Mercurial > hg
view tests/test-convert.t @ 30155:b7a966ce89ed
changelog: disable delta chains
This patch disables delta chains on changelogs. After this patch, new
entries on changelogs - including existing changelogs - will be stored
as the fulltext of that data (likely compressed). No delta computation
will be performed.
An overview of delta chains and data justifying this change follows.
Revlogs try to store entries as a delta against a previous entry (either
a parent revision in the case of generaldelta or the previous physical
revision when not using generaldelta). Most of the time this is the
correct thing to do: it frequently results in less CPU usage and smaller
storage.
Delta chains are most effective when the base revision being deltad
against is similar to the current data. This tends to occur naturally
for manifests and file data, since only small parts of each tend to
change with each revision. Changelogs, however, are a different story.
Changelog entries represent changesets/commits. And unless commits in a
repository are homogonous (same author, changing same files, similar
commit messages, etc), a delta from one entry to the next tends to be
relatively large compared to the size of the entry. This means that
delta chains tend to be short. How short? Here is the full vs delta
revision breakdown on some real world repos:
Repo % Full % Delta Max Length
hg 45.8 54.2 6
mozilla-central 42.4 57.6 8
mozilla-unified 42.5 57.5 17
pypy 46.1 53.9 6
python-zstandard 46.1 53.9 3
(I threw in python-zstandard as an example of a repo that is homogonous.
It contains a small Python project with changes all from the same
author.)
Contrast this with the manifest revlog for these repos, where 99+% of
revisions are deltas and delta chains run into the thousands.
So delta chains aren't as useful on changelogs. But even a short delta
chain may provide benefits. Let's measure that.
Delta chains may require less CPU to read revisions if the CPU time
spent reading smaller deltas is less than the CPU time used to
decompress larger individual entries. We can measure this via
`hg perfrevlog -c -d 1` to iterate a revlog to resolve each revision's
fulltext. Here are the results of that command on a repo using delta
chains in its changelog and on a repo without delta chains:
hg (forward)
! wall 0.407008 comb 0.410000 user 0.410000 sys 0.000000 (best of 25)
! wall 0.390061 comb 0.390000 user 0.390000 sys 0.000000 (best of 26)
hg (reverse)
! wall 0.515221 comb 0.520000 user 0.520000 sys 0.000000 (best of 19)
! wall 0.400018 comb 0.400000 user 0.390000 sys 0.010000 (best of 25)
mozilla-central (forward)
! wall 4.508296 comb 4.490000 user 4.490000 sys 0.000000 (best of 3)
! wall 4.370222 comb 4.370000 user 4.350000 sys 0.020000 (best of 3)
mozilla-central (reverse)
! wall 5.758995 comb 5.760000 user 5.720000 sys 0.040000 (best of 3)
! wall 4.346503 comb 4.340000 user 4.320000 sys 0.020000 (best of 3)
mozilla-unified (forward)
! wall 4.957088 comb 4.950000 user 4.940000 sys 0.010000 (best of 3)
! wall 4.660528 comb 4.650000 user 4.630000 sys 0.020000 (best of 3)
mozilla-unified (reverse)
! wall 6.119827 comb 6.110000 user 6.090000 sys 0.020000 (best of 3)
! wall 4.675136 comb 4.670000 user 4.670000 sys 0.000000 (best of 3)
pypy (forward)
! wall 1.231122 comb 1.240000 user 1.230000 sys 0.010000 (best of 8)
! wall 1.164896 comb 1.160000 user 1.160000 sys 0.000000 (best of 9)
pypy (reverse)
! wall 1.467049 comb 1.460000 user 1.460000 sys 0.000000 (best of 7)
! wall 1.160200 comb 1.170000 user 1.160000 sys 0.010000 (best of 9)
The data clearly shows that it takes less wall and CPU time to resolve
revisions when there are no delta chains in the changelogs, regardless
of the direction of traversal. Furthermore, not using a delta chain
means that fulltext resolution in reverse is as fast as iterating
forward. So not using delta chains on the changelog is a clear CPU win
for reading operations.
An example of a user-visible operation showing this speed-up is revset
evaluation. Here are results for
`hg perfrevset 'author(gps) or author(mpm)'`:
hg
! wall 1.655506 comb 1.660000 user 1.650000 sys 0.010000 (best of 6)
! wall 1.612723 comb 1.610000 user 1.600000 sys 0.010000 (best of 7)
mozilla-central
! wall 17.629826 comb 17.640000 user 17.600000 sys 0.040000 (best of 3)
! wall 17.311033 comb 17.300000 user 17.260000 sys 0.040000 (best of 3)
What about 00changelog.i size?
