Some groups of added/deleted lines in diffs can be slid up or down,
because lines at the edges of the group are not unique. Picking good
shifts for such groups is not a matter of correctness but definitely has
a big effect on aesthetics. For example, consider the following two
diffs. The first is what standard Git emits:
--- a/9c572b21dd090a1e5c5bb397053bf8043ffe7fb4:git-send-email.perl
+++ b/6dcfa306f2b67b733a7eb2d7ded1bc9987809edb:git-send-email.perl
@@ -231,6 +231,9 @@ if (!defined $initial_reply_to && $prompting) {
}
if (!$smtp_server) {
+ $smtp_server = $repo->config('sendemail.smtpserver');
+}
+if (!$smtp_server) {
foreach (qw( /usr/sbin/sendmail /usr/lib/sendmail )) {
if (-x $_) {
$smtp_server = $_;
The following diff is equivalent, but is obviously preferable from an
aesthetic point of view:
--- a/9c572b21dd090a1e5c5bb397053bf8043ffe7fb4:git-send-email.perl
+++ b/6dcfa306f2b67b733a7eb2d7ded1bc9987809edb:git-send-email.perl
@@ -230,6 +230,9 @@ if (!defined $initial_reply_to && $prompting) {
$initial_reply_to =~ s/(^\s+|\s+$)//g;
}
+if (!$smtp_server) {
+ $smtp_server = $repo->config('sendemail.smtpserver');
+}
if (!$smtp_server) {
foreach (qw( /usr/sbin/sendmail /usr/lib/sendmail )) {
if (-x $_) {
This patch teaches Git to pick better positions for such "diff sliders"
using heuristics that take the positions of nearby blank lines and the
indentation of nearby lines into account.
The existing Git code basically always shifts such "sliders" as far down
in the file as possible. The only exception is when the slider can be
aligned with a group of changed lines in the other file, in which case
Git favors depicting the change as one add+delete block rather than one
add and a slightly offset delete block. This naive algorithm often
yields ugly diffs.
Commit d634d61ed6 improved the situation somewhat by preferring to
position add/delete groups to make their last line a blank line, when
that is possible. This heuristic does more good than harm, but (1) it
can only help if there are blank lines in the right places, and (2)
always picks the last blank line, even if there are others that might be
better. The end result is that it makes perhaps 1/3 as many errors as
the default Git algorithm, but that still leaves a lot of ugly diffs.
This commit implements a new and much better heuristic for picking
optimal "slider" positions using the following approach: First observe
that each hypothetical positioning of a diff slider introduces two
splits: one between the context lines preceding the group and the first
added/deleted line, and the other between the last added/deleted line
and the first line of context following it. It tries to find the
positioning that creates the least bad splits.
Splits are evaluated based only on the presence and locations of nearby
blank lines, and the indentation of lines near the split. Basically, it
prefers to introduce splits adjacent to blank lines, between lines that
are indented less, and between lines with the same level of indentation.
In more detail:
1. It measures the following characteristics of a proposed splitting
position in a `struct split_measurement`:
* the number of blank lines above the proposed split
* whether the line directly after the split is blank
* the number of blank lines following that line
* the indentation of the nearest non-blank line above the split
* the indentation of the line directly below the split
* the indentation of the nearest non-blank line after that line
2. It combines the measured attributes using a bunch of
empirically-optimized weighting factors to derive a `struct
split_score` that measures the "badness" of splitting the text at
that position.
3. It combines the `split_score` for the top and the bottom of the
slider at each of its possible positions, and selects the position
that has the best `split_score`.
I determined the initial set of weighting factors by collecting a corpus
of Git histories from 29 open-source software projects in various
programming languages. I generated many diffs from this corpus, and
determined the best positioning "by eye" for about 6600 diff sliders. I
used about half of the repositories in the corpus (corresponding to
about 2/3 of the sliders) as a training set, and optimized the weights
against this corpus using a crude automated search of the parameter
space to get the best agreement with the manually-determined values.
