poky/scripts/lib/build_perf/report.py

#
# Copyright (c) 2017, Intel Corporation.
#
# This program is free software; you can redistribute it and/or modify it
# under the terms and conditions of the GNU General Public License,
# version 2, as published by the Free Software Foundation.
#
# This program is distributed in the hope it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
# more details.
#
"""Handling of build perf test reports"""
from collections import OrderedDict, Mapping, namedtuple
from datetime import datetime, timezone
from numbers import Number
from statistics import mean, stdev, variance


AggregateTestData = namedtuple('AggregateTestData', ['metadata', 'results'])


def isofmt_to_timestamp(string):
    """Convert timestamp string in ISO 8601 format into unix timestamp"""
    if '.' in string:
        dt = datetime.strptime(string, '%Y-%m-%dT%H:%M:%S.%f')
    else:
        dt = datetime.strptime(string, '%Y-%m-%dT%H:%M:%S')
    return dt.replace(tzinfo=timezone.utc).timestamp()


def metadata_xml_to_json(elem):
    """Convert metadata xml into JSON format"""
    assert elem.tag == 'metadata', "Invalid metadata file format"

    def _xml_to_json(elem):
        """Convert xml element to JSON object"""
        out = OrderedDict()
        for child in elem.getchildren():
            key = child.attrib.get('name', child.tag)
            if len(child):
                out[key] = _xml_to_json(child)
            else:
                out[key] = child.text
        return out
    return _xml_to_json(elem)


def results_xml_to_json(elem):
    """Convert results xml into JSON format"""
    rusage_fields = ('ru_utime', 'ru_stime', 'ru_maxrss', 'ru_minflt',
                     'ru_majflt', 'ru_inblock', 'ru_oublock', 'ru_nvcsw',
                     'ru_nivcsw')
    iostat_fields = ('rchar', 'wchar', 'syscr', 'syscw', 'read_bytes',
                     'write_bytes', 'cancelled_write_bytes')

    def _read_measurement(elem):
        """Convert measurement to JSON"""
        data = OrderedDict()
        data['type'] = elem.tag
        data['name'] = elem.attrib['name']
        data['legend'] = elem.attrib['legend']
        values = OrderedDict()

        # SYSRES measurement
        if elem.tag == 'sysres':
            for subel in elem:
                if subel.tag == 'time':
                    values['start_time'] = isofmt_to_timestamp(subel.attrib['timestamp'])
                    values['elapsed_time'] = float(subel.text)
                elif subel.tag == 'rusage':
                    rusage = OrderedDict()
                    for field in rusage_fields:
                        if 'time' in field:
                            rusage[field] = float(subel.attrib[field])
                        else:
                            rusage[field] = int(subel.attrib[field])
                    values['rusage'] = rusage
                elif subel.tag == 'iostat':
                    values['iostat'] = OrderedDict([(f, int(subel.attrib[f]))
                        for f in iostat_fields])
                elif subel.tag == 'buildstats_file':
                    values['buildstats_file'] = subel.text
                else:
                    raise TypeError("Unknown sysres value element '{}'".format(subel.tag))
        # DISKUSAGE measurement
        elif elem.tag == 'diskusage':
            values['size'] = int(elem.find('size').text)
        else:
            raise Exception("Unknown measurement tag '{}'".format(elem.tag))
        data['values'] = values
        return data

    def _read_testcase(elem):
        """Convert testcase into JSON"""
        assert elem.tag == 'testcase', "Expecting 'testcase' element instead of {}".format(elem.tag)

        data = OrderedDict()
        data['name'] = elem.attrib['name']
        data['description'] = elem.attrib['description']
        data['status'] = 'SUCCESS'
        data['start_time'] = isofmt_to_timestamp(elem.attrib['timestamp'])
        data['elapsed_time'] = float(elem.attrib['time'])
        measurements = OrderedDict()

