statistics.py 20.8 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
# 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.

"""
A command that generates various statistics.
"""

19
20
21
22
23
from matplotlib import scale as mscale
from matplotlib import transforms as mtransforms
import matplotlib.pyplot as plt
import matplotlib.ticker as ticker
import numpy as np
24
import click
Markus Scheidgen's avatar
Markus Scheidgen committed
25
import json
26
from datetime import datetime
27

28
from .client import client
29
30


Markus Scheidgen's avatar
Markus Scheidgen committed
31
32
def codes(client, minimum=1, **kwargs):
    data = client.repo.search(per_page=1, **kwargs).response().result
33

34
    x_values = sorted([
35
        code for code, values in data.statistics['code_name'].items()
36
        if code != 'not processed' and values.get('calculations', 1000) >= minimum], key=lambda x: x.lower())
37

38
    return data.statistics, x_values, 'code_name', 'code'
39
40


Markus Scheidgen's avatar
Markus Scheidgen committed
41
42
def dates(client, minimum=1, **kwargs):
    data = client.repo.search(per_page=1, date_histogram=True, **kwargs).response().result
43
44

    x_values = list([
45
        x for x in data.statistics['date_histogram'].keys()])
46

47
    return data.statistics, x_values, 'date_histogram', 'month'
48

49

50
def error_fig(client):
51
    _, labels, _, _ = codes(client)
52
53
54
55
56
57
58
59
60
61
62

    def code_values(metric='code_runs', **kwargs):
        result = client.repo.search(
            per_page=1,
            owner='admin',
            metrics=[] if metric == 'code_runs' else metric,
            **kwargs).response().result

        return {
            code: values[metric]
            for code, values in result.quantities['code_name'].items()
63
            if code != 'not processed' and (not labels or code in labels) > 0}
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

    # get the data
    all_entries = code_values()
    parser_failure_label = 'parser failure'
    error_types = [
        {'name': parser_failure_label, 'search': dict(system='not processed')},
        {'name': 'failed system classification', 'search': dict(system='unavailable')},
        {'name': 'no basis set available', 'search': dict(basis_set='unavailable')},
        {'name': 'no XC functional available', 'search': dict(xc_functional='unavailable')}
    ]
    errors = {
        error_type['name']: {
            code: failures
            for code, failures in code_values(**error_type['search']).items()}
        for error_type in error_types}
    errors_rates = {
        error_type['name']: {
            code: 0 if all_entries[code] == 0 else failures / all_entries[code]
            for code, failures in code_values(**error_type['search']).items()}
        for error_type in error_types}

    fig, axs = plt.subplots(figsize=(15, 12), dpi=72, nrows=2)

    def draw_error_chart(errors, ax, colors, entries=None, mul=1, scale=0.5):
        n_bars = len(errors) - 1
        leg_colors = list(colors)

        x = np.arange(len(labels))  # the label locations
        width = 0.7 / n_bars  # the width of the bars
        plt.sca(ax)
        plt.xticks(rotation=90)

        if entries is not None:
            ax.bar(x, [entries[code] for code in labels], width * n_bars, label='all entries', color=colors.pop(0))

        i = -1
        not_processed = [errors[parser_failure_label][code] * mul for code in labels]
        ax.bar(x, not_processed, width * n_bars, label=parser_failure_label, color=colors.pop(0))
        for key, values in errors.items():
            if key != parser_failure_label:
                ax.bar(x + i * width, [values[code] * mul for code in labels], width, label=key, bottom=not_processed, color=colors.pop(0))
                i += 1

        # Add some text for labels, title and custom x-axis tick labels, etc.
        ax.set_yscale('power', exponent=scale)
        ax.set_xticks(x)
        ax.set_xticklabels(labels)
        ax.legend()
        leg = ax.get_legend()
        for i in range(0, len(leg_colors)):
            leg.legendHandles[i].set_color(leg_colors[i])

        fig.tight_layout()

    ax = axs[0]
    ax.set_title('Absolute number of entries with parser errors or missing repository metadata compared to all entries per code')
    ax.set_ylabel('number of entries', )
    colors = ['grey', 'red', 'yellow', 'orange', 'brown']
    draw_error_chart(errors, ax, entries=all_entries, mul=1, scale=0.25, colors=colors)
    ax.yaxis.set_major_formatter(ticker.StrMethodFormatter('{x:,.0f}'))

    ax = axs[1]
    ax.set_title('Relative rates of entries with parser errors or missing repository metadata per code')
    ax.set_ylabel('rate in %', )
    colors = ['red', 'yellow', 'orange', 'brown']
    draw_error_chart(errors_rates, ax, mul=100, colors=colors)

    plt.show()

