using three aggregations:
            # "Composite" to enable scrolling, "Terms" to enable selecting
            # by material_id and "Top Hits" to fetch a single
            # representative material document. Unnecessary fields are
            # filtered to reduce data transfer.
            terms_agg = A("terms", field="encyclopedia.material.material_id")
            composite_kwargs = {"sources": {"materials": terms_agg}, "size": per_page}

            # The number of matched materials is only requested on the first
            # search, not for each page.
            if after is not None:
                composite_kwargs["after"] = after
            else:
                cardinality_agg = A("cardinality", field="encyclopedia.material.material_id", precision_threshold=1000)
                s.aggs.metric("n_materials", cardinality_agg)

            composite_agg = A("composite", **composite_kwargs)
            composite_agg.metric("representative", A(
                "top_hits",
                size=1,
                _source={"includes": list(material_prop_map.values())},
            ))
            s.aggs.bucket("materials", composite_agg)

            # We ignore the top level hits and sort by reduced material formula.
            s = s.extra(**{
                "size": 0,
            })

            response = s.execute()
            materials = response.aggs.materials.buckets
            if len(materials) == 0:
                abort(404, message="No materials found for the given search criteria or pagination.")
            after_new = response.aggs.materials["after_key"]

            # Gather results from aggregations
            result_list = []
            materials = response.aggs.materials.buckets
            keys = list(material_prop_map.keys())
            for material in materials:
                representative = material["representative"][0]
                mat_dict = get_es_doc_values(representative, material_prop_map, keys)
                mat_dict["n_matches"] = material.doc_count
                result_list.append(mat_dict)

            # Page information is incomplete for aggregations
            pages = {
                "page": page,
                "per_page": per_page,
                "after": after_new,
            }
            if after is None:
                n_materials = response.aggs.n_materials.value
                pages["total"] = n_materials

        # 2. Collapse approach. Quickly provides a list of materials
483
484
        # corresponding to the query, offers full pagination, the number of
        # matches per material needs to be requested with a separate query.
485
486
487
488
489
        elif mode == "collapse":
            s = Search(index=config.elastic.index_name)
            s = s.query(bool_query)

            # Add cardinality aggregation that gives out the total number of materials
490
            cardinality_agg = A("cardinality", field="encyclopedia.material.material_id", precision_threshold=1000)
491
            s.aggs.metric("n_materials", cardinality_agg)
492

493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
            s = s.extra(**{
                "collapse": {"field": "encyclopedia.material.material_id"},
                "size": per_page,
                "from": (page - 1) * per_page,
                "sort": [{"encyclopedia.material.formula_reduced": {"order": "asc"}}],
                "explain": True,
            })

            # Execute query
            response = s.execute()

            # No matches
            if len(response) == 0:
                abort(404, message="No materials found for the given search criteria or pagination.")

            # Gather number of entries per material with a separate query
            material_ids = [x.encyclopedia.material.material_id for x in response]
            s = Search(index=config.elastic.index_name)
            bool_query = Q(
                "bool",
                filter=Q("terms", encyclopedia__material__material_id=material_ids),
            )
            s2 = s.query(bool_query)
            s2.aggs.bucket("n_matches", A("terms", field="encyclopedia.material.material_id"))
            response2 = s2.execute()
            matmap = {x.key: x.doc_count for x in response2.aggs.n_matches}
519

520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
            # Loop over materials
            result_list = []
            keys = list(material_prop_map.keys())
            for material in response:
                # Get values from the collapsed doc
                mat_result = get_es_doc_values(material, material_prop_map, keys)
                mat_id = material.encyclopedia.material.material_id
                mat_result["n_matches"] = matmap[mat_id]
                result_list.append(mat_result)

            # Full page information available for collapse
            pages = {
                "page": page,
                "per_page": per_page,
                "pages": math.ceil(response.hits.total / per_page),
                "total": response.aggs.n_materials.value,
            }
537
538
539

        result = {
            "results": result_list,
540
            "pages": pages,
541
        }
542
        return result, 200
543
544


