• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1# Copyright (c) 2012 The Chromium OS Authors. All rights reserved.
2# Use of this source code is governed by a BSD-style license that can be
3# found in the LICENSE file.
4
5"""A module handling the logs.
6
7The structure of this module:
8
9    RoundLog: the test results of every round are saved in a log file.
10              includes: fw, and round_name (i.e., the date time of the round
11
12      --> GestureLogs: includes gesture name, and variation
13
14            --> ValidatorLogs: includes name, details, criteria, score, metrics
15
16
17    SummaryLog: derived from multiple RoundLogs
18      --> SimpleTable: (key, vlog) pairs
19            key: (fw, round_name, gesture_name, variation_name, validator_name)
20            vlog: name, details, criteria, score, metrics
21
22    TestResult: encapsulation of scores and metrics
23                used by a client program to query the test results
24      --> StatisticsScores: includes average, ssd, and count
25      --> StatisticsMetrics: includes average, min, max, and more
26
27
28How the logs work:
29    (1) ValidatorLogs are contained in a GestureLog.
30    (2) Multiple GestureLogs are packed in a RoundLog which is saved in a
31        separate pickle log file.
32    (3) To construct a SummaryLog, it reads RoundLogs from all pickle logs
33        in the specified log directory. It then creates a SimpleTable
34        consisting of (key, ValidatorLog) pairs, where
35        key is a 5-tuple:
36            (fw, round_name, gesture_name, variation_name, validator_name).
37    (4) The client program, i.e., firmware_summary module, contains a
38        SummaryLog, and queries all statistics using get_result() which returns
39        a TestResult object containing both StatisticsScores and
40        StatisticsMetrics.
41
42"""
43
44
45import glob
46import numpy as np
47import pickle
48import os
49
50import test_conf as conf
51import validators as val
52
53from collections import defaultdict, namedtuple
54
55from common_util import Debug, print_and_exit
56from firmware_constants import AXIS
57
58
59MetricProps = namedtuple('MetricProps', ['description', 'note', 'stat_func'])
60
61
62def _setup_debug(debug_flag):
63    """Set up the global debug_print function."""
64    if 'debug_print' not in globals():
65        global debug_print
66        debug = Debug(debug_flag)
67        debug_print = debug.print_msg
68
69
70def _calc_sample_standard_deviation(sample):
71    """Calculate the sample standard deviation (ssd) from a given sample.
72
73    To compute a sample standard deviation, the following formula is used:
74        sqrt(sum((x_i - x_average)^2) / N-1)
75
76    Note that N-1 is used in the denominator for sample standard deviation,
77    where N-1 is the degree of freedom. We need to set ddof=1 below;
78    otherwise, N would be used in the denominator as ddof's default value
79    is 0.
80
81    Reference:
82        http://en.wikipedia.org/wiki/Standard_deviation
83    """
84    return np.std(np.array(sample), ddof=1)
85
86
87class float_d2(float):
88    """A float type with special __repr__ and __str__ methods that display
89    the float number to the 2nd decimal place."""
90    template = '%.2f'
91
92    def __str__(self):
93        """Display the float to the 2nd decimal place."""
94        return self.template % self.real
95
96    def __repr__(self):
97        """Display the float to the 2nd decimal place."""
98        return self.template % self.real
99
100
101def convert_float_to_float_d2(value):
102    """Convert the float(s) in value to float_d2."""
103    if isinstance(value, float):
104        return float_d2(value)
105    elif isinstance(value, tuple):
106        return tuple(float_d2(v) if isinstance(v, float) else v for v in value)
107    else:
108        return value
109
110
111class Metric:
112    """A class to handle the name and the value of a metric."""
113    def __init__(self, name, value):
114        self.name = name
115        self.value = convert_float_to_float_d2(value)
116
117    def insert_key(self, key):
118        """Insert the key to this metric."""
119        self.key = key
120        return self
121
122
123class MetricNameProps:
124    """A class keeping the information of metric name templates, descriptions,
125    and statistic functions.
126    """
127
128    def __init__(self):
129        self._init_raw_metrics_props()
130        self._derive_metrics_props()
131
132    def _init_raw_metrics_props(self):
133        """Initialize raw_metrics_props.