Repo Delta Chains No Delta Chains
hg 7,033,250 6,976,771
mozilla-central 82,978,748 81,574,623
mozilla-unified 88,112,349 86,702,162
pypy 20,740,699 20,659,741
The data shows that removing delta chains from the changelog makes the
changelog smaller.
Delta chains are also used during changegroup generation. This
operation essentially converts a series of revisions to one large
delta chain. And changegroup generation is smart: if the delta in
the revlog matches what the changegroup is emitting, it will reuse
the delta instead of recalculating it. We can measure the impact
removing changelog delta chains has on changegroup generation via
`hg perfchangegroupchangelog`:
hg
! wall 1.589245 comb 1.590000 user 1.590000 sys 0.000000 (best of 7)
! wall 1.788060 comb 1.790000 user 1.790000 sys 0.000000 (best of 6)
mozilla-central
! wall 17.382585 comb 17.380000 user 17.340000 sys 0.040000 (best of 3)
! wall 20.161357 comb 20.160000 user 20.120000 sys 0.040000 (best of 3)
mozilla-unified
! wall 18.722839 comb 18.720000 user 18.680000 sys 0.040000 (best of 3)
! wall 21.168075 comb 21.170000 user 21.130000 sys 0.040000 (best of 3)
pypy
! wall 4.828317 comb 4.830000 user 4.820000 sys 0.010000 (best of 3)
! wall 5.415455 comb 5.420000 user 5.410000 sys 0.010000 (best of 3)
The data shows eliminating delta chains makes the changelog part of
changegroup generation slower. This is expected since we now have to
compute deltas for revisions where we could recycle the delta before.
It is worth putting this regression into context of overall changegroup
times. Here is the rough total CPU time spent in changegroup generation
for various repos while using delta chains on the changelog:
Repo CPU Time (s) CPU Time w/ compression
hg 4.50 7.05
mozilla-central 111.1 222.0
pypy 28.68 75.5
Before compression, removing delta chains from the changegroup adds
~4.4% overhead to hg changegroup generation, 1.3% to mozilla-central,
and 2.0% to pypy. When you factor in zlib compression, these percentages
are roughly divided by 2.
While the increased CPU usage for changegroup generation is unfortunate,
I think it is acceptable because the percentage is small, server
operators (those likely impacted most by this) have other mechanisms
to mitigate CPU consumption (namely reducing zlib compression level and
pre-generated clone bundles), and because there is room to optimize this
in the future. For example, we could use the nullid as the base revision,
effectively encoding the full revision for each entry in the changegroup.
When doing this, `hg perfchangegroupchangelog` nearly halves:
mozilla-unified
! wall 21.168075 comb 21.170000 user 21.130000 sys 0.040000 (best of 3)
! wall 11.196461 comb 11.200000 user 11.190000 sys 0.010000 (best of 3)
This looks very promising as a future optimization opportunity.
It's worth that the changes in test-acl.t to the changegroup part size.
This is because revision 6 in the changegroup had a delta chain of
length 2 before and after this patch the base revision is nullrev.
When the base revision is nullrev, cg2packer.deltaparent() hardcodes
the *previous* revision from the changegroup as the delta parent.
This caused the delta in the changegroup to switch base revisions,
the delta to change, and the size to change accordingly. While the
size increased in this case, I think sizes will remain the same
on average, as the delta base for changelog revisions doesn't matter
too much (as this patch shows). So, I don't consider this a regression.