545
groups_result = api.model("groups_result", {
546
547
    "groups_eos": fields.Raw,
    "groups_par": fields.Raw,
548
549
550
})


551
@ns.route("/materials/<string:material_id>/groups")
Lauri Himanen's avatar
Lauri Himanen committed
552
class EncGroupsResource(Resource):
553
554
555
    @api.response(404, "Material not found")
    @api.response(400, "Bad request")
    @api.response(200, "Metadata send", fields.Raw)
556
    @api.marshal_with(groups_result)
557
    @api.doc("enc_materials")
Lauri Himanen's avatar
Lauri Himanen committed
558
    def get(self, material_id):
559
560
561
        """Returns a summary of the calculation groups that were identified for
        this material.
        """
562
563
564
        # Find entries for the given material, which have EOS or parameter
        # variation hashes set.
        bool_query = Q(
565
            "bool",
566
            filter=get_enc_filter() + [Q("term", encyclopedia__material__material_id=material_id)],
567
568
            must=[
                Q("exists", field="encyclopedia.properties.energies.energy_total"),
569
                Q("exists", field="encyclopedia.material.idealized_structure.cell_volume"),
570
571
            ],
            should=[
572
573
                Q("exists", field="encyclopedia.method.group_eos_id"),
                Q("exists", field="encyclopedia.method.group_parametervariation_id"),
574
575
576
            ],
            minimum_should_match=1,  # At least one of the should query must match
        )
Lauri Himanen's avatar
Lauri Himanen committed
577
578

        s = Search(index=config.elastic.index_name)
579
580
581
582
        s = s.query(bool_query)

        # Bucket the calculations by the group hashes. Only create a bucket if an
        # above-minimum number of documents are found.
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
        group_eos_bucket = A("terms", field="encyclopedia.method.group_eos_id", min_doc_count=4)
        group_param_bucket = A("terms", field="encyclopedia.method.group_parametervariation_id", min_doc_count=2)
        calc_aggregation = A(
            "top_hits",
            _source={"includes": ["calc_id"]},
            sort=[{"encyclopedia.properties.energies.energy_total": {"order": "asc"}}],
            size=100,
        )
        group_eos_bucket.bucket("calculations", calc_aggregation)
        group_param_bucket.bucket("calculations", calc_aggregation)
        s.aggs.bucket("groups_eos", group_eos_bucket)
        s.aggs.bucket("groups_param", group_param_bucket)

        # We ignore the top level hits
        s = s.extra(**{
            "size": 0,
        })

        # Collect information for each group from the aggregations
        response = s.execute()
        groups_eos = {group.key: [calc.calc_id for calc in group.calculations.hits] for group in response.aggs.groups_eos.buckets}
        groups_param = {group.key: [calc.calc_id for calc in group.calculations.hits] for group in response.aggs.groups_param.buckets}

        # Return results
        result = {
            "groups_eos": groups_eos,
            "groups_par": groups_param,
        }

        return result, 200


group_result = api.model("group_result", {
    "calculations": fields.List(fields.String),
    "energies": fields.List(fields.Float),
    "volumes": fields.List(fields.Float),
})
group_source = {
    "includes": [
        "calc_id",
        "encyclopedia.properties.energies.energy_total",
        "encyclopedia.material.idealized_structure.cell_volume",
    ]
}


@ns.route("/materials/<string:material_id>/groups/<string:group_type>/<string:group_id>")
class EncGroupResource(Resource):
    @api.response(404, "Group not found")
    @api.response(400, "Bad request")
    @api.response(200, "Metadata send", fields.Raw)
    @api.marshal_with(group_result)
    @api.doc("enc_group")
    def get(self, material_id, group_type, group_id):
        """Used to query detailed information for a specific calculation group.
        """
        # Find entries for the given material, which have EOS or parameter
        # variation hashes set.
        if group_type == "eos":
            group_id_source = "encyclopedia.method.group_eos_id"
        elif group_type == "par":
            group_id_source = "encyclopedia.method.group_parametervariation_id"
        else:
            abort(400, message="Unsupported group type.")