Then I tested the resulting heuristic against the full corpus. The
results are summarized in the following table, in column `indent-1`:
| repository | count | Git 2.9.0 | compaction | compaction-fixed | indent-1 | indent-2 |
| --------------------- | ----- | -------------- | -------------- | ---------------- | -------------- | -------------- |
| afnetworking | 109 | 89 (81.7%) | 37 (33.9%) | 37 (33.9%) | 2 (1.8%) | 2 (1.8%) |
| alamofire | 30 | 18 (60.0%) | 14 (46.7%) | 15 (50.0%) | 0 (0.0%) | 0 (0.0%) |
| angular | 184 | 127 (69.0%) | 39 (21.2%) | 23 (12.5%) | 5 (2.7%) | 5 (2.7%) |
| animate | 313 | 2 (0.6%) | 2 (0.6%) | 2 (0.6%) | 2 (0.6%) | 2 (0.6%) |
| ant | 380 | 356 (93.7%) | 152 (40.0%) | 148 (38.9%) | 15 (3.9%) | 15 (3.9%) | *
| bugzilla | 306 | 263 (85.9%) | 109 (35.6%) | 99 (32.4%) | 14 (4.6%) | 15 (4.9%) | *
| corefx | 126 | 91 (72.2%) | 22 (17.5%) | 21 (16.7%) | 6 (4.8%) | 6 (4.8%) |
| couchdb | 78 | 44 (56.4%) | 26 (33.3%) | 28 (35.9%) | 6 (7.7%) | 6 (7.7%) | *
| cpython | 937 | 158 (16.9%) | 50 (5.3%) | 49 (5.2%) | 5 (0.5%) | 5 (0.5%) | *
| discourse | 160 | 95 (59.4%) | 42 (26.2%) | 36 (22.5%) | 18 (11.2%) | 13 (8.1%) |
| docker | 307 | 194 (63.2%) | 198 (64.5%) | 253 (82.4%) | 8 (2.6%) | 8 (2.6%) | *
| electron | 163 | 132 (81.0%) | 38 (23.3%) | 39 (23.9%) | 6 (3.7%) | 6 (3.7%) |
| git | 536 | 470 (87.7%) | 73 (13.6%) | 78 (14.6%) | 16 (3.0%) | 16 (3.0%) | *
| gitflow | 127 | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) |
| ionic | 133 | 89 (66.9%) | 29 (21.8%) | 38 (28.6%) | 1 (0.8%) | 1 (0.8%) |
| ipython | 482 | 362 (75.1%) | 167 (34.6%) | 169 (35.1%) | 11 (2.3%) | 11 (2.3%) | *
| junit | 161 | 147 (91.3%) | 67 (41.6%) | 66 (41.0%) | 1 (0.6%) | 1 (0.6%) | *
| lighttable | 15 | 5 (33.3%) | 0 (0.0%) | 2 (13.3%) | 0 (0.0%) | 0 (0.0%) |
| magit | 88 | 75 (85.2%) | 11 (12.5%) | 9 (10.2%) | 1 (1.1%) | 0 (0.0%) |
| neural-style | 28 | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) |
| nodejs | 781 | 649 (83.1%) | 118 (15.1%) | 111 (14.2%) | 4 (0.5%) | 5 (0.6%) | *
| phpmyadmin | 491 | 481 (98.0%) | 75 (15.3%) | 48 (9.8%) | 2 (0.4%) | 2 (0.4%) | *
| react-native | 168 | 130 (77.4%) | 79 (47.0%) | 81 (48.2%) | 0 (0.0%) | 0 (0.0%) |
| rust | 171 | 128 (74.9%) | 30 (17.5%) | 27 (15.8%) | 16 (9.4%) | 14 (8.2%) |
| spark | 186 | 149 (80.1%) | 52 (28.0%) | 52 (28.0%) | 2 (1.1%) | 2 (1.1%) |
| tensorflow | 115 | 66 (57.4%) | 48 (41.7%) | 48 (41.7%) | 5 (4.3%) | 5 (4.3%) |
| test-more | 19 | 15 (78.9%) | 2 (10.5%) | 2 (10.5%) | 1 (5.3%) | 1 (5.3%) | *
| test-unit | 51 | 34 (66.7%) | 14 (27.5%) | 8 (15.7%) | 2 (3.9%) | 2 (3.9%) | *
| xmonad | 23 | 22 (95.7%) | 2 (8.7%) | 2 (8.7%) | 1 (4.3%) | 1 (4.3%) | *
| --------------------- | ----- | -------------- | -------------- | ---------------- | -------------- | -------------- |
| totals | 6668 | 4391 (65.9%) | 1496 (22.4%) | 1491 (22.4%) | 150 (2.2%) | 144 (2.2%) |
| totals (training set) | 4552 | 3195 (70.2%) | 1053 (23.1%) | 1061 (23.3%) | 86 (1.9%) | 88 (1.9%) |
| totals (test set) | 2116 | 1196 (56.5%) | 443 (20.9%) | 430 (20.3%) | 64 (3.0%) | 56 (2.6%) |