| author | Gregory Szorc <gregory.szorc@gmail.com> |
|---|---|
| date | Thu, 13 Oct 2016 12:50:27 +0200 |
| parents | d65e246100ed |
| children | ea3540e66fd8 |
line wrap: on
line source
$ cat >> $HGRCPATH <<EOF > [extensions] > convert= > [convert] > hg.saverev=False > EOF $ hg help convert hg convert [OPTION]... SOURCE [DEST [REVMAP]] convert a foreign SCM repository to a Mercurial one. Accepted source formats [identifiers]: - Mercurial [hg] - CVS [cvs] - Darcs [darcs] - git [git] - Subversion [svn] - Monotone [mtn] - GNU Arch [gnuarch] - Bazaar [bzr] - Perforce [p4] Accepted destination formats [identifiers]: - Mercurial [hg] - Subversion [svn] (history on branches is not preserved) If no revision is given, all revisions will be converted. Otherwise, convert will only import up to the named revision (given in a format understood by the source). If no destination directory name is specified, it defaults to the basename of the source with "-hg" appended. If the destination repository doesn't exist, it will be created. By default, all sources except Mercurial will use --branchsort. Mercurial uses --sourcesort to preserve original revision numbers order. Sort modes have the following effects: --branchsort convert from parent to child revision when possible, which means branches are usually converted one after the other. It generates more compact repositories. --datesort sort revisions by date. Converted repositories have good- looking changelogs but are often an order of magnitude larger than the same ones generated by --branchsort. --sourcesort try to preserve source revisions order, only supported by Mercurial sources. --closesort try to move closed revisions as close as possible to parent branches, only supported by Mercurial sources. If "REVMAP" isn't given, it will be put in a default location ("<dest>/.hg/shamap" by default). The "REVMAP" is a simple text file that maps each source commit ID to the destination ID for that revision, like so: <source ID> <destination ID> If the file doesn't exist, it's automatically created. It's updated on each commit copied, so 'hg convert' can be interrupted and can be run repeatedly to copy new commits. The authormap is a simple text file that maps each source commit author to a destination commit author. It is handy for source SCMs that use unix logins to identify authors (e.g.: CVS). One line per author mapping and the line format is: source author = destination author Empty lines and lines starting with a "#" are ignored. The filemap is a file that allows filtering and remapping of files and directories. Each line can contain one of the following directives: include path/to/file-or-dir exclude path/to/file-or-dir rename path/to/source path/to/destination Comment lines start with "#". A specified path matches if it equals the full relative name of a file or one of its parent directories. The "include" or "exclude" directive with the longest matching path applies, so line order does not matter. The "include" directive causes a file, or all files under a directory, to be included in the destination repository. The default if there are no "include" statements is to include everything. If there are any "include" statements, nothing else is included. The "exclude" directive causes files or directories to be omitted. The "rename" directive renames a file or directory if it is converted. To rename from a subdirectory into the root of the repository, use "." as the path to rename to. "--full" will make sure the converted changesets contain exactly the right files with the right content. It will make a full conversion of all files, not just the ones that have changed. Files that already are correct will not be changed. This can be used to apply filemap changes when converting incrementally. This is currently only supported for Mercurial and Subversion. The splicemap is a file that allows insertion of synthetic history, letting you specify the parents of a revision. This is useful if you want to e.g. give a Subversion merge two parents, or graft two disconnected series of history together. Each entry contains a key, followed by a space, followed by one or two comma-separated values: key parent1, parent2 The key is the revision ID in the source revision control system whose parents should be modified (same format as a key in .hg/shamap). The values are the revision IDs (in either the source or destination revision control system) that should be used as the new parents for that node. For example, if you have merged "release-1.0" into "trunk", then you should specify the revision on "trunk" as the first parent and the one on the "release-1.0" branch as the second. The branchmap is a file that allows you to rename a branch when it is being brought in from whatever external repository. When used in conjunction with a splicemap, it allows for a powerful combination to help fix even the most badly mismanaged repositories and turn them into nicely structured Mercurial repositories. The branchmap contains lines of the form: original_branch_name new_branch_name where "original_branch_name" is the name of the branch in the source repository, and "new_branch_name" is the name of the branch is the destination repository. No whitespace is allowed in the branch names. This can be used to (for instance) move code in one repository from "default" to a named branch. Mercurial Source ################ The Mercurial source recognizes the following configuration options, which you can set on the command line with "--config": convert.hg.ignoreerrors ignore integrity errors when reading. Use it to fix Mercurial repositories with missing revlogs, by converting from and to Mercurial. Default is False. convert.hg.saverev store original revision ID in