        bool_query = Q(
            "bool",
650
            filter=get_enc_filter() + [
651
652
653
654
655
656
657
                Q("term", encyclopedia__material__material_id=material_id),
                Q("term", **{group_id_source: group_id}),
            ],
        )

        s = Search(index=config.elastic.index_name)
        s = s.query(bool_query)
658
659
660
661
662

        # calc_id and energy should be extracted for each matched document. The
        # documents are sorted by energy so that the minimum energy one can be
        # easily extracted. A maximum request size is set in order to limit the
        # result size. ES also has an index-level property
663
        # "index.max_inner_result_window" that limits the number of results
664
665
666
667
668
669
670
        # that an inner result can contain.
        energy_aggregation = A(
            "top_hits",
            _source=group_source,
            sort=[{"encyclopedia.properties.energies.energy_total": {"order": "asc"}}],
            size=100,
        )
671
        s.aggs.bucket("groups_eos", energy_aggregation)
672

673
674
675
676
        # We ignore the top level hits
        s = s.extra(**{
            "size": 0,
        })
677

678
        # Collect information for each group from the aggregations
679
        response = s.execute()
680

681
682
683
684
685
686
687
688
        hits = response.aggs.groups_eos.hits
        calculations = [doc.calc_id for doc in hits]
        energies = [doc.encyclopedia.properties.energies.energy_total for doc in hits]
        volumes = [doc.encyclopedia.material.idealized_structure.cell_volume for doc in hits]
        group_dict = {
            "calculations": calculations,
            "energies": energies,
            "volumes": volumes,
689
        }
690
691

        return group_dict, 200
692
693


694
695
696
697
suggestions_map = {
    "code_name": "dft.code_name",
    "structure_type": "encyclopedia.material.bulk.structure_type",
}
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
suggestions_query = api.parser()
suggestions_query.add_argument(
    "property",
    type=str,
    choices=("code_name", "structure_type"),
    help="The property name for which suggestions are returned.",
    location="args"
)
suggestions_result = api.model("suggestions_result", {
    "code_name": fields.List(fields.String),
    "structure_type": fields.List(fields.String),
})


@ns.route("/suggestions")
class EncSuggestionsResource(Resource):
    @api.response(404, "Suggestion not found")
    @api.response(400, "Bad request")
    @api.response(200, "Metadata send", fields.Raw)
    @api.expect(suggestions_query, validate=False)
    @api.marshal_with(suggestions_result, skip_none=True)
    @api.doc("enc_suggestions")
    def get(self):

        # Parse request arguments
        args = suggestions_query.parse_args()
        prop = args.get("property", None)

726
727
728
729
730
731
732
        # Use aggregation to return all unique terms for the requested field.
        # Without using composite aggregations there is a size limit for the
        # number of aggregation buckets. This should, however, not be a problem
        # since the number of unique values is low for all supported properties.
        s = Search(index=config.elastic.index_name)
        query = Q(
            "bool",
733
            filter=get_enc_filter()
734
735
736
737
738
739
740
741
742
743
744
745
746
747
        )
        s = s.query(query)
        s = s.extra(**{
            "size": 0,
        })

        terms_agg = A("terms", field=suggestions_map[prop])
        s.aggs.bucket("suggestions", terms_agg)

        # Gather unique values into a list
        response = s.execute()
        suggestions = [x.key for x in response.aggs.suggestions.buckets]