In this table, the numbers are the count and percentage of human-rated
sliders that the corresponding algorithm got *wrong*. The columns are
* "repository" - the name of the repository used. I used the diffs
between successive non-merge commits on the HEAD branch of the
corresponding repository.
* "count" - the number of sliders that were human-rated. I chose most,
but not all, sliders to rate from those among which the various
algorithms gave different answers.
* "Git 2.9.0" - the default algorithm used by `git diff` in Git 2.9.0.
* "compaction" - the heuristic used by `git diff --compaction-heuristic`
in Git 2.9.0.
* "compaction-fixed" - the heuristic used by `git diff
--compaction-heuristic` after the fixes from earlier in this patch
series. Note that the results are not dramatically different than
those for "compaction". Both produce non-ideal diffs only about 1/3 as
often as the default `git diff`.
* "indent-1" - the new `--indent-heuristic` algorithm, using the first
set of weighting factors, determined as described above.
* "indent-2" - the new `--indent-heuristic` algorithm, using the final
set of weighting factors, determined as described below.
* `*` - indicates that repo was part of training set used to determine
the first set of weighting factors.
The fact that the heuristic performed nearly as well on the test set as
on the training set in column "indent-1" is a good indication that the
heuristic was not over-trained. Given that fact, I ran a second round of
optimization, using the entire corpus as the training set. The resulting
set of weights gave the results in column "indent-2". These are the
weights included in this patch.
The final result gives consistently and significantly better results
across the whole corpus than either `git diff` or `git diff
--compaction-heuristic`. It makes only about 1/30 as many errors as the
former and about 1/10 as many errors as the latter. (And a good fraction
of the remaining errors are for diffs that involve weirdly-formatted
code, sometimes apparently machine-generated.)
The tools that were used to do this optimization and analysis, along
with the human-generated data values, are recorded in a separate project
[1].
This patch adds a new command-line option `--indent-heuristic`, and a
new configuration setting `diff.indentHeuristic`, that activate this
heuristic. This interface is only meant for testing purposes, and should
be finalized before including this change in any release.
[1] https://github.com/mhagger/diff-slider-tools
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The idea of xdl_change_compact() is fairly simple:
* Proceed through groups of changed lines in the file to be compacted,
keeping track of the corresponding location in the "other" file.
* If possible, slide the group up and down to try to give the most
aesthetically pleasing diff. Whenever it is slid, the current location
in the other file needs to be adjusted.
But these simple concepts are obfuscated by a lot of index handling that
is written in terse, subtle, and varied patterns. I found it very hard
to convince myself that the function was correct.