Markus Scheidgen's avatar
Markus Scheidgen committed
133
134
    return fig, plt

135

136
class Metric:
Markus Scheidgen's avatar
Markus Scheidgen committed
137
    def __init__(self, metric, label=None, power=None, multiplier=1, format=None, cumulate=False):
138
139
        if label is None:
            label = metric
140

141
142
143
144
145
146
147
148
        self.metric = metric
        self.agg = None
        self.label = label
        self.multiplier = multiplier
        self.power = power
        self.format = format
        self.cumulate = cumulate

Markus Scheidgen's avatar
Markus Scheidgen committed
149
150
    def draw_axis(self, axis, data, x_values, x_positions, width, color, only=False):
        label_color = 'black' if only else color
151
152
153
154
155
        value_map = {
            x: values[self.metric]
            for x, values in data[self.agg].items()
            if x in x_values}

Markus Scheidgen's avatar
Markus Scheidgen committed
156
157
158
159
160
        if self.power is not None:
            axis.set_yscale('power', exponent=self.power)
        else:
            axis.set_yscale('log')
        axis.set_ylabel(self.label, color=label_color)
161
162
163
164
165
166
167

        if self.format is not None:
            axis.yaxis.set_major_formatter(ticker.StrMethodFormatter(self.format))

        y_values = [value_map[x] * self.multiplier for x in x_values]
        if self.cumulate:
            y_values = np.array(y_values).cumsum()
Markus Scheidgen's avatar
Markus Scheidgen committed
168
169
170
171
        axis.bar(x_positions, y_values, width, label=self.label, color=color, align='edge')
        axis.tick_params(axis='y', labelcolor=label_color)

        # TODO remove
172
173
174
175
176
        # for x, v in zip(x_positions, y_values):
        #     axis.text(x + .1, v, ' {:,}'.format(int(v)), color=color, fontweight='bold', rotation=90)
        # import matplotlib.lines as mlines
        # line = mlines.Line2D([min(x_positions), max(x_positions)], [72, 72], color=color)
        # axis.add_line(line)
177

Markus Scheidgen's avatar
Markus Scheidgen committed
178

179
180
181
182
183
184
def bar_plot(
        client, retrieve, metric1, metric2=None, title=None, format_xlabel=None,
        xlim={}, ylim=dict(bottom=1), **kwargs):
    if format_xlabel is None:
        format_xlabel = lambda x: x

185
186
187
188
    metrics = [] if metric1.metric == 'code_runs' else [metric1.metric]
    if metric2 is not None:
        metrics += [] if metric2.metric == 'code_runs' else [metric2.metric]

189
    data, x_values, agg, agg_label = retrieve(client, metrics=metrics, statistics=True, **kwargs)
190
191
192
    metric1.agg = agg
    if metric2 is not None:
        metric2.agg = agg
193

194
    fig, ax1 = plt.subplots(figsize=(5, 4), dpi=72)
195
    x = np.arange(len(x_values))
Markus Scheidgen's avatar
Markus Scheidgen committed
196
    width = 0.8 / 2
197
    if metric2 is None:
Markus Scheidgen's avatar
Markus Scheidgen committed
198
        width = 0.8
199
200
201
    plt.sca(ax1)
    plt.xticks(rotation=90)
    ax1.set_xticks(x)
202
    ax1.set_xticklabels([format_xlabel(value) if value != 'Quantum Espresso' else 'Q. Espresso' for value in x_values])
Markus Scheidgen's avatar
Markus Scheidgen committed
203
    ax1.margins(x=0.01)
204
205
206
207
208
209
    ax1.set_xlim(**xlim)
    # i = 0
    # for label in ax1.xaxis.get_ticklabels():
    #     label.set_visible(i % 4 == 0)
    #     i += 1

210
211
212
213
214
215
    if title is None:
        title = 'Number of %s' % metric1.label
        if metric2 is not None:
            title += ' and %s' % metric2.label
        title += ' per %s' % agg_label
        ax1.set_title(title)
216
217
    elif title != '':
        ax1.set_title(title)
218

