libcxx/utils/parse-google-benchmark-results

#!/usr/bin/env python3

import argparse
import csv
import json
import sys

def main(argv):
    parser = argparse.ArgumentParser(
        prog='parse-google-benchmark-results',
        description='Parse Google Benchmark result files (in JSON format) into CSV or LNT compatible output.')
    parser.add_argument('filename', type=argparse.FileType('r'), nargs='+',
        help='One of more JSON files to extract the results from. The results parsed from each '
             'file are concatenated together.')
    parser.add_argument('--timing', type=str, choices=['real_time', 'cpu_time'], default='real_time',
        help='The timing to extract from the Google Benchmark results. This can either be the '
             '"real time" or the "CPU time". Default is "real time".')
    parser.add_argument('--output-format', type=str, choices=['csv', 'lnt'], default='csv',
        help='The desired output format for the data. `csv` is CSV format and `lnt` is a format compatible with '
             '`lnt importreport` (see https://llvm.org/docs/lnt/importing_data.html#importing-data-in-a-text-file).')
    args = parser.parse_args(argv)

    # Parse the data from all files, aggregating the results
    headers = ['Benchmark', args.timing]
    rows = []
    for file in args.filename:
        js = json.load(file)
        for bm in js['benchmarks']:
            if args.timing not in bm:
                raise RuntimeError(f'Benchmark does not contain key for {args.timing}: {bm}')
            row = [bm['name'], bm[args.timing]]
            rows.append(row)

    # Print the results in the right format
    if args.output_format == 'csv':
        writer = csv.writer(sys.stdout)
        writer.writerow(headers)
        for row in rows:
            writer.writerow(row)
    elif args.output_format == 'lnt':
        benchmark = headers.index('Benchmark')
        time = headers.index(args.timing)
        for row in rows:
            # LNT format uses '.' to separate the benchmark name from the metric, and ' '
            # to separate the benchmark name + metric from the numerical value. Escape both.
            escaped = row[benchmark].replace(".", "_").replace(" ", "_")
            print(f'{escaped}.execution_time {row[time]}')

if __name__ == '__main__':
    main(sys.argv[1:])