        return {prop: suggestions}, 200
748
749
750
751
752
753
754
755


calc_prop_map = {
    "calc_id": "calc_id",
    "code_name": "dft.code_name",
    "code_version": "dft.code_version",
    "functional_type": "encyclopedia.method.functional_type",
    "basis_set_type": "dft.basis_set",
756
    "core_electron_treatment": "encyclopedia.method.core_electron_treatment",
757
758
759
760
    "run_type": "encyclopedia.calculation.calculation_type",
    "has_dos": "encyclopedia.properties.electronic_dos",
    "has_band_structure": "encyclopedia.properties.electronic_band_structure",
    "has_thermal_properties": "encyclopedia.properties.thermodynamical_properties",
761
762
    "has_phonon_dos": "encyclopedia.properties.phonon_dos",
    "has_phonon_band_structure": "encyclopedia.properties.phonon_band_structure",
763
764
765
766
767
768
769
}
calculation_result = api.model("calculation_result", {
    "calc_id": fields.String,
    "code_name": fields.String,
    "code_version": fields.String,
    "functional_type": fields.String,
    "basis_set_type": fields.String,
770
    "core_electron_treatment": fields.String,
771
772
773
774
    "run_type": fields.String,
    "has_dos": fields.Boolean,
    "has_band_structure": fields.Boolean,
    "has_thermal_properties": fields.Boolean,
775
776
    "has_phonon_dos": fields.Boolean,
    "has_phonon_band_structure": fields.Boolean,
777
})
778
779
780
781
782
783
representatives_result = api.model("representatives_result", {
    "idealized_structure": fields.String,
    "electronic_band_structure": fields.String,
    "electronic_dos": fields.String,
    "thermodynamical_properties": fields.String,
})
784
785
786
787
calculations_result = api.model("calculations_result", {
    "total_results": fields.Integer,
    "pages": fields.Nested(pages_result),
    "results": fields.List(fields.Nested(calculation_result)),
788
    "representatives": fields.Nested(representatives_result, skip_none=True),
789
790
791
792
})


@ns.route("/materials/<string:material_id>/calculations")
793
class EncCalculationsResource(Resource):
794
    @enable_gzip()
795
796
797
798
799
    @api.response(404, "Suggestion not found")
    @api.response(400, "Bad request")
    @api.response(200, "Metadata send", fields.Raw)
    @api.doc("enc_calculations")
    def get(self, material_id):
800
801
802
        """Used to return all calculations related to the given material. Also
        returns a representative calculation for each property shown in the
        overview page.
803
804
805
806
        """
        s = Search(index=config.elastic.index_name)
        query = Q(
            "bool",
807
            filter=get_enc_filter() + [
808
809
810
811
812
813
814
815
                Q("term", encyclopedia__material__material_id=material_id),
            ]
        )
        s = s.query(query)

        # The query is filtered already on the ES side so we don"t need to
        # transfer so much data.
        s = s.extra(**{
816
817
818
            "_source": {"includes": list(calc_prop_map.values()) + ["dft.xc_functional"]},
            "size": 10000,
            "from": 0,
819
820
821
822
823
824
825
        })
        response = s.execute()

        # No such material
        if len(response) == 0:
            abort(404, message="There is no material {}".format(material_id))

826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
        # Add representative properties. It might be possible to write a custom
        # ES scoring mechanism or aggregation to also perform the selection.
        representatives = {}

        def calc_score(entry):
            """Custom scoring function used to sort results by their
            "quality". Currently built to mimic the scoring that was used
            in the old Encyclopedia GUI.
            """
            score = 0
            functional_score = {
                "GGA": 100
            }
            code_score = {
                "FHI-aims": 3,
                "VASP": 2,
                "Quantum Espresso": 1,
            }
            code_name = entry.dft.code_name
            functional = entry.dft.xc_functional
            has_dos = rgetattr(entry, "encyclopedia.properties.electronic_band_structure") is not None
            has_bs = rgetattr(entry, "encyclopedia.properties.electronic_dos") is not None
            score += functional_score.get(functional, 0)
            score += code_score.get(code_name, 0)
            if has_dos and has_bs:
                score += 10

            return score

        # The calculations are first sorted by "quality"
        sorted_calc = sorted(response, key=lambda x: calc_score(x), reverse=True)

        # Get the requested representative properties
        representatives["idealized_structure"] = sorted_calc[0].calc_id
        thermo_found = False
        bs_found = False
        dos_found = False
        for calc in sorted_calc:
            if rgetattr(calc, "encyclopedia.properties.thermodynamical_properties") is not None:
                representatives["thermodynamical_properties"] = calc.calc_id
                thermo_found = True
            if rgetattr(calc, "encyclopedia.properties.electronic_band_structure") is not None:
                representatives["electronic_band_structure"] = calc.calc_id
                bs_found = True
            if rgetattr(calc, "encyclopedia.properties.electronic_dos") is not None:
                representatives["electronic_dos"] = calc.calc_id
                dos_found = True
            if thermo_found and bs_found and dos_found:
                break