Markus Scheidgen's avatar
Markus Scheidgen committed
219
    metric1.draw_axis(ax1, data, x_values, x - (width / 2), width, 'tab:blue', only=metric2 is None)
220
221
222
    ax1.set_ylim(**ylim)
    ax1.set_yticks([40, 30, 20, 10, 5, 1, 0.5, 0.1])
    ax1.grid(which='major', axis='y', linestyle='--')
223

224
225
    if metric2:
        ax2 = ax1.twinx()  # instantiate a second axes that shares the same x-axis
226
227
228
        metric2.draw_axis(
            ax2, data, x_values, x + width / 2, width, 'tab:red')
        ax2.set_ylim(bottom=1)
229
230

    fig.tight_layout()
Markus Scheidgen's avatar
Markus Scheidgen committed
231
232

    return fig, plt
233
234


235
236
@client.command(help='Generate various matplotlib charts')
@click.option('--errors', is_flag=True, help='Two charts with relative and absolute parser/normalizer errors per code.')
237
@click.option('--x-axis', type=str, help='Aggregation used for x-axis, values are "code" and "time".')
238
@click.option('--y-axis', multiple=True, type=str, help='Metrics used for y-axis, values are "entries", "energies", "calculations", "users".')
239
240
241
@click.option('--cumulate', is_flag=True, help='Cumulate over x-axis.')
@click.option('--title', type=str, help='Override chart title with given value.')
@click.option('--total', is_flag=True, help='Provide total sums of key metrics.')
Markus Scheidgen's avatar
Markus Scheidgen committed
242
243
244
245
@click.option('--save', type=str, help='Save to given file instead of showing the plot.')
@click.option('--power', type=float, help='User power scale instead of log with the given inverse power.')
@click.option('--open-access', is_flag=True, help='Only consider Open-Access data.')
@click.option('--minimum', type=int, default=1, help='Only consider codes with at least the given ammount of entries.')
246
def statistics_plot(errors, title, x_axis, y_axis, cumulate, total, save, power, open_access, minimum):
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
    from .client import create_client
    client = create_client()

    class PowerScale(mscale.ScaleBase):
        name = 'power'

        def __init__(self, axis, exponent, **kwargs):
            mscale.ScaleBase.__init__(self, axis, **kwargs)
            self.exponent = exponent

        def set_default_locators_and_formatters(self, axis):
            axis.set_major_locator(ticker.AutoLocator())
            axis.set_major_formatter(ticker.ScalarFormatter())
            axis.set_minor_locator(ticker.NullLocator())
            axis.set_minor_formatter(ticker.NullFormatter())

        def limit_range_for_scale(self, vmin, vmax, minpos):
            return max(0., vmin), vmax

        class Transform(mtransforms.Transform):
            input_dims = 1
            output_dims = 1
            is_separable = True

            def __init__(self, exponent):
                super().__init__()
                self.exponent = exponent

            def transform_non_affine(self, a):
                return np.array(a)**self.exponent

            def inverted(self):
                return PowerScale.Transform(1 / self.exponent)

        def get_transform(self):
            return self.Transform(self.exponent)

    mscale.register_scale(PowerScale)

Markus Scheidgen's avatar
Markus Scheidgen committed
286
287
288
289
290
291
292
293
294
295
296
297
    kwargs = {}
    if cumulate:
        kwargs.update(
            power=1,
            multiplier=1e-6,
            format='{x:,.1f}M')
    elif power is not None:
        kwargs.update(
            power=1 / power,
            multiplier=1e-6,
            format='{x:,.1f}M')

298
299
300
301
302
    metrics = {
        'entries': Metric(
            'code_runs',
            label='entries (code runs)',
            cumulate=cumulate,
Markus Scheidgen's avatar
Markus Scheidgen committed
303
            **kwargs),
304
305
306
307
308
309
310
311
        'users': Metric(
            'users',
            cumulate=cumulate,
            label='users that provided data'),
        'energies': Metric(
            'total_energies',
            label='total energy calculations',
            cumulate=cumulate,
Markus Scheidgen's avatar
Markus Scheidgen committed
312
            **kwargs),
313
314
        'calculations': Metric(
            'calculations',
Markus Scheidgen's avatar
Markus Scheidgen committed
315
            label='calculations (e.g. total energy)',
316
            cumulate=cumulate,
Markus Scheidgen's avatar
Markus Scheidgen committed
317
            **kwargs)
318
319
    }

