view tests/test-remotefilelog-datapack.py @ 44363:f7459da77f23

nodemap: introduce an option to use mmap to read the nodemap mapping

The performance and memory benefit is much greater if we don't have to copy all
the data in memory for each information. So we introduce an option (on by
default) to read the data using mmap.

This changeset is the last one definition the API for index support nodemap
data. (they have to be able to use the mmaping).

Below are some benchmark comparing the best we currently have in 5.3 with the
final step of this series (using the persistent nodemap implementation in
Rust). The benchmark run `hg perfindex` with various revset and the following
variants:

Before:

 * do not use the persistent nodemap
 * use the CPython implementation of the index for nodemap
 * use mmapping of the changelog index

After:

 * use the MixedIndex Rust code, with the NodeTree object for nodemap access
   (still in review)
 * use the persistent nodemap data from disk
 * access the persistent nodemap data through mmap
 * use mmapping of the changelog index

The persistent nodemap greatly speed up  most operation on very large
repositories. Some of the previously very fast lookup end up a bit slower because
the persistent nodemap has to be setup. However the absolute slowdown is very
small and won't matters in the big picture.

Here are some numbers (in seconds) for the reference copy of mozilla-try:

Revset            Before   After     abs-change speedup

-10000:           0.004622 0.005532   0.000910  ×   0.83
-10:              0.000050 0.000132   0.000082  ×   0.37
tip               0.000052 0.000085   0.000033  ×   0.61
0 + (-10000:)     0.028222 0.005337  -0.022885  ×   5.29
0                 0.023521 0.000084  -0.023437  × 280.01
(-10000:) + 0     0.235539 0.005308  -0.230231  ×  44.37
(-10:) + :9       0.232883 0.000180  -0.232703  ×1293.79
(-10000:) + (:99) 0.238735 0.005358  -0.233377  ×  44.55
:99 + (-10000:)   0.317942 0.005593  -0.312349  ×  56.84
:9 + (-10:)       0.313372 0.000179  -0.313193  ×1750.68
:9                0.316450 0.000143  -0.316307  ×2212.93


On smaller repositories, the cost of nodemap related operation is not as big, so
the win is much more modest. Yet it helps shaving a handful of millisecond here
and there.

Here are some numbers (in seconds) for the reference copy of mercurial:

Revset            Before   After     abs-change speedup
-10:              0.000065 0.000097   0.000032  × 0.67
tip               0.000063 0.000078   0.000015  × 0.80
0                 0.000561 0.000079  -0.000482  × 7.10
-10000:           0.004609 0.003648  -0.000961  × 1.26
0 + (-10000:)     0.005023 0.003715  -0.001307  × 1.35
(-10:) + :9       0.002187 0.000108  -0.002079  ×20.25
(-10000:) + 0     0.006252 0.003716  -0.002536  × 1.68
(-10000:) + (:99) 0.006367 0.003707  -0.002660  × 1.71
:9 + (-10:)       0.003846 0.000110  -0.003736  ×34.96
:9                0.003854 0.000099  -0.003755  ×38.92
:99 + (-10000:)   0.007644 0.003778  -0.003866  × 2.02

Differential Revision: https://phab.mercurial-scm.org/D7894

#!/usr/bin/env python
from __future__ import absolute_import, print_function

import hashlib
import os
import random
import shutil
import stat
import struct
import sys
import tempfile
import time
import unittest

import silenttestrunner

# Load the local remotefilelog, not the system one
sys.path[0:0] = [os.path.join(os.path.dirname(__file__), '..')]
from mercurial.node import nullid
from mercurial import (
    pycompat,
    ui as uimod,
)
from hgext.remotefilelog import (
    basepack,
    constants,
    datapack,
)


class datapacktestsbase(object):
    def __init__(self, datapackreader, paramsavailable):
        self.datapackreader = datapackreader
        self.paramsavailable = paramsavailable

    def setUp(self):
        self.tempdirs = []

    def tearDown(self):
        for d in self.tempdirs:
            shutil.rmtree(d)

    def makeTempDir(self):
        tempdir = pycompat.bytestr(tempfile.mkdtemp())
        self.tempdirs.append(tempdir)
        return tempdir

    def getHash(self, content):
        return hashlib.sha1(content).digest()

    def getFakeHash(self):
        return b''.join(
            pycompat.bytechr(random.randint(0, 255)) for _ in range(20)
        )

    def createPack(self, revisions=None, packdir=None):
        if revisions is None:
            revisions = [(b"filename", self.getFakeHash(), nullid, b"content")]

        if packdir is None:
            packdir = self.makeTempDir()

        packer = datapack.mutabledatapack(uimod.ui(), packdir, version=2)

        for args in revisions:
            filename, node, base, content = args[0:4]
            # meta is optional
            meta = None
            if len(args) > 4:
                meta = args[4]
            packer.add(filename, node, base, content, metadata=meta)