876
877
878
879
880
        # Create result JSON
        results = []
        for entry in response:
            calc_dict = get_es_doc_values(entry, calc_prop_map)
            calc_dict["has_dos"] = calc_dict["has_dos"] is not None
881
            calc_dict["has_band_structure"] = calc_dict["has_band_structure"] is not None
882
            calc_dict["has_thermal_properties"] = calc_dict["has_thermal_properties"] is not None
883
884
            calc_dict["has_phonon_dos"] = calc_dict["has_phonon_dos"] is not None
            calc_dict["has_phonon_band_structure"] = calc_dict["has_phonon_band_structure"] is not None
885
886
887
888
889
            results.append(calc_dict)

        result = {
            "total_results": len(results),
            "results": results,
890
            "representatives": representatives,
891
892
893
894
895
        }

        return result, 200


896
897
898
899
histogram = api.model("histogram", {
    "occurrences": fields.List(fields.Integer),
    "values": fields.List(fields.Float),
})
900
901
statistics_query = api.model("statistics_query", {
    "calculations": fields.List(fields.String),
902
    "properties": fields.List(fields.String),
903
    "n_histogram_bins": fields.Integer,
904
905
906
907
908
})
statistics = api.model("statistics", {
    "min": fields.Float,
    "max": fields.Float,
    "avg": fields.Float,
909
    "histogram": fields.Nested(histogram, skip_none=True)
910
911
})
statistics_result = api.model("statistics_result", {
912
913
914
915
916
917
918
919
920
921
    "cell_volume": fields.Nested(statistics, skip_none=True),
    "atomic_density": fields.Nested(statistics, skip_none=True),
    "mass_density": fields.Nested(statistics, skip_none=True),
    "lattice_a": fields.Nested(statistics, skip_none=True),
    "lattice_b": fields.Nested(statistics, skip_none=True),
    "lattice_c": fields.Nested(statistics, skip_none=True),
    "alpha": fields.Nested(statistics, skip_none=True),
    "beta": fields.Nested(statistics, skip_none=True),
    "gamma": fields.Nested(statistics, skip_none=True),
    "band_gap": fields.Nested(statistics, skip_none=True),
922
})
923
924
925
926
927
928
929
930
931
932
property_map = {
    "cell_volume": "encyclopedia.material.idealized_structure.cell_volume",
    "atomic_density": "encyclopedia.properties.atomic_density",
    "mass_density": "encyclopedia.properties.mass_density",
    "lattice_a": "encyclopedia.material.idealized_structure.lattice_parameters.a",
    "lattice_b": "encyclopedia.material.idealized_structure.lattice_parameters.b",
    "lattice_c": "encyclopedia.material.idealized_structure.lattice_parameters.c",
    "alpha": "encyclopedia.material.idealized_structure.lattice_parameters.alpha",
    "beta": "encyclopedia.material.idealized_structure.lattice_parameters.beta",
    "gamma": "encyclopedia.material.idealized_structure.lattice_parameters.gamma",
Lauri Himanen's avatar
Lauri Himanen committed
933
    "band_gap": "encyclopedia.properties.band_gap",
934
}
935
936
937


@ns.route("/materials/<string:material_id>/statistics")
938
class EncStatisticsResource(Resource):
939
940
941
    @api.response(404, "Suggestion not found")
    @api.response(400, "Bad request")
    @api.response(200, "Metadata send", fields.Raw)
942
943
    @api.expect(statistics_query, validate=False)
    @api.marshal_with(statistics_result, skip_none=True)
944
945
    @api.doc("enc_statistics")
    def post(self, material_id):
946
947
        """Used to return statistics related to the specified material and
        calculations.
948
        """
949
950
951
952
953
954
955
956
957
        # Get query parameters as json
        try:
            data = marshal(request.get_json(), statistics_query)
        except Exception as e:
            abort(400, message=str(e))