320
    if errors:
Markus Scheidgen's avatar
Markus Scheidgen committed
321
322
323
        fig, plt = error_fig(client)

    owner = 'all' if open_access else 'admin'
324
325
326
327

    if x_axis is not None:
        assert 1 <= len(y_axis) <= 2, 'Need 1 or 2 y axis'

328
        kwargs = {}
329
330
        if x_axis == 'code':
            x_axis = codes
331
            kwargs.update(ylim=dict(bottom=0))
332
333
        elif x_axis == 'time':
            x_axis = dates
334
335
336
            kwargs.update(
                ylim=dict(bottom=0),
                format_xlabel=lambda x: datetime.fromtimestamp(int(x) / 1000).strftime('%b %y'))
337
338
339
340
341
        else:
            assert False, 'x axis can only be "code" or "time"'

        y_axis = [metrics[y] for y in y_axis]

342
343
        fig, plt = bar_plot(
            client, x_axis, *y_axis, title=title, owner=owner, minimum=minimum, **kwargs)
Markus Scheidgen's avatar
Markus Scheidgen committed
344
345
346
347
348
349

    if errors or x_axis is not None:
        if save is not None:
            fig.savefig(save, bbox_inches='tight')
        else:
            plt.show()
350
351

    if total:
352
        data = client.repo.search(
353
354
355
            per_page=1, owner=owner, statistics=True,
            metrics=['total_energies', 'calculations', 'uploaders', 'authors', 'datasets']).response().result
        print(json.dumps(data.statistics['total'], indent=4))
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
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
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
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545


@client.command(help='Generate table with basic statistics summary.')
@click.option('--html', is_flag=True, help='Output HTML instead of plain text table.')
@click.option('--geometries', is_flag=True, help='Use geometries not unique geometries.')
def statistics_table(html, geometries):
    def get_statistic(response, quantity, value, metric):
        quantity_data = response.statistics.get(quantity)
        if quantity_data is None:
            return 0
        value_data = quantity_data.get(value)
        if value_data is None:
            return 0

        value = value_data.get(metric)
        return value if value is not None else 0

    from nomad.cli.client import create_client
    client = create_client()

    geometry_metric = 'unique_geometries' if not geometries else 'geometries'

    # search scc with system type
    data_all = client.repo.search(
        per_page=1, metrics=['calculations'], statistics=True).response().result

    entries = get_statistic(data_all, 'total', 'all', 'code_runs')
    calculations = get_statistic(data_all, 'total', 'all', 'calculations')
    calculations_1d = get_statistic(data_all, 'system', '1D', 'calculations') \
        + get_statistic(data_all, 'system', 'atom', 'calculations') \
        + get_statistic(data_all, 'system', 'molecule / cluster', 'calculations')

    calculations_2d = get_statistic(data_all, 'system', '2D / surface', 'calculations')
    calculations_3d = get_statistic(data_all, 'system', 'bulk', 'calculations')

    metrics_all = client.repo.search(per_page=1, metrics=[geometry_metric]).response().result
    geometries = get_statistic(metrics_all, 'total', 'all', geometry_metric)
    quantities = get_statistic(metrics_all, 'total', 'all', 'quantities')

    # search calcs quantities=section_k_band
    band_structures = get_statistic(
        client.repo.search(per_page=1, quantities=['section_k_band']).response().result,
        'total', 'all', 'code_runs')

    # search calcs quantities=section_dos
    dos = get_statistic(
        client.repo.search(per_page=1, quantities=['section_dos']).response().result,
        'total', 'all', 'code_runs')

    phonons = get_statistic(
        client.repo.search(per_page=1, code_name='Phonopy').response().result,
        'total', 'all', 'code_runs')

    if not html:
        print('''
            Entries: {:,},
            Calculations, e.g. total energies: {:,},
            Unique geometries: {:,},
            Bulk crystals: {:,},
            2D / Surfaces: {:,},
            Atoms / Molecules: {:,},
            DOS: {:,},
            Band structures: {:,}
            Total parsed quantities: {:,}
        '''.format(
            entries,
            calculations,
            geometries,
            calculations_3d,
            calculations_2d,
            calculations_1d,
            dos,
            band_structures,
            quantities
        ))