134
135        The raw_metrics_props is a dictionary from metric attribute to the
136        corresponding metric properties. Take MAX_ERR as an example of metric
137        attribute. Its metric properties include
138          . metric name template: 'max error in {} (mm)'
139            The metric name template will be expanded later. For example,
140            with name variations ['x', 'y'], the above template will be
141            expanded to:
142                'max error in x (mm)', and
143                'max error in y (mm)'
144          . name variations: for example, ['x', 'y'] for MAX_ERR
145          . metric name description: 'The max err of all samples'
146          . metric note: None
147          . the stat function used to calculate the statistics for the metric:
148            we use max() to calculate MAX_ERR in x/y for linearity.
149
150        About metric note:
151            We show tuples instead of percentages if the metrics values are
152            percentages. This is because such a tuple unveils more information
153            (i.e., the values of the nominator and the denominator) than a mere
154            percentage value. For examples,
155
156            1f-click miss rate (%):
157                one_finger_physical_click.center (20130710_063117) : (0, 1)
158                  the tuple means (the number of missed clicks, total clicks)
159
160            intervals > xxx ms (%)
161                one_finger_tap.top_left (20130710_063117) : (1, 6)
162                  the tuple means (the number of long intervals, total packets)
163        """
164        # stat_functions include: max, average,
165        #                         pct_by_numbers, pct_by_missed_numbers,
166        #                         pct_by_cases_neq, and pct_by_cases_less
167        average = lambda lst: float(sum(lst)) / len(lst)
168        get_sums = lambda lst: [sum(count) for count in zip(*lst)]
169        _pct = lambda lst: float(lst[0]) / lst[1] * 100
170        # The following lambda function is used to compute the missed pct of
171        #
172        #       '(clicks with correct finger IDs, actual clicks)'
173        #
174        # In some cases when the number of actual clicks is 0, there are no
175        # missed finger IDs. So just return 0 for this special case to prevent
176        # the devision by 0 error.
177        _missed_pct = lambda lst: (float(lst[1] - lst[0]) / lst[1] * 100
178                                   if lst[1] != 0 else 0)
179
180        # pct by numbers: lst means [(incorrect number, total number), ...]
181        #  E.g., lst = [(2, 10), (0, 10), (0, 10), (0, 10)]
182        #  pct_by_numbers would be (2 + 0 + 0 + 0) / (10 + 10 + 10 + 10) * 100%
183        pct_by_numbers = lambda lst: _pct(get_sums(lst))
184
185        # pct by misssed numbers: lst means
186        #                         [(actual number, expected number), ...]
187        #  E.g., lst = [(0, 1), (1, 1), (1, 1), (1, 1)]
188        #  pct_by_missed_numbers would be
189        #       0 + 1 + 1 + 1  = 3
190        #       1 + 1 + 1 + 1  = 4
191        #       missed pct = (4 - 3) / 4 * 100% = 25%
192        pct_by_missed_numbers = lambda lst: _missed_pct(get_sums(lst))
193
194        # pct of incorrect cases in [(acutal_value, expected_value), ...]
195        #   E.g., lst = [(1, 1), (0, 1), (1, 1), (1, 1)]
196        #   pct_by_cases_neq would be 1 / 4 * 100%
197        # This is used for CountTrackingIDValidator
198        pct_by_cases_neq = lambda lst: _pct(
199                [len([pair for pair in lst if pair[0] != pair[1]]), len(lst)])
200
201        # pct of incorrect cases in [(acutal_value, min expected_value), ...]