        # Find entries for the given material.
        bool_query = Q(
            "bool",
958
            filter=get_enc_filter() + [
959
960
961
962
963
964
965
966
967
968
969
                Q("term", encyclopedia__material__material_id=material_id),
                Q("terms", calc_id=data["calculations"]),
            ]
        )

        s = Search(index=config.elastic.index_name)
        s = s.query(bool_query)
        s = s.extra(**{
            "size": 0,
        })

970
971
972
973
974
975
        # Add statistics aggregations for each requested property
        properties = data["properties"]
        for prop in properties:
            stats_agg = A("stats", field=property_map[prop])
            s.aggs.bucket("{}_stats".format(prop), stats_agg)

976
977
978
979
980
        # No hits on the top query level
        response = s.execute()
        if response.hits.total == 0:
            abort(404, message="Could not find matching calculations.")

981
982
983
984
985
986
987
988
        # Run a second query that creates histograms with fixed size buckets
        # based on the min and max from previous query. Might make more sense
        # to use the mean and sigma to define the range?
        s = Search(index=config.elastic.index_name)
        s = s.query(bool_query)
        s = s.extra(**{
            "size": 0,
        })
989
        n_bins = data["n_histogram_bins"]
990
991
        for prop in properties:
            stats = getattr(response.aggs, "{}_stats".format(prop))
992
993
            if stats.count == 0:
                continue
994
            interval = (stats.max * 1.001 - stats.min) / n_bins
995
996
            if interval == 0:
                interval = 1
997
            hist_agg = A("histogram", field=property_map[prop], interval=interval, offset=stats.min, min_doc_count=0)
998
999
1000
            s.aggs.bucket("{}_hist".format(prop), hist_agg)
        response_hist = s.execute()

1001
        # Return results
1002
1003
1004
        result = {}
        for prop in properties:
            stats = getattr(response.aggs, "{}_stats".format(prop))
1005
1006
            if stats.count == 0:
                continue
1007
1008
1009
1010
1011
1012
1013
            hist = getattr(response_hist.aggs, "{}_hist".format(prop))
            occurrences = [x.doc_count for x in hist.buckets]
            values = [x.key for x in hist.buckets]
            result[prop] = {
                "min": stats.min,
                "max": stats.max,
                "avg": stats.avg,
1014
                "histogram": {
1015
1016
1017
                    "occurrences": occurrences,
                    "values": values,
                }
1018
            }
1019

1020
        return result, 200
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031


wyckoff_variables_result = api.model("wyckoff_variables_result", {
    "x": fields.Float,
    "y": fields.Float,
    "z": fields.Float,
})
wyckoff_set_result = api.model("wyckoff_set_result", {
    "wyckoff_letter": fields.String,
    "indices": fields.List(fields.Integer),
    "element": fields.String,
1032
    "variables": fields.Nested(wyckoff_variables_result, skip_none=True),
1033
})
1034
1035
1036
1037
1038
1039
1040
1041
lattice_parameters = api.model("lattice_parameters", {
    "a": fields.Float,
    "b": fields.Float,
    "c": fields.Float,
    "alpha": fields.Float,
    "beta": fields.Float,
    "gamma": fields.Float,
})
1042
1043
1044
1045
1046
1047

idealized_structure_result = api.model("idealized_structure_result", {
    "atom_labels": fields.List(fields.String),
    "atom_positions": fields.List(fields.List(fields.Float)),
    "lattice_vectors": fields.List(fields.List(fields.Float)),
    "lattice_vectors_primitive": fields.List(fields.List(fields.Float)),
1048
    "lattice_parameters": fields.Nested(lattice_parameters, skip_none=True),
1049
1050
1051
    "periodicity": fields.List(fields.Boolean),
    "number_of_atoms": fields.Integer,
    "cell_volume": fields.Float,
1052
    "wyckoff_sets": fields.List(fields.Nested(wyckoff_set_result, skip_none=True)),
1053
1054
})