So introduce a "struct group" that represents a group of changed lines
in a file. Add some functions that perform elementary operations on
groups:
* Initialize a group to the first group in a file
* Move to the next or previous group in a file
* Slide a group up or down
Even though the resulting code is longer, I think it is easier to
understand and review. Its performance is not changed
appreciably (though it would be if `group_next()` and `group_previous()`
were not inlined).
...and in fact, the rewriting helped me discover another bug in the
--compaction-heuristic code: The update of blank_lines was never done
for the highest possible position of the group. This means that it could
fail to slide the group to its highest possible position, even if that
position had a blank line as its last line. So for example, it yielded
the following diff:
$ git diff --no-index --compaction-heuristic a.txt b.txt
diff --git a/a.txt b/b.txt
index e53969f..0d60c5fe 100644
--- a/a.txt
+++ b/b.txt
@@ -1,3 +1,7 @@
1
A
+
+B
+
+A
2
when in fact the following diff is better (according to the rules of
--compaction-heuristic):
$ git diff --no-index --compaction-heuristic a.txt b.txt
diff --git a/a.txt b/b.txt
index e53969f..0d60c5fe 100644
--- a/a.txt
+++ b/b.txt
@@ -1,3 +1,7 @@
1
+A
+
+B
+
A
2
The new code gives the bottom answer.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
There is no reason for it to take an array and two indexes as argument,
as it only accesses two elements of the array.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
There is no reason for it to take an array and index as argument, as it
only accesses a single element of the array.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If the changed group of lines can be matched to a group in the other
file, then that positioning should take precedence over the compaction
heuristic.
The old code tried the heuristic unconditionally, which cost redundant
effort and also was broken if the matching code had already shifted the
group higher than the blank line.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The code branch used for the compaction heuristic forgot to keep ixo in
sync while the group was shifted. This is certainly wrong, as it causes
the two counters to get out of sync.
I *think* that this bug could also have caused the function to read past
the end of the rchgo array, though I haven't done the work to prove it
for sure. Here is my reasoning:
If ixo is not decremented correctly during one iteration of the outer
while loop, then it will loose sync with the ix counter. In particular,
ixo will be too large.
Suppose that the next iterations of the outer while loop (i.e.,
processing the next block of add/delete lines) don't have any sliders.
Then the ixo counter would be incremented by the number of non-changed
lines in xdf, which is the same as the number of non-changed lines in
xdfo that *should have* followed the group that experienced the
malfunction. But since ixo was too large at the end of that iteration,
it will be incremented past the end of the xdfo->rchg array, and will
try to read that memory illegally.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git add -i/-p" learned to honor diff.compactionHeuristic
experimental knob, so that the user can work on the same hunk split
as "git diff" output.
* jk/add-i-diff-compact-heuristics:
add--interactive: respect diff.compactionHeuristic
The ownership rule for the piece of memory that hold references to
be fetched in "git fetch" was screwy, which has been cleaned up.
* km/fetch-do-not-free-remote-name:
builtin/fetch.c: don't free remote->name after fetch
HTTPd tests learned to show the server error log to help diagnosing
a failing tests.
* nd/test-lib-httpd-show-error-log-in-verbose:
lib-httpd.sh: print error.log on error
Instead of taking advantage of a struct string_list that is
allocated with all NULs happens to be STRING_LIST_INIT_NODUP kind,
initialize them explicitly as such, to document their behaviour
better.
* jk/string-list-static-init:
use string_list initializer consistently
blame,shortlog: don't make local option variables static
interpret-trailers: don't duplicate option strings
parse_opt_string_list: stop allocating new strings
HTTP transport gained an option to produce more detailed debugging
trace.
* ep/http-curl-trace:
imap-send.c: introduce the GIT_TRACE_CURL enviroment variable
http.c: implement the GIT_TRACE_CURL environment variable
The completion script (in contrib/) learned to complete "git
status" options.