        for subel in elem.getchildren():
            if subel.tag == 'error' or subel.tag == 'failure':
                data['status'] = subel.tag.upper()
                data['message'] = subel.attrib['message']
                data['err_type'] = subel.attrib['type']
                data['err_output'] = subel.text
            elif subel.tag == 'skipped':
                data['status'] = 'SKIPPED'
                data['message'] = subel.text
            else:
                measurements[subel.attrib['name']] = _read_measurement(subel)
        data['measurements'] = measurements
        return data

    def _read_testsuite(elem):
        """Convert suite to JSON"""
        assert elem.tag == 'testsuite', \
                "Expecting 'testsuite' element instead of {}".format(elem.tag)

        data = OrderedDict()
        if 'hostname' in elem.attrib:
            data['tester_host'] = elem.attrib['hostname']
        data['start_time'] = isofmt_to_timestamp(elem.attrib['timestamp'])
        data['elapsed_time'] = float(elem.attrib['time'])
        tests = OrderedDict()

        for case in elem.getchildren():
            tests[case.attrib['name']] = _read_testcase(case)
        data['tests'] = tests
        return data

    # Main function
    assert elem.tag == 'testsuites', "Invalid test report format"
    assert len(elem) == 1, "Too many testsuites"

    return _read_testsuite(elem.getchildren()[0])


def aggregate_metadata(metadata):
    """Aggregate metadata into one, basically a sanity check"""
    mutable_keys = ('pretty_name', 'version_id')

    def aggregate_obj(aggregate, obj, assert_str=True):
        """Aggregate objects together"""
        assert type(aggregate) is type(obj), \
                "Type mismatch: {} != {}".format(type(aggregate), type(obj))
        if isinstance(obj, Mapping):
            assert set(aggregate.keys()) == set(obj.keys())
            for key, val in obj.items():
                aggregate_obj(aggregate[key], val, key not in mutable_keys)
        elif isinstance(obj, list):
            assert len(aggregate) == len(obj)
            for i, val in enumerate(obj):
                aggregate_obj(aggregate[i], val)
        elif not isinstance(obj, str) or (isinstance(obj, str) and assert_str):
            assert aggregate == obj, "Data mismatch {} != {}".format(aggregate, obj)

    if not metadata:
        return {}

    # Do the aggregation
    aggregate = metadata[0].copy()
    for testrun in metadata[1:]:
        aggregate_obj(aggregate, testrun)
    aggregate['testrun_count'] = len(metadata)
    return aggregate


def aggregate_data(data):
    """Aggregate multiple test results JSON structures into one"""

    mutable_keys = ('status', 'message', 'err_type', 'err_output')

    class SampleList(list):
        """Container for numerical samples"""
        pass

    def new_aggregate_obj(obj):
        """Create new object for aggregate"""
        if isinstance(obj, Number):
            new_obj = SampleList()
            new_obj.append(obj)
        elif isinstance(obj, str):
            new_obj = obj
        else:
            # Lists and and dicts are kept as is
            new_obj = obj.__class__()
            aggregate_obj(new_obj, obj)
        return new_obj

    def aggregate_obj(aggregate, obj, assert_str=True):
        """Recursive "aggregation" of JSON objects"""
        if isinstance(obj, Number):
            assert isinstance(aggregate, SampleList)
            aggregate.append(obj)
            return

        assert type(aggregate) == type(obj), \
                "Type mismatch: {} != {}".format(type(aggregate), type(obj))
        if isinstance(obj, Mapping):
            for key, val in obj.items():
                if not key in aggregate:
                    aggregate[key] = new_aggregate_obj(val)
                else:
                    aggregate_obj(aggregate[key], val, key not in mutable_keys)
        elif isinstance(obj, list):
            for i, val in enumerate(obj):
                if i >= len(aggregate):
                    aggregate[key] = new_aggregate_obj(val)
                else:
                    aggregate_obj(aggregate[i], val)
        elif isinstance(obj, str):
            # Sanity check for data
            if assert_str:
                assert aggregate == obj, "Data mismatch {} != {}".format(aggregate, obj)
        else:
            raise Exception("BUG: unable to aggregate '{}' ({})".format(type(obj), str(obj)))

