view contrib/perf-utils/search-discovery-case @ 48178:f12a19d03d2c

fix: reduce number of tool executions

By grouping together (path, ctx) pairs according to the inputs they would
provide to fixer tools, we can deduplicate executions of fixer tools to
significantly reduce the amount of time spent running slow tools.

This change does not handle clean files in the working copy, which could still
be deduplicated against the files in the checked out commit. It's a little
harder to do that because the filerev is not available in the workingfilectx
(and it doesn't exist for added files).

Anecdotally, this change makes some real uses cases at Google 10x faster. I
think we were originally hesitant to do this because the benefits weren't
obvious, and implementing it efficiently is kind of tricky. If we simply
memoized the formatter execution function, we would be keeping tons of file
content in memory.

Also included is a regression test for a corner case that I broke with my first
attempt at optimizing this code.

Differential Revision: https://phab.mercurial-scm.org/D11280

#!/usr/bin/env python3
# Search for interesting discovery instance
#
#  search-discovery-case REPO [REPO]…
#
# This use a subsetmaker extension (next to this script) to generate a steam of
# random discovery instance. When interesting case are discovered, information
# about them are print on the stdout.
from __future__ import print_function

import json
import os
import queue
import random
import signal
import subprocess
import sys
import threading

this_script = os.path.abspath(sys.argv[0])
this_dir = os.path.dirname(this_script)
hg_dir = os.path.join(this_dir, '..', '..')
HG_REPO = os.path.normpath(hg_dir)
HG_BIN = os.path.join(HG_REPO, 'hg')

JOB = int(os.environ.get('NUMBER_OF_PROCESSORS', 8))


SLICING = ('scratch', 'randomantichain', 'rev')


def nb_revs(repo_path):
    cmd = [
        HG_BIN,
        '--repository',
        repo_path,
        'log',
        '--template',
        '{rev}',
        '--rev',
        'tip',
    ]
    s = subprocess.Popen(cmd, stdout=subprocess.PIPE)
    out, err = s.communicate()
    return int(out)


repos = []
for repo in sys.argv[1:]:
    size = nb_revs(repo)
    repos.append((repo, size))


def pick_one(repo):
    pick = random.choice(SLICING)
    seed = random.randint(0, 100000)
    if pick == 'scratch':
        start = int(repo[1] * 0.3)
        end = int(repo[1] * 0.7)
        nb = random.randint(start, end)
        return ('scratch', nb, seed)
    elif pick == 'randomantichain':
        return ('randomantichain', seed)
    elif pick == 'rev':
        start = int(repo[1] * 0.3)
        end = int(repo[1])
        rev = random.randint(start, end)
        return ('rev', rev)
    else:
        assert False


done = threading.Event()
cases = queue.Queue(maxsize=10 * JOB)
results = queue.Queue()


def worker():
    while not done.is_set():
        c = cases.get()
        if c is None:
            return
        try:
            res = process(c)
            results.put((c, res))
        except Exception as exc:
            print('processing-failed: %s %s' % (c, exc), file=sys.stderr)
        c = (c[0], c[2], c[1])
        try:
            res = process(c)
            results.put((c, res))
        except Exception as exc:
            print('processing-failed: %s %s' % (c, exc), file=sys.stderr)


SUBSET_PATH = os.path.join(HG_REPO, 'contrib', 'perf-utils', 'subsetmaker.py')


CMD_BASE = (
    HG_BIN,
    'debugdiscovery',
    '--template',
    'json',
    '--config',
    'extensions.subset=%s' % SUBSET_PATH,
)
#    '--local-as-revs "$left" --local-as-revs "$right"'
#    > /data/discovery-references/results/disco-mozilla-unified-$1-$2.txt
#        )


def to_revsets(case):
    t = case[0]
    if t == 'scratch':
        return 'not scratch(all(), %d, "%d")' % (case[1], case[2])
    elif t == 'randomantichain':
        return '::randomantichain(all(), "%d")' % case[1]
    elif t == 'rev':
        return '::%d' % case[1]
    else:
        assert False


def process(case):
    (repo, left, right) = case
    cmd = list(CMD_BASE)
    cmd.append('-R')
    cmd.append(repo[0])
    cmd.append('--local-as-revs')
    cmd.append(to_revsets(left))
    cmd.append('--remote-as-revs')
    cmd.append(to_revsets(right))
    s = subprocess.Popen(cmd, stdout=subprocess.PIPE)
    out, err = s.communicate()
    return json.loads(out)[0]


def interesting_boundary(res):
    """check if a case is interesting or not

    For now we are mostly interrested in case were we do multiple roundstrip
    and where the boundary is somewhere in the middle of the undecided set.

    Ideally, we would make this configurable, but this is not a focus for now

    return None or (round-trip, undecided-common, undecided-missing)
    """
    roundtrips = res["total-roundtrips"]
    if roundtrips <= 1:
        return None
    undecided_common = res["nb-ini_und-common"]
    undecided_missing = res["nb-ini_und-missing"]
    if undecided_common == 0:
        return None
    if undecided_missing == 0:
        return None
    return (roundtrips, undecided_common, undecided_missing)


def end(*args, **kwargs):
    done.set()


def format_case(case):
    return '-'.join(str(s) for s in case)


signal.signal(signal.SIGINT, end)

for i in range(JOB):
    threading.Thread(target=worker).start()

nb_cases = 0
while not done.is_set():
    repo = random.choice(repos)
    left = pick_one(repo)
    right = pick_one(repo)
    cases.put((repo, left, right))
    while not results.empty():
        # results has a single reader so this is fine
        c, res = results.get_nowait()
        boundary = interesting_boundary(res)
        if boundary is not None:
            print(c[0][0], format_case(c[1]), format_case(c[2]), *boundary)
            sys.stdout.flush()

    nb_cases += 1
    if not nb_cases % 100:
        print('[%d cases generated]' % nb_cases, file=sys.stderr)

for i in range(JOB):
    try:
        cases.put_nowait(None)
    except queue.Full:
        pass

print('[%d cases generated]' % nb_cases, file=sys.stderr)
print('[ouput generation is over]' % nb_cases, file=sys.stderr)