        path = packer.close()
        return self.datapackreader(path)

    def _testAddSingle(self, content):
        """Test putting a simple blob into a pack and reading it out.
        """
        filename = b"foo"
        node = self.getHash(content)

        revisions = [(filename, node, nullid, content)]
        pack = self.createPack(revisions)
        if self.paramsavailable:
            self.assertEqual(
                pack.params.fanoutprefix, basepack.SMALLFANOUTPREFIX
            )

        chain = pack.getdeltachain(filename, node)
        self.assertEqual(content, chain[0][4])

    def testAddSingle(self):
        self._testAddSingle(b'')

    def testAddSingleEmpty(self):
        self._testAddSingle(b'abcdef')

    def testAddMultiple(self):
        """Test putting multiple unrelated blobs into a pack and reading them
        out.
        """
        revisions = []
        for i in range(10):
            filename = b"foo%d" % i
            content = b"abcdef%d" % i
            node = self.getHash(content)
            revisions.append((filename, node, self.getFakeHash(), content))

        pack = self.createPack(revisions)

        for filename, node, base, content in revisions:
            entry = pack.getdelta(filename, node)
            self.assertEqual((content, filename, base, {}), entry)

            chain = pack.getdeltachain(filename, node)
            self.assertEqual(content, chain[0][4])

    def testAddDeltas(self):
        """Test putting multiple delta blobs into a pack and read the chain.
        """
        revisions = []
        filename = b"foo"
        lastnode = nullid
        for i in range(10):
            content = b"abcdef%d" % i
            node = self.getHash(content)
            revisions.append((filename, node, lastnode, content))
            lastnode = node

        pack = self.createPack(revisions)

        entry = pack.getdelta(filename, revisions[0][1])
        realvalue = (revisions[0][3], filename, revisions[0][2], {})
        self.assertEqual(entry, realvalue)

        # Test that the chain for the final entry has all the others
        chain = pack.getdeltachain(filename, node)
        for i in range(10):
            content = b"abcdef%d" % i
            self.assertEqual(content, chain[-i - 1][4])

    def testPackMany(self):
        """Pack many related and unrelated objects.
        """
        # Build a random pack file
        revisions = []
        blobs = {}
        random.seed(0)
        for i in range(100):
            filename = b"filename-%d" % i
            filerevs = []
            for j in range(random.randint(1, 100)):
                content = b"content-%d" % j
                node = self.getHash(content)
                lastnode = nullid
                if len(filerevs) > 0:
                    lastnode = filerevs[random.randint(0, len(filerevs) - 1)]
                filerevs.append(node)
                blobs[(filename, node, lastnode)] = content
                revisions.append((filename, node, lastnode, content))

        pack = self.createPack(revisions)

        # Verify the pack contents
        for (filename, node, lastnode), content in sorted(blobs.items()):
            chain = pack.getdeltachain(filename, node)
            for entry in chain:
                expectedcontent = blobs[(entry[0], entry[1], entry[3])]
                self.assertEqual(entry[4], expectedcontent)

    def testPackMetadata(self):
        revisions = []
        for i in range(100):
            filename = b'%d.txt' % i
            content = b'put-something-here \n' * i
            node = self.getHash(content)
            meta = {
                constants.METAKEYFLAG: i ** 4,
                constants.METAKEYSIZE: len(content),
                b'Z': b'random_string',
                b'_': b'\0' * i,
            }
            revisions.append((filename, node, nullid, content, meta))
        pack = self.createPack(revisions)
        for name, node, x, content, origmeta in revisions:
            parsedmeta = pack.getmeta(name, node)
            # flag == 0 should be optimized out
            if origmeta[constants.METAKEYFLAG] == 0:
                del origmeta[constants.METAKEYFLAG]
            self.assertEqual(parsedmeta, origmeta)

    def testGetMissing(self):
        """Test the getmissing() api.
        """
        revisions = []
        filename = b"foo"
        lastnode = nullid
        for i in range(10):
            content = b"abcdef%d" % i
            node = self.getHash(content)
            revisions.append((filename, node, lastnode, content))
            lastnode = node

        pack = self.createPack(revisions)

        missing = pack.getmissing([(b"foo", revisions[0][1])])
        self.assertFalse(missing)

        missing = pack.getmissing(
            [(b"foo", revisions[0][1]), (b"foo", revisions[1][1])]
        )
        self.assertFalse(missing)

        fakenode = self.getFakeHash()
        missing = pack.getmissing(
            [(b"foo", revisions[0][1]), (b"foo", fakenode)]
        )
        self.assertEqual(missing, [(b"foo", fakenode)])

    def testAddThrows(self):
        pack = self.createPack()