1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
calculation_property_map = {
    "lattice_parameters": {
        "es_source": "encyclopedia.material.idealized_structure.lattice_parameters"
    },
    "energies": {
        "es_source": "encyclopedia.properties.energies",
    },
    "mass_density": {
        "es_source": "encyclopedia.properties.mass_density",
    },
    "atomic_density": {
        "es_source": "encyclopedia.properties.atomic_density",
    },
    "cell_volume": {
        "es_source": "encyclopedia.material.idealized_structure.cell_volume"
    },
Lauri Himanen's avatar
Lauri Himanen committed
1071
1072
1073
    "band_gap": {
        "es_source": "encyclopedia.properties.band_gap"
    },
1074
1075
1076
1077
1078
1079
    "electronic_band_structure": {
        "es_source": "encyclopedia.properties.electronic_band_structure"
    },
    "electronic_dos": {
        "es_source": "encyclopedia.properties.electronic_dos"
    },
1080
1081
1082
1083
1084
1085
1086
1087
1088
    "phonon_band_structure": {
        "es_source": "encyclopedia.properties.phonon_band_structure"
    },
    "phonon_dos": {
        "es_source": "encyclopedia.properties.phonon_dos"
    },
    "thermodynamical_properties": {
        "es_source": "encyclopedia.properties.thermodynamical_properties"
    },
1089
1090
1091
    "wyckoff_sets": {
        "arch_source": "section_metadata/encyclopedia/material/idealized_structure/wyckoff_sets"
    },
1092
1093
1094
    "idealized_structure": {
        "arch_source": "section_metadata/encyclopedia/material/idealized_structure"
    },
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
}

calculation_property_query = api.model("calculation_query", {
    "properties": fields.List(fields.String),
})
energies = api.model("energies", {
    "energy_total": fields.Float,
    "energy_total_T0": fields.Float,
    "energy_free": fields.Float,
})
Lauri Himanen's avatar
Lauri Himanen committed
1105
1106
1107
1108
electronic_band_structure = api.model("electronic_band_structure", {
    "reciprocal_cell": fields.List(fields.List(fields.Float)),
    "brillouin_zone": fields.Raw,
    "section_k_band_segment": fields.Raw,
1109
    "section_band_gap": fields.Raw,
Lauri Himanen's avatar
Lauri Himanen committed
1110
1111
1112
})
electronic_dos = api.model("electronic_dos", {
    "dos_energies": fields.List(fields.Float),
1113
    "dos_values": fields.List(fields.List(fields.Float)),
Lauri Himanen's avatar
Lauri Himanen committed
1114
})
1115
1116
1117
calculation_property_result = api.model("calculation_property_result", {
    "lattice_parameters": fields.Nested(lattice_parameters, skip_none=True),
    "energies": fields.Nested(energies, skip_none=True),
1118
1119
1120
    "mass_density": fields.Float,
    "atomic_density": fields.Float,
    "cell_volume": fields.Float,
1121
    "wyckoff_sets": fields.Nested(wyckoff_set_result, skip_none=True),
1122
    "idealized_structure": fields.Nested(idealized_structure_result, skip_none=True),
1123
1124
1125
    "band_gap": fields.Float,
    "electronic_band_structure": fields.Nested(electronic_band_structure, skip_none=True),
    "electronic_dos": fields.Nested(electronic_dos, skip_none=True),
1126
1127
1128
    "phonon_band_structure": fields.Raw,
    "phonon_dos": fields.Raw,
    "thermodynamical_properties": fields.Raw,
1129
1130
1131
})


1132
1133
@ns.route("/materials/<string:material_id>/calculations/<string:calc_id>")
class EncCalculationResource(Resource):
1134
    @enable_gzip()
1135
1136
1137
    @api.response(404, "Material or calculation not found")
    @api.response(400, "Bad request")
    @api.response(200, "Metadata send", fields.Raw)
1138
    @api.expect(calculation_property_query, validate=False)
1139
    @api.marshal_with(calculation_property_result, skip_none=True)
1140
    @api.doc("enc_calculation")
1141
1142
1143
1144
    def post(self, material_id, calc_id):
        """Used to return calculation details. Some properties are not
        available in the ES index and are instead read from the Archive
        directly.
1145
        """
1146
1147
1148
1149
1150
1151
        # Get query parameters as json
        try:
            data = marshal(request.get_json(), calculation_property_query)
        except Exception as e:
            abort(400, message=str(e))