202        #   E.g., lst = [(3, 3), (4, 3)]
203        #     pct_by_cases_less would be 0 / 2 * 100%
204        #   E.g., lst = [(2, 3), (5, 3)]
205        #     pct_by_cases_less would be 1 / 2 * 100%
206        # This is used for CountPacketsIDValidator and PinchValidator
207        pct_by_cases_less = lambda lst: _pct(
208                [len([pair for pair in lst if pair[0] < pair[1]]), len(lst)])
209
210        self.max_report_interval_str = '%.2f' % conf.max_report_interval
211
212        # A dictionary from metric attribute to its properties:
213        #    {metric_attr: (template,
214        #                   name_variations,
215        #                   description,
216        #                   metric_note,
217        #                   stat_func)
218        #    }
219        # Ordered by validators
220        self.raw_metrics_props = {
221            # Count Packets Validator
222            'COUNT_PACKETS': (
223                'pct of incorrect cases (%)--packets',
224                None,
225                'an incorrect case is one where a swipe has less than '
226                    '3 packets reported',
227                '(actual number of packets, expected number of packets)',
228                pct_by_cases_less),
229            # Count TrackingID Validator
230            'TID': (
231                'pct of incorrect cases (%)--tids',
232                None,
233                'an incorrect case is one where there are an incorrect number '
234                    'of fingers detected',
235                '(actual tracking IDs, expected tracking IDs)',
236                pct_by_cases_neq),
237            # Drag Latency Validator
238            'AVG_LATENCY': (
239                'average latency (ms)',
240                None,
241                'The average drag-latency in milliseconds',
242                None,
243                average),
244            # Drumroll Validator
245            'CIRCLE_RADIUS': (
246                'circle radius (mm)',
247                None,
248                'the max radius of enclosing circles of tapping points',
249                None,
250                max),
251            # Hysteresis Validator
252            'MAX_INIT_GAP_RATIO': (
253                'max init gap ratio',
254                None,
255                'the max ratio of dist(p0,p1) / dist(p1,p2)',
256                None,
257                max),
258            'AVE_INIT_GAP_RATIO': (
259                'ave init gap ratio',
260                None,
261                'the average ratio of dist(p0,p1) / dist(p1,p2)',
262                None,
263                average),
264            # Linearity Validator
265            'MAX_ERR': (
266                'max error in {} (mm)',
267                AXIS.LIST,
268                'The max err of all samples',
269                None,
270                max),
271            'RMS_ERR': (
272                'rms error in {} (mm)',
273                AXIS.LIST,
274                'The mean of all rms means of all trials',
275                None,
276                average),
277            # MTB Sanity Validator
278            'MTB_SANITY_ERR': (
279                'pct of MTB errors (%)',
280                None,
281                'pct of MTB errors',
282                '(MTB errors, expected errors)',
283                pct_by_cases_neq),
284            # No Ghost Finger Validator
285            'GHOST_FINGERS': (
286                'pct of ghost fingers (%)',
287                None,
288                'pct of ghost fingers',
289                '(ghost fingers, expected fingers)',
290                pct_by_cases_neq),
291            # Physical Click Validator
292            'CLICK_CHECK_CLICK': (
293                '{}f-click miss rate (%)',
294                conf.fingers_physical_click,
295                'the pct of finger IDs w/o a click',
296                '(acutual clicks, expected clicks)',
297                pct_by_missed_numbers),
298            'CLICK_CHECK_TIDS': (
299                '{}f-click w/o finger IDs (%)',
300                conf.fingers_physical_click,
301                'the pct of clicks w/o correct finger IDs',
302                '(clicks with correct finger IDs, actual clicks)',
303                pct_by_missed_numbers),
304            # Pinch Validator
305            'PINCH': (
306                'pct of incorrect cases (%)--pinch',
307                None,
308                'pct of incorrect cases over total cases',
309                '(actual relative motion (px), expected relative motion (px))',
310                pct_by_cases_less),
311            # Range Validator
312            'RANGE': (
313                '{} edge not reached (mm)',
314                ['left', 'right', 'top', 'bottom'],
315                'Min unreachable distance',
316                None,
317                max),
318            # Report Rate Validator
319            'LONG_INTERVALS': (
320                'pct of large intervals (%)',
321                None,
322                'pct of intervals larger than expected',
323                '(the number of long intervals, total packets)',
324                pct_by_numbers),
325            'AVE_TIME_INTERVAL': (
326                'average time interval (ms)',
327                None,
328                'the average of report intervals',
329                None,
330                average),
331            'MAX_TIME_INTERVAL': (
332                'max time interval (ms)',
333                None,
334                'the max report interval',
335                None,
336                max),
337            # Stationary Finger Validator
338            'MAX_DISTANCE': (
339                'max distance (mm)',
340                None,
341                'max distance of any two points from any run',
342                None,
343                max),
344        }
345
346        # Set the metric attribute to its template
347        #   E.g., self.MAX_ERR = 'max error in {} (mm)'
348        for key, props in self.raw_metrics_props.items():
349            template = props[0]
350            setattr(self, key, template)
351
352    def _derive_metrics_props(self):
353        """Expand the metric name templates to the metric names, and then
354        derive the expanded metrics_props.
355
356        In _init_raw_metrics_props():
357            The raw_metrics_props is defined as:
358                'MAX_ERR': (
359                    'max error in {} (mm)',             # template
360                    ['x', 'y'],                         # name variations
361                    'The max err of all samples',       # description
362                    max),                               # stat_func
363                ...