    else:
        print('''
            <div class="container">
                <p>The <i>NOMAD Archive</i> stores in a code-independent format calculations performed
                with all the most important and widely used electronic-structure and force-field codes.
                </p>
                <p>Summary statistics of the Archive content (last update in {}):</p>
                <table class="table">
                    <thead>
                        <tr>
                        <th scope="col">Metric</th>
                        <th scope="col">Value</th>
                        </tr>
                    </thead>
                    <tbody>
                        <tr>
                        <th scope="row">Entries, i.e. code runs</th>
                        <td>{:,}</td>
                        </tr>
                        <tr>
                        <th scope="row">Calculations, e.g. total energies</th>
                        <td>{:,}</td>
                        </tr>
                        <tr>
                        <th scope="row">Unique geometries</th>
                        <td>{:,}</td>
                        </tr>
                        <tr>
                        <th scope="row">Bulk Crystals</th>
                        <td>{:,}</td>
                        </tr>
                        <tr>
                        <th scope="row">Surfaces</th>
                        <td>{:,}</td>
                        </tr>
                        <tr>
                        <th scope="row">Molecules/Clusters</th>
                        <td>{:,}</td>
                        </tr>
                        <tr>
                        <th scope="row">DOS</th>
                        <td>{:,}</td>
                        </tr>
                        <tr>
                        <th scope="row">Band Structures</th>
                        <td>{:,}</td>
                        </tr>
                        <tr>
                        <th scope="row">Phonon Calculations</th>
                        <td>{:,}</td>
                        </tr>
                        <tr>
                        <th scope="row">Overall parsed quantities</th>
                        <td>{:,}</td>
                        </tr>
                    </tbody>
                </table>
                <p>
                    Furthermore:
                </p>
                <ul>
                    <li><b>9,274</b> Zip Archives for parsing: <b>16.5 TB</b> of data (compressed)</li>
                    <li>Data extracted with parsing: <b>5.6 TB</b> of HDF5 files (compressed)</li>
                    <li>Data classified using <b>168</b> public metadata of the NOMAD Meta Info and <b>2,360</b> code-specific metadata</li>
                    <li>Number of parsed quantities <b>871,497,996</b></li>
                </ul>
                <p>
                    90% of VASP calculations are provided by
                        <a href="http://aflowlib.org">AFLOWlib</a> (S. Curtarolo),
                        <a href="http://oqmd.org"> OQMD</a> (C. Wolverton) and
                        <a href="https://materialsproject.org">Materials Project</a> (K. Persson).
                </p>
                <p>
                    You can further explore the statistics in the below dynamic histograms. To
                    change the displayed quantity, select from the "Quantities" drop-down. To
                    filter the data, click histogram bars for different filter combinations.
                    To reset filters, click "Reset Filters".
                </p>
                <p>
                    The archive data is represented in a code-independent, structured
                    form. The archive structure and all quantities are described via the
                    <a href="https://www.nomad-coe.eu/the-project/nomad-archive/archive-meta-info">NOMAD Metainfo</a>.
                    The NOMAD Metainfo defines a conceptual model to store the values connected
                    to atomistic or <i>ab initio</i> calculations. A clear and usable metadata definition
                    is a prerequisites to preparing the data for analysis that everybody
                    can contribute to.
                </p>
                <p>
                    In collaboration with the <a href="http://www.bbdc.berlin/">Berlin Big Data Center (BBDC)</a>,
                    we use the Apache Flink infrastructure to support and go beyond the standard MapReduce model to enable
                    rapid and complex queries.
                </p>
                <p>
                    Contact concerning general aspects of the CoE: <a href="mailto:pietsch@fhi-berlin.mpg.de">Jessica Pietsch</a>
                </p>
                <p>
                    Contact concerning the NOMAD Archive:
                        <a href="mailto:markus.scheidgen@physik.hu-berlin.de">Markus Scheidgen</a>,
                        <a href="mailto:ghiringhelli@fhi-berlin.mpg.de">Luca Ghiringhelli</a>
                </p>
            </div>
        '''.format(
            datetime.now().strftime('%b %y'),
            entries,
            calculations,
            geometries,
            calculations_3d,
            calculations_2d,
            calculations_1d,
            dos,
            band_structures,
            phonons,
            quantities
        ))