* tb/complete-status:
completion: add git status
completion: add __git_get_option_value helper
completion: factor out untracked file modes into a variable
"git cherry-pick A" worked on an unborn branch, but "git
cherry-pick A..B" didn't.
* mg/cherry-pick-multi-on-unborn:
cherry-pick: allow to pick to unborn branches
Allow messages that are generated by auto gc during "git push" on
the receiving end to be explicitly passed back to the sending end
over sideband, so that they are shown with "remote: " prefix to
avoid confusing the users.
* lf/receive-pack-auto-gc-to-client:
receive-pack: send auto-gc output over sideband 2
Comments about misbehaving FreeBSD shells have been clarified with
the version number (9.x and before are broken, newer ones are OK).
* em/newer-freebsd-shells-are-fine-with-returns:
rebase: update comment about FreeBSD /bin/sh
"git status" used to say "working directory" when it meant "working
tree".
* lv/status-say-working-tree-not-directory:
Use "working tree" instead of "working directory" for git status
"git update-index --add --chmod=+x file" may be usable as an escape
hatch, but not a friendly thing to force for people who do need to
use it regularly. "git add --chmod=+x file" can be used instead.
* et/add-chmod-x:
add: add --chmod=+x / --chmod=-x options
The git-prompt scriptlet (in contrib/) was not friendly with those
who uses "set -u", which has been fixed.
* vs/prompt-avoid-unset-variable:
git-prompt.sh: Don't error on null ${ZSH,BASH}_VERSION, $short_sha
"git reflog" stopped upon seeing an entry that denotes a branch
creation event (aka "unborn"), which made it appear as if the
reflog was truncated.
* sg/reflog-past-root:
reflog: continue walking the reflog past root commits
The documentation tries to consistently spell "GPG"; when
referring to the specific program name, "gpg" is used.
* dn/gpg-doc:
Documentation: GPG capitalization
The documentation set has been updated so that literal commands,
configuration variables and environment variables are consistently
typeset in fixed-width font and bold in manpages.
* tr/doc-tt:
doc: change configuration variables format
doc: more consistency in environment variables format
doc: change environment variables format
doc: clearer rule about formatting literals
The "git apply" standalone program is being libified; this is the
first step to move many state variables into a structure that can
be explicitly (re)initialized to make the machinery callable more
than once.
The next step that moves some remaining state variables into the
structure and turns die()s into an error return that propagates up
to the caller is not queued yet but in flight. It would be good to
review the above first and give the remainder of the series a solid
base to build on.
* cc/apply-introduce-state: (50 commits)
builtin/apply: remove misleading comment on lock_file field
builtin/apply: move 'newfd' global into 'struct apply_state'
builtin/apply: add 'lock_file' pointer into 'struct apply_state'
builtin/apply: move applying patches into apply_all_patches()
builtin/apply: move 'state' check into check_apply_state()
builtin/apply: move 'symlink_changes' global into 'struct apply_state'
builtin/apply: move 'fn_table' global into 'struct apply_state'
builtin/apply: move 'state_linenr' global into 'struct apply_state'
builtin/apply: move 'max_change' and 'max_len' into 'struct apply_state'
builtin/apply: move 'ws_ignore_action' into 'struct apply_state'
builtin/apply: move 'ws_error_action' into 'struct apply_state'
builtin/apply: move 'applied_after_fixing_ws' into 'struct apply_state'
builtin/apply: move 'squelch_whitespace_errors' into 'struct apply_state'
builtin/apply: remove whitespace_option arg from set_default_whitespace_mode()
builtin/apply: move 'whitespace_option' into 'struct apply_state'
builtin/apply: move 'whitespace_error' global into 'struct apply_state'
builtin/apply: move 'root' global into 'struct apply_state'
builtin/apply: move 'p_value_known' global into 'struct apply_state'
builtin/apply: move 'p_value' global into 'struct apply_state'
builtin/apply: move 'has_include' global into 'struct apply_state'
...