        try:
            pack.add(b'filename', nullid, b'contents')
            self.assertTrue(False, "datapack.add should throw")
        except RuntimeError:
            pass

    def testBadVersionThrows(self):
        pack = self.createPack()
        path = pack.path + b'.datapack'
        with open(path, 'rb') as f:
            raw = f.read()
        raw = struct.pack('!B', 255) + raw[1:]
        os.chmod(path, os.stat(path).st_mode | stat.S_IWRITE)
        with open(path, 'wb+') as f:
            f.write(raw)

        try:
            self.datapackreader(pack.path)
            self.assertTrue(False, "bad version number should have thrown")
        except RuntimeError:
            pass

    def testMissingDeltabase(self):
        fakenode = self.getFakeHash()
        revisions = [(b"filename", fakenode, self.getFakeHash(), b"content")]
        pack = self.createPack(revisions)
        chain = pack.getdeltachain(b"filename", fakenode)
        self.assertEqual(len(chain), 1)

    def testLargePack(self):
        """Test creating and reading from a large pack with over X entries.
        This causes it to use a 2^16 fanout table instead."""
        revisions = []
        blobs = {}
        total = basepack.SMALLFANOUTCUTOFF + 1
        for i in pycompat.xrange(total):
            filename = b"filename-%d" % i
            content = filename
            node = self.getHash(content)
            blobs[(filename, node)] = content
            revisions.append((filename, node, nullid, content))

        pack = self.createPack(revisions)
        if self.paramsavailable:
            self.assertEqual(
                pack.params.fanoutprefix, basepack.LARGEFANOUTPREFIX
            )

        for (filename, node), content in blobs.items():
            actualcontent = pack.getdeltachain(filename, node)[0][4]
            self.assertEqual(actualcontent, content)

    def testPacksCache(self):
        """Test that we remember the most recent packs while fetching the delta
        chain."""

        packdir = self.makeTempDir()
        deltachains = []

        numpacks = 10
        revisionsperpack = 100

        for i in range(numpacks):
            chain = []
            revision = (b'%d' % i, self.getFakeHash(), nullid, b"content")

            for _ in range(revisionsperpack):
                chain.append(revision)
                revision = (
                    b'%d' % i,
                    self.getFakeHash(),
                    revision[1],
                    self.getFakeHash(),
                )

            self.createPack(chain, packdir)
            deltachains.append(chain)

        class testdatapackstore(datapack.datapackstore):
            # Ensures that we are not keeping everything in the cache.
            DEFAULTCACHESIZE = numpacks // 2

        store = testdatapackstore(uimod.ui(), packdir)

        random.shuffle(deltachains)
        for randomchain in deltachains:
            revision = random.choice(randomchain)
            chain = store.getdeltachain(revision[0], revision[1])

            mostrecentpack = next(iter(store.packs), None)
            self.assertEqual(
                mostrecentpack.getdeltachain(revision[0], revision[1]), chain
            )

            self.assertEqual(randomchain.index(revision) + 1, len(chain))

    # perf test off by default since it's slow
    def _testIndexPerf(self):
        random.seed(0)
        print("Multi-get perf test")
        packsizes = [
            100,
            10000,
            100000,
            500000,
            1000000,
            3000000,
        ]
        lookupsizes = [
            10,
            100,
            1000,
            10000,
            100000,
            1000000,
        ]
        for packsize in packsizes:
            revisions = []
            for i in pycompat.xrange(packsize):
                filename = b"filename-%d" % i
                content = b"content-%d" % i
                node = self.getHash(content)
                revisions.append((filename, node, nullid, content))

            path = self.createPack(revisions).path

            # Perf of large multi-get
            import gc

            gc.disable()
            pack = self.datapackreader(path)
            for lookupsize in lookupsizes:
                if lookupsize > packsize:
                    continue
                random.shuffle(revisions)
                findnodes = [(rev[0], rev[1]) for rev in revisions]

                start = time.time()
                pack.getmissing(findnodes[:lookupsize])
                elapsed = time.time() - start
                print(
                    "%s pack %d lookups = %0.04f"
                    % (
                        ('%d' % packsize).rjust(7),
                        ('%d' % lookupsize).rjust(7),
                        elapsed,
                    )
                )

            print("")
            gc.enable()

        # The perf test is meant to produce output, so we always fail the test
        # so the user sees the output.
        raise RuntimeError("perf test always fails")


class datapacktests(datapacktestsbase, unittest.TestCase):
    def __init__(self, *args, **kwargs):
        datapacktestsbase.__init__(self, datapack.datapack, True)
        unittest.TestCase.__init__(self, *args, **kwargs)


# TODO:
# datapack store:
# - getmissing
# - GC two packs into one

if __name__ == '__main__':
    if pycompat.iswindows:
        sys.exit(80)  # Skip on Windows
    silenttestrunner.main(__name__)