    if not data:
        return {}

    # Do the aggregation
    aggregate = data[0].__class__()
    for testrun in data:
        aggregate_obj(aggregate, testrun)
    return aggregate


class MeasurementVal(float):
    """Base class representing measurement values"""
    gv_data_type = 'number'

    def gv_value(self):
        """Value formatting for visualization"""
        if self != self:
            return "null"
        else:
            return self


class TimeVal(MeasurementVal):
    """Class representing time values"""
    quantity = 'time'
    gv_title = 'elapsed time'
    gv_data_type = 'timeofday'

    def hms(self):
        """Split time into hours, minutes and seconeds"""
        hhh = int(abs(self) / 3600)
        mmm = int((abs(self) % 3600) / 60)
        sss = abs(self) % 60
        return hhh, mmm, sss

    def __str__(self):
        if self != self:
            return "nan"
        hh, mm, ss = self.hms()
        sign = '-' if self < 0 else ''
        if hh > 0:
            return '{}{:d}:{:02d}:{:02.0f}'.format(sign, hh, mm, ss)
        elif mm > 0:
            return '{}{:d}:{:04.1f}'.format(sign, mm, ss)
        elif ss > 1:
            return '{}{:.1f} s'.format(sign, ss)
        else:
            return '{}{:.2f} s'.format(sign, ss)

    def gv_value(self):
        """Value formatting for visualization"""
        if self != self:
            return "null"
        hh, mm, ss = self.hms()
        return [hh, mm, int(ss), int(ss*1000) % 1000]


class SizeVal(MeasurementVal):
    """Class representing time values"""
    quantity = 'size'
    gv_title = 'size in MiB'
    gv_data_type = 'number'

    def __str__(self):
        if self != self:
            return "nan"
        if abs(self) < 1024:
            return '{:.1f} kiB'.format(self)
        elif abs(self) < 1048576:
            return '{:.2f} MiB'.format(self / 1024)
        else:
            return '{:.2f} GiB'.format(self / 1048576)

    def gv_value(self):
        """Value formatting for visualization"""
        if self != self:
            return "null"
        return self / 1024

def measurement_stats(meas, prefix=''):
    """Get statistics of a measurement"""
    if not meas:
        return {prefix + 'sample_cnt': 0,
                prefix + 'mean': MeasurementVal('nan'),
                prefix + 'stdev': MeasurementVal('nan'),
                prefix + 'variance': MeasurementVal('nan'),
                prefix + 'min': MeasurementVal('nan'),
                prefix + 'max': MeasurementVal('nan'),
                prefix + 'minus': MeasurementVal('nan'),
                prefix + 'plus': MeasurementVal('nan')}

    stats = {'name': meas['name']}
    if meas['type'] == 'sysres':
        val_cls = TimeVal
        values = meas['values']['elapsed_time']
    elif meas['type'] == 'diskusage':
        val_cls = SizeVal
        values = meas['values']['size']
    else:
        raise Exception("Unknown measurement type '{}'".format(meas['type']))
    stats['val_cls'] = val_cls
    stats['quantity'] = val_cls.quantity
    stats[prefix + 'sample_cnt'] = len(values)

    mean_val = val_cls(mean(values))
    min_val = val_cls(min(values))
    max_val = val_cls(max(values))

    stats[prefix + 'mean'] = mean_val
    if len(values) > 1:
        stats[prefix + 'stdev'] = val_cls(stdev(values))
        stats[prefix + 'variance'] = val_cls(variance(values))
    else:
        stats[prefix + 'stdev'] = float('nan')
        stats[prefix + 'variance'] = float('nan')
    stats[prefix + 'min'] = min_val
    stats[prefix + 'max'] = max_val
    stats[prefix + 'minus'] = val_cls(mean_val - min_val)
    stats[prefix + 'plus'] = val_cls(max_val - mean_val)

    return stats