"git show -W" (extend hunks to cover the entire function, delimited
by lines that match the "funcname" pattern) used to show the entire
file when a change added an entire function at the end of the file,
which has been fixed.
* rs/xdiff-hunk-with-func-line:
xdiff: fix merging of appended hunk with -W
grep: -W: don't extend context to trailing empty lines
t7810: add test for grep -W and trailing empty context lines
xdiff: don't trim common tail with -W
xdiff: -W: don't include common trailing empty lines in context
xdiff: ignore empty lines before added functions with -W
xdiff: handle appended chunks better with -W
xdiff: factor out match_func_rec()
t4051: rewrite, add more tests
"git rev-list --count" whose walk-length is limited with "-n"
option did not work well with the counting optimized to look at the
bitmap index.
* jk/rev-list-count-with-bitmap:
rev-list: disable bitmaps when "-n" is used with listing objects
rev-list: "adjust" results of "--count --use-bitmap-index -n"
The commands in `git log` family take %C(auto) in a custom format
string. This unconditionally turned the color on, ignoring
--no-color or with --color=auto when the output is not connected to
a tty; this was corrected to make the format truly behave as
"auto".
* et/pretty-format-c-auto:
format_commit_message: honor `color=auto` for `%C(auto)`
When "git daemon" is run without --[init-]timeout specified, a
connection from a client that silently goes offline can hang around
for a long time, wasting resources. The socket-level KEEPALIVE has
been enabled to allow the OS to notice such failed connections.
* ew/daemon-socket-keepalive:
daemon: enable SO_KEEPALIVE for all sockets
For repositories without parent following enabled, finding
git parents through svn:mergeinfo or svk::parents can be
expensive and pointless.
Reported-by: Александр Овчинников <proff@proff.email>
http://mid.gmane.org/4094761466408188@web24o.yandex.ru
Signed-off-by: Eric Wong <e@80x24.org>
"git show -W" (extend hunks to cover the entire function, delimited
by lines that match the "funcname" pattern) used to show the entire
file when a change added an entire function at the end of the file,
which has been fixed.
* rs/xdiff-hunk-with-func-line:
xdiff: fix merging of appended hunk with -W
grep: -W: don't extend context to trailing empty lines
t7810: add test for grep -W and trailing empty context lines
xdiff: don't trim common tail with -W
xdiff: -W: don't include common trailing empty lines in context
xdiff: ignore empty lines before added functions with -W
xdiff: handle appended chunks better with -W
xdiff: factor out match_func_rec()
t4051: rewrite, add more tests
"git rev-list --count" whose walk-length is limited with "-n"
option did not work well with the counting optimized to look at the
bitmap index.
* jk/rev-list-count-with-bitmap:
rev-list: disable bitmaps when "-n" is used with listing objects
rev-list: "adjust" results of "--count --use-bitmap-index -n"
We usually call a function that clears the contents a data
structure X without freeing the structure itself clear_X(), and
call a function that does clear_X() and also frees it free_X().
free_pathspec() function has been renamed to clear_pathspec()
to avoid confusion.
* jc/clear-pathspec:
pathspec: rename free_pathspec() to clear_pathspec()
"git worktree add" learned that '-' can be used as a short-hand for
"@{-1}", the previous branch.
* jg/dash-is-last-branch-in-worktree-add:
worktree: allow "-" short-hand for @{-1} in add command
The commands in `git log` family take %C(auto) in a custom format
string. This unconditionally turned the color on, ignoring
--no-color or with --color=auto when the output is not connected to
a tty; this was corrected to make the format truly behave as
"auto".
* et/pretty-format-c-auto:
format_commit_message: honor `color=auto` for `%C(auto)`
An upstream project can make a recommendation to shallowly clone
some submodules in the .gitmodules file it ships.
* sb/submodule-recommend-shallowness:
submodule update: learn `--[no-]recommend-shallow` option
submodule-config: keep shallow recommendation around
Michael Haggerty