364
365            By expanding the template with its corresponding name variations,
366            the names related with MAX_ERR will be:
367                'max error in x (mm)', and
368                'max error in y (mm)'
369
370        Here we are going to derive metrics_props as:
371                metrics_props = {
372                    'max error in x (mm)':
373                        MetricProps('The max err of all samples', max),
374                    ...
375                }
376        """
377        self.metrics_props = {}
378        for raw_props in self.raw_metrics_props.values():
379            template, name_variations, description, note, stat_func = raw_props
380            metric_props = MetricProps(description, note, stat_func)
381            if name_variations:
382                # Expand the template with every variations.
383                #   E.g., template = 'max error in {} (mm)' is expanded to
384                #         name = 'max error in x (mm)'
385                for variation in name_variations:
386                    name = template.format(variation)
387                    self.metrics_props[name] = metric_props
388            else:
389                # Otherwise, the template is already the name.
390                #   E.g., the template 'max distance (mm)' is same as the name.
391                self.metrics_props[template] = metric_props
392
393
394class ValidatorLog:
395    """A class handling the logs reported by validators."""
396    def __init__(self):
397        self.name = None
398        self.details = []
399        self.criteria = None
400        self.score = None
401        self.metrics = []
402        self.error = None
403
404    def reset(self):
405        """Reset all attributes."""
406        self.details = []
407        self.score = None
408        self.metrics = []
409        self.error = None
410
411    def insert_details(self, msg):
412        """Insert a msg into the details."""
413        self.details.append(msg)
414
415
416class GestureLog:
417    """A class handling the logs related with a gesture."""
418    def __init__(self):
419        self.name = None
420        self.variation = None
421        self.prompt = None
422        self.vlogs = []
423
424
425class RoundLog:
426    """Manipulate the test result log generated in a single round."""
427    def __init__(self, test_version, fw=None, round_name=None):
428        self._test_version = test_version
429        self._fw = fw
430        self._round_name = round_name
431        self._glogs = []
432
433    def dump(self, filename):
434        """Dump the log to the specified filename."""
435        try:
436            with open(filename, 'w') as log_file:
437                pickle.dump([self._fw, self._round_name, self._test_version,
438                             self._glogs], log_file)
439        except Exception, e:
440            msg = 'Error in dumping to the log file (%s): %s' % (filename, e)
441            print_and_exit(msg)
442
443    @staticmethod
444    def load(filename):
445        """Load the log from the pickle file."""
446        try:
447            with open(filename) as log_file:
448                return pickle.load(log_file)
449        except Exception, e:
450            msg = 'Error in loading the log file (%s): %s' % (filename, e)
451            print_and_exit(msg)
452
453    def insert_glog(self, glog):
454        """Insert the gesture log into the round log."""
455        if glog.vlogs:
456            self._glogs.append(glog)
457
458
459class StatisticsScores:
460    """A statistics class to compute the average, ssd, and count of
461    aggregate scores.
462    """
463    def __init__(self, scores):
464        self.all_data = ()
465        if scores:
466            self.average = np.average(np.array(scores))
467            self.ssd = _calc_sample_standard_deviation(scores)
468            self.count = len(scores)
469            self.all_data = (self.average, self.ssd, self.count)
470
471
472class StatisticsMetrics:
473    """A statistics class to compute the statistics including the min, max, or
474    average of aggregate metrics.
475    """
476
477    def __init__(self, metrics):
478        """Collect all values for every metric.
479
480        @param metrics: a list of Metric objects.
481        """
482        # metrics_values: the raw metrics values
483        self.metrics_values = defaultdict(list)
484        self.metrics_dict = defaultdict(list)
485        for metric in metrics:
486            self.metrics_values[metric.name].append(metric.value)
487            self.metrics_dict[metric.name].append(metric)
488
489        # Calculate the statistics of metrics using corresponding stat functions
490        self._calc_statistics(MetricNameProps().metrics_props)
491
492    def _calc_statistics(self, metrics_props):
493        """Calculate the desired statistics for every metric.