changeset (forces target IDs to change). It takes a boolean argument and defaults to False. convert.hg.startrev specify the initial Mercurial revision. The default is 0. convert.hg.revs revset specifying the source revisions to convert. CVS Source ########## CVS source will use a sandbox (i.e. a checked-out copy) from CVS to indicate the starting point of what will be converted. Direct access to the repository files is not needed, unless of course the repository is ":local:". The conversion uses the top level directory in the sandbox to find the CVS repository, and then uses CVS rlog commands to find files to convert. This means that unless a filemap is given, all files under the starting directory will be converted, and that any directory reorganization in the CVS sandbox is ignored. The following options can be used with "--config": convert.cvsps.cache Set to False to disable remote log caching, for testing and debugging purposes. Default is True. convert.cvsps.fuzz Specify the maximum time (in seconds) that is allowed between commits with identical user and log message in a single changeset. When very large files were checked in as part of a changeset then the default may not be long enough. The default is 60. convert.cvsps.mergeto Specify a regular expression to which commit log messages are matched. If a match occurs, then the conversion process will insert a dummy revision merging the branch on which this log message occurs to the branch indicated in the regex. Default is "{{mergetobranch ([-\w]+)}}" convert.cvsps.mergefrom Specify a regular expression to which commit log messages are matched. If a match occurs, then the conversion process will add the most recent revision on the branch indicated in the regex as the second parent of the changeset. Default is "{{mergefrombranch ([-\w]+)}}" convert.localtimezone use local time (as determined by the TZ environment variable) for changeset date/times. The default is False (use UTC). hooks.cvslog Specify a Python function to be called at the end of gathering the CVS log. The function is passed a list with the log entries, and can modify the entries in-place, or add or delete them. hooks.cvschangesets Specify a Python function to be called after the changesets are calculated from the CVS log. The function is passed a list with the changeset entries, and can modify the changesets in-place, or add or delete them. An additional "debugcvsps" Mercurial command allows the builtin changeset merging code to be run without doing a conversion. Its parameters and output are similar to that of cvsps 2.1. Please see the command help for more details. Subversion Source ################# Subversion source detects classical trunk/branches/tags layouts. By default, the supplied "svn://repo/path/" source URL is converted as a single branch. If "svn://repo/path/trunk" exists it replaces the default branch. If "svn://repo/path/branches" exists, its subdirectories are listed as possible branches. If "svn://repo/path/tags" exists, it is looked for tags referencing converted branches. Default "trunk", "branches" and "tags" values can be overridden with following options. Set them to paths relative to the source URL, or leave them blank to disable auto detection. The following options can be set with "--config": convert.svn.branches specify the directory containing branches. The default is "branches". convert.svn.tags specify the directory containing tags. The default is "tags". convert.svn.trunk specify the name of the trunk branch. The default is "trunk". convert.localtimezone use local time (as determined by the TZ environment variable) for changeset date/times. The default is False (use UTC). Source history can be retrieved starting at a specific revision, instead of being integrally converted. Only single branch conversions are supported. convert.svn.startrev specify start Subversion revision number. The default is 0. Git Source ########## The Git importer converts commits from all reachable branches (refs in refs/heads) and remotes (refs in refs/remotes) to Mercurial. Branches are converted to bookmarks with the same name, with the leading 'refs/heads' stripped. Git submodules are converted to Git subrepos in Mercurial. The following options can be set with "--config": convert.git.similarity specify how similar files modified in a commit must be to be imported as renames or copies, as a percentage between "0" (disabled) and "100" (files must be identical). For example, "90" means that a delete/add pair will be imported as a rename if more than 90% of the file hasn't changed. The default is "50". convert.git.findcopiesharder while detecting copies, look at all files in the working copy instead of just changed ones. This is very expensive for large projects, and is only effective when "convert.git.similarity" is greater than 0. The default is False. convert.git.remoteprefix remote refs are converted as bookmarks with "convert.git.remoteprefix" as a prefix followed by a /. The default is 'remote'. convert.git.skipsubmodules does not convert root level .gitmodules files or files with 160000 mode indicating a submodule. Default is False. Perforce Source ############### The Perforce (P4) importer can be given a p4 depot path or a client specification as source. It will convert all files in the source to a flat Mercurial repository, ignoring labels, branches and integrations. Note that when a depot path is given you then usually should specify a target directory, because otherwise the target may be named "...