1152
1153
1154
        s = Search(index=config.elastic.index_name)
        query = Q(
            "bool",
1155
            filter=get_enc_filter() + [
1156
1157
1158
1159
1160
1161
                Q("term", encyclopedia__material__material_id=material_id),
                Q("term", calc_id=calc_id),
            ]
        )
        s = s.query(query)

1162
        # Create dictionaries for requested properties
1163
        references = []
1164
1165
1166
        properties = data["properties"]
        arch_properties = {}
        es_properties = {}
1167
1168
1169
1170
1171
1172
1173
        ref_properties = set((
            "electronic_dos",
            "electronic_band_structure",
            "thermodynamical_properties",
            "phonon_dos",
            "phonon_band_structure",
        ))
1174
1175
1176
1177
        for prop in properties:
            es_source = calculation_property_map[prop].get("es_source")
            if es_source is not None:
                es_properties[prop] = es_source
1178
                if prop in ref_properties:
1179
                    references.append(prop)
1180
1181
1182
1183
            arch_source = calculation_property_map[prop].get("arch_source")
            if arch_source is not None:
                arch_properties[prop] = arch_source

1184
        # The query is filtered already on the ES side so we don't need to
1185
        # transfer so much data.
1186
1187
1188
        sources = [
            "upload_id",
            "calc_id",
1189
            "encyclopedia",
1190
1191
1192
        ]
        sources += list(es_properties.values())

1193
        s = s.extra(**{
1194
            "_source": {"includes": sources},
1195
1196
1197
1198
1199
1200
1201
1202
1203
            "size": 1,
        })

        response = s.execute()

        # No such material
        if len(response) == 0:
            abort(404, message="There is no material {} with calculation {}".format(material_id, calc_id))

1204
1205
1206
        # Add references that are to be read from the archive
        for ref in references:
            arch_path = response[0]
1207
1208
            arch_path = rgetattr(arch_path, es_properties[ref])
            if arch_path is not None:
Lauri Himanen's avatar
Lauri Himanen committed
1209
                arch_properties[ref] = arch_path
1210
1211
            del es_properties[ref]

1212
1213
1214
1215
1216
1217
1218
        # If any of the requested properties require data from the Archive, the
        # file is opened and read.
        result = {}
        if len(arch_properties) != 0:
            entry = response[0]
            upload_id = entry.upload_id
            calc_id = entry.calc_id
1219
            root = read_archive(
1220
1221
1222
1223
1224
                upload_id,
                calc_id,
            )

            # Add results from archive
1225
1226
1227
            for key, arch_path in arch_properties.items():
                value = root[arch_path]

1228
1229
                # Replace unnormalized thermodynamical properties with
                # normalized ones and turn into dict
1230
1231
1232
                if key == "thermodynamical_properties":
                    specific_heat_capacity = value.specific_heat_capacity.magnitude.tolist()
                    specific_free_energy = value.specific_vibrational_free_energy_at_constant_volume.magnitude.tolist()
1233
1234
                    specific_heat_capacity = [x if np.isfinite(x) else None for x in specific_heat_capacity]
                    specific_free_energy = [x if np.isfinite(x) else None for x in specific_free_energy]
1235
1236
1237
1238
                if isinstance(value, list):
                    value = [x.m_to_dict() for x in value]
                else:
                    value = value.m_to_dict()
1239
                if key == "thermodynamical_properties":
1240
1241
                    del value["thermodynamical_property_heat_capacity_C_v"]
                    del value["vibrational_free_energy_at_constant_volume"]
1242
1243
                    value["specific_heat_capacity"] = specific_heat_capacity
                    value["specific_vibrational_free_energy_at_constant_volume"] = specific_free_energy
1244

1245
                # DOS results are simplified.