494
495        @param metrics_props: a dictionary mapping a metric name to a
496                metric props including the description and stat_func
497        """
498        self.metrics_props = metrics_props
499        self.stats_values = {}
500        for metric_name, values in self.metrics_values.items():
501            assert metric_name in metrics_props, (
502                    'The metric name "%s" cannot be found.' % metric_name)
503            stat_func = metrics_props[metric_name].stat_func
504            self.stats_values[metric_name] = stat_func(values)
505
506
507class TestResult:
508    """A class includes the statistics of the score and the metrics."""
509    def __init__(self, scores, metrics):
510        self.stat_scores = StatisticsScores(scores)
511        self.stat_metrics = StatisticsMetrics(metrics)
512
513
514class SimpleTable:
515    """A very simple data table."""
516    def __init__(self):
517        """This initializes a simple table."""
518        self._table = defaultdict(list)
519
520    def insert(self, key, value):
521        """Insert a row. If the key exists already, the value is appended."""
522        self._table[key].append(value)
523        debug_print('    key: %s' % str(key))
524
525    def search(self, key):
526        """Search rows with the specified key.
527
528        A key is a list of attributes.
529        If any attribute is None, it means no need to match this attribute.
530        """
531        match = lambda i, j: i == j or j is None
532        return filter(lambda (k, vlog): all(map(match, k, key)),
533                      self._table.items())
534
535    def items(self):
536        """Return the table items."""
537        return self._table.items()
538
539
540class SummaryLog:
541    """A class to manipulate the summary logs.
542
543    A summary log may consist of result logs of different firmware versions
544    where every firmware version may consist of multiple rounds.
545    """
546    def __init__(self, log_dir, segment_weights, validator_weights,
547                 individual_round_flag, debug_flag):
548        self.log_dir = log_dir
549        self.segment_weights = segment_weights
550        self.validator_weights = validator_weights
551        self.individual_round_flag = individual_round_flag
552        _setup_debug(debug_flag)
553        self._read_logs()
554        self.ext_validator_weights = {}
555        for fw, validators in self.fw_validators.items():
556            self.ext_validator_weights[fw] = \
557                    self._compute_extended_validator_weight(validators)
558
559    def _get_firmware_version(self, filename):
560        """Get the firmware version from the given filename."""
561        return filename.split('-')[2]
562
563    def _read_logs(self):
564        """Read the result logs in the specified log directory."""
565        # Get logs in the log_dir or its sub-directories.
566        log_filenames = glob.glob(os.path.join(self.log_dir, '*.log'))
567        if not log_filenames:
568            log_filenames = glob.glob(os.path.join(self.log_dir, '*', '*.log'))
569
570        if not log_filenames:
571            err_msg = 'Error: no log files in the test result directory: %s'
572            print_and_exit(err_msg % self.log_dir)
573
574        self.log_table = SimpleTable()
575        self.fws = set()
576        self.gestures = set()
577        # fw_validators keeps track of the validators of every firmware
578        self.fw_validators = defaultdict(set)
579
580        for i, log_filename in enumerate(log_filenames):
581            round_no = i if self.individual_round_flag else None
582            self._add_round_log(log_filename, round_no)
583
584        # Convert set to list below
585        self.fws = sorted(list(self.fws))
586        self.gestures = sorted(list(self.gestures))
587        # Construct validators by taking the union of the validators of
588        # all firmwares.
589        self.validators = sorted(list(set.union(*self.fw_validators.values())))
590
591        for fw in self.fws:
592            self.fw_validators[fw] = sorted(list(self.fw_validators[fw]))
593
594    def _add_round_log(self, log_filename, round_no):
595        """Add the round log, decompose the validator logs, and build
596        a flat summary log.
597        """
598        log_data = RoundLog.load(log_filename)
599        if len(log_data) == 3:
600            fw, round_name, glogs = log_data
601            self.test_version = 'test_version: NA'
602        elif len(log_data) == 4:
603            fw, round_name, self.test_version, glogs = log_data
604        else:
605            print 'Error: the log format is unknown.'
606            sys.exit(1)
607
608        if round_no is not None:
609            fw = '%s_%d' % (fw, round_no)
610        self.fws.add(fw)
611        debug_print('  fw(%s) round(%s)' % (fw, round_name))
612
613        # Iterate through every gesture_variation of the round log,
614        # and generate a flat dictionary of the validator logs.
615        for glog in glogs:
616            self.gestures.add(glog.name)
617            for vlog in glog.vlogs:
618                self.fw_validators[fw].add(vlog.name)
619                key = (fw, round_name, glog.name, glog.variation, vlog.name)
620                self.log_table.insert(key, vlog)
621
622    def _compute_extended_validator_weight(self, validators):
623        """Compute extended validator weight from validator weight and segment
624        weight. The purpose is to merge the weights of split validators, e.g.