-hg". The following options can be set with "--config": convert.p4.encoding specify the encoding to use when decoding standard output of the Perforce command line tool. The default is default system encoding. convert.p4.startrev specify initial Perforce revision (a Perforce changelist number). Mercurial Destination ##################### The Mercurial destination will recognize Mercurial subrepositories in the destination directory, and update the .hgsubstate file automatically if the destination subrepositories contain the <dest>/<sub>/.hg/shamap file. Converting a repository with subrepositories requires converting a single repository at a time, from the bottom up. The following options are supported: convert.hg.clonebranches dispatch source branches in separate clones. The default is False. convert.hg.tagsbranch branch name for tag revisions, defaults to "default". convert.hg.usebranchnames preserve branch names. The default is True. convert.hg.sourcename records the given string as a 'convert_source' extra value on each commit made in the target repository. The default is None. All Destinations ################ All destination types accept the following options: convert.skiptags does not convert tags from the source repo to the target repo. The default is False. options ([+] can be repeated): -s --source-type TYPE source repository type -d --dest-type TYPE destination repository type -r --rev REV [+] import up to source revision REV -A --authormap FILE remap usernames using this file --filemap FILE remap file names using contents of file --full apply filemap changes by converting all files again --splicemap FILE splice synthesized history into place --branchmap FILE change branch names while converting --branchsort try to sort changesets by branches --datesort try to sort changesets by date --sourcesort preserve source changesets order --closesort try to reorder closed revisions (some details hidden, use --verbose to show complete help) $ hg init a $ cd a $ echo a > a $ hg ci -d'0 0' -Ama adding a $ hg cp a b $ hg ci -d'1 0' -mb $ hg rm a $ hg ci -d'2 0' -mc $ hg mv b a $ hg ci -d'3 0' -md $ echo a >> a $ hg ci -d'4 0' -me $ cd .. $ hg convert a 2>&1 | grep -v 'subversion python bindings could not be loaded' assuming destination a-hg initializing destination a-hg repository scanning source... sorting... converting... 4 a 3 b 2 c 1 d 0 e $ hg --cwd a-hg pull ../a pulling from ../a searching for changes no changes found conversion to existing file should fail $ touch bogusfile $ hg convert a bogusfile initializing destination bogusfile repository abort: cannot create new bundle repository [255] #if unix-permissions no-root conversion to dir without permissions should fail $ mkdir bogusdir $ chmod 000 bogusdir $ hg convert a bogusdir abort: Permission denied: 'bogusdir' [255] user permissions should succeed $ chmod 700 bogusdir $ hg convert a bogusdir initializing destination bogusdir repository scanning source... sorting... converting... 4 a 3 b 2 c 1 d 0 e #endif test pre and post conversion actions $ echo 'include b' > filemap $ hg convert --debug --filemap filemap a partialb | \ > grep 'run hg' run hg source pre-conversion action run hg sink pre-conversion action run hg sink post-conversion action run hg source post-conversion action converting empty dir should fail "nicely $ mkdir emptydir override $PATH to ensure p4 not visible; use $PYTHON in case we're running from a devel copy, not a temp installation $ PATH="$BINDIR" $PYTHON "$BINDIR"/hg convert emptydir assuming destination emptydir-hg initializing destination emptydir-hg repository emptydir does not look like a CVS checkout $TESTTMP/emptydir does not look like a Git repository (glob) emptydir does not look like a Subversion repository emptydir is not a local Mercurial repository emptydir does not look like a darcs repository emptydir does not look like a monotone repository emptydir does not look like a GNU Arch repository emptydir does not look like a Bazaar repository cannot find required "p4" tool abort: emptydir: missing or unsupported repository [255] convert with imaginary source type $ hg convert --source-type foo a a-foo initializing destination a-foo repository abort: foo: invalid source repository type [255] convert with imaginary sink type $ hg convert --dest-type foo a a-foo abort: foo: invalid destination repository type [255] testing: convert must not produce duplicate entries in fncache $ hg convert a b initializing destination b repository scanning source... sorting... converting... 4 a 3 b 2 c 1 d 0 e contents of fncache file: $ cat b/.hg/store/fncache | sort data/a.i data/b.i test bogus URL $ hg convert -q bzr+ssh://foobar@selenic.com/baz baz abort: bzr+ssh://foobar@selenic.com/baz: missing or unsupported repository [255] test revset converted() lookup $ hg --config convert.hg.saverev=True convert a c initializing destination c repository scanning source... sorting... converting... 4 a 3 b 2 c 1 d 0 e $ echo f > c/f $ hg -R c ci -d'0 0' -Amf adding f created new head $ hg -R c log -r "converted(09d945a62ce6)" changeset: 1:98c3dd46a874 user: test date: Thu Jan 01 00:00:01 1970 +0000 summary: b $ hg -R c log -r "converted()" changeset: 0:31ed57b2037c user: test date: Thu Jan 01 00:00:00 1970 +0000 summary: a changeset: 1:98c3dd46a874 user: test date: Thu Jan 01 00:00:01 1970 +0000 summary: b changeset: 2:3b9ca06ef716 user: test date: Thu Jan 01 00:00:02 1970 +0000 summary: c changeset: 3:4e0debd37cf2 user: test date: Thu Jan 01 00:00:03 1970 +0000 summary: d changeset: 4:9de3bc9349c5 user: test date: Thu Jan 01 00:00:04 1970 +0000 summary: e test specifying a sourcename $ echo g > a/g $ hg -R a ci -d'0 0' -Amg adding g $ hg --config convert.hg.sourcename=mysource --config convert.hg.saverev=True convert a c scanning source... sorting... converting... 0 g $ hg -R c log -r tip --template '{extras % "{extra}\n"}' branch=default convert_revision=a3bc6100aa8ec03e00aaf271f1f50046fb432072 convert_source=mysource