625        Linearity(*)Validator, so that their weights are not counted multiple
626        times.
627
628        Example:
629          validators = ['CountTrackingIDValidator',
630                        'Linearity(BothEnds)Validator',
631                        'Linearity(Middle)Validator',
632                        'NoGapValidator']
633
634          Note that both names of the validators
635                'Linearity(BothEnds)Validator' and
636                'Linearity(Middle)Validator'
637          are created at run time from LinearityValidator and use
638          the relative weights defined by segment_weights.
639
640          validator_weights = {'CountTrackingIDValidator': 12,
641                               'LinearityValidator': 10,
642                               'NoGapValidator': 10}
643
644          segment_weights = {'Middle': 0.7,
645                             'BothEnds': 0.3}
646
647          split_validator = {'Linearity': ['BothEnds', 'Middle'],}
648
649          adjusted_weight of Lineary(*)Validator:
650            Linearity(BothEnds)Validator = 0.3 / (0.3 + 0.7) * 10 = 3
651            Linearity(Middle)Validator =   0.7 / (0.3 + 0.7) * 10 = 7
652
653          extended_validator_weights: {'CountTrackingIDValidator': 12,
654                                       'Linearity(BothEnds)Validator': 3,
655                                       'Linearity(Middle)Validator': 7,
656                                       'NoGapValidator': 10}
657        """
658        extended_validator_weights = {}
659        split_validator = {}
660
661        # Copy the base validator weight into extended_validator_weights.
662        # For the split validators, collect them in split_validator.
663        for v in validators:
664            base_name, segment = val.get_base_name_and_segment(v)
665            if segment is None:
666                # It is a base validator. Just copy it into the
667                # extended_validaotr_weight dict.
668                extended_validator_weights[v] = self.validator_weights[v]
669            else:
670                # It is a derived validator, e.g., Linearity(BothEnds)Validator
671                # Needs to compute its adjusted weight.
672
673                # Initialize the split_validator for this base_name if not yet.
674                if split_validator.get(base_name) is None:
675                    split_validator[base_name] = []
676
677                # Append this segment name so that we know all segments for
678                # the base_name.
679                split_validator[base_name].append(segment)
680
681        # Compute the adjusted weight for split_validator
682        for base_name in split_validator:
683            name = val.get_validator_name(base_name)
684            weight_list = [self.segment_weights[segment]
685                           for segment in split_validator[base_name]]
686            weight_sum = sum(weight_list)
687            for segment in split_validator[base_name]:
688                derived_name = val.get_derived_name(name, segment)
689                adjusted_weight = (self.segment_weights[segment] / weight_sum *
690                                   self.validator_weights[name])
691                extended_validator_weights[derived_name] = adjusted_weight
692
693        return extended_validator_weights
694
695    def get_result(self, fw=None, round=None, gesture=None, variation=None,
696                   validators=None):
697        """Get the result statistics of a validator which include both
698        the score and the metrics.
699
700        If validators is a list, every validator in the list is used to query
701        the log table, and all results are merged to get the final result.
702        For example, both StationaryFingerValidator and StationaryTapValidator
703        inherit StationaryValidator. The results of those two extended classes
704        will be merged into StationaryValidator.
705        """
706        if not isinstance(validators, list):
707            validators = [validators,]
708
709        rows = []
710        for validator in validators:
711            key = (fw, round, gesture, variation, validator)
712            rows.extend(self.log_table.search(key))
713
714        scores = [vlog.score for _key, vlogs in rows for vlog in vlogs]
715        metrics = [metric.insert_key(_key) for _key, vlogs in rows
716                                               for vlog in vlogs
717                                                   for metric in vlog.metrics]
718        return TestResult(scores, metrics)
719
720    def get_final_weighted_average(self):
721        """Calculate the final weighted average."""
722        weighted_average = {}
723        # for fw in self.fws:
724        for fw, validators in self.fw_validators.items():
725            scores = [self.get_result(fw=fw, validators=val).stat_scores.average
726                      for val in validators]
727            _, weights = zip(*sorted(self.ext_validator_weights[fw].items()))
728            weighted_average[fw] = np.average(scores, weights=weights)
729        return weighted_average
730