[announce+patch] git-forest 20080402

[announce+patch] git-forest 20080402

[Jan Engelhardt]

Hi,


here is the current snapshot of git-forest, a text-based tree 
visualizer, as a patch to the git.git/contrib/ directory.

What's new since the last posting is that it now displays [HEAD]
along with a commit, since HEAD does not always need to be the
first commit that is graphed.
    The --all option now includes graphing a detached HEAD, as can
happen on switching to a non-branch and during rebases.
    Speaking of rebases, git-forest will now also show where you are in 
a rebase operation -- quite handy if you forget whether you actually are 
in one (happened to me time and again and when you finally notice, you 
need to play with HEAD@{x} to restore it) -- it can be suppressed 
with --no-rebase.

I updated the screenshot at 
http://jengelh.hopto.org/images/git-forest.png (URL unchanged).
Color-agnostic people may use the --no-color option. (Tag colors 
follow gitk as far as the 16 colors permit.)

The direct grab URL for the script is still
http://dev.computergmbh.de/gitweb.cgi?p=hxtools;a=blob_plain;f=bin/git-forest
but of course a patch is provided below with the intent of getting it 
into contrib/ at least.


thanks,
Jan

N.B.: This was a ridiculous command to get it right...
can't this be done easier?
`git-format-patch -C -M --stat=72 --summary -p --thread -n HEAD^..HEAD &&
git-send-email --suppress-from --no-chain --compose --to git@xxx 00*`

[PATCH 1/1] Import git-forest

[Jan Engelhardt]

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
---
 contrib/git-forest/git-forest     |  422 +++++++++++++++++++++++++++++
 contrib/git-forest/git-forest.txt |   53 ++++
 2 files changed, 475 insertions(+), 0 deletions(-)
 create mode 100755 contrib/git-forest/git-forest
 create mode 100644 contrib/git-forest/git-forest.txt

diff --git a/contrib/git-forest/git-forest b/contrib/git-forest/git-forest
new file mode 100755
index 0000000..c38b218
--- /dev/null
+++ b/contrib/git-forest/git-forest
@@ -0,0 +1,422 @@
+#!/usr/bin/perl
+#
+#	git-森林
+#	text-based tree visualisation
+#	Copyright © Jan Engelhardt <jengelh [at] gmx de>, 2008
+#
+#	This program is free software; you can redistribute it and/or modify
+#	it under the terms of the GNU General Public License as published by
+#	the Free Software Foundation; either version 2 or 3 of the license.
+#
+use Getopt::Long;
+use Git;
+use strict;
+use encoding "utf8";
+my $Repo     = Git->repository($ENV{"GIT_DIR"} || ".");
+my $Show_all = 0;
+my $Show_rebase = 1;
+my $Style    = 2;
+my $With_sha = 0;
+my %Color    = (
+	"default" => "\e[0m",
+	"at"      => "\e[1;30m",
+	"hhead"   => "\e[1;31m",
+	"head"    => "\e[1;32m",
+	"ref"     => "\e[1;34m",
+	"remote"  => "\e[1;35m",
+	"sha"     => "\e[0;31m",
+	"tag"     => "\e[1;33m",
+	"tree"    => "\e[0;33m",
+);
+
+&main();
+
+sub main ()
+{
+	&Getopt::Long::Configure(qw(bundling pass_through));
+	&GetOptions(
+		"all"      => \$Show_all,
+		"no-color" => sub { %Color = (); },
+		"no-rebase" => sub { $Show_rebase = 0; },
+		"style=i"  => \$Style,
+		"sha"      => \$With_sha,
+	);
+	if ($Show_all) {
+		unshift(@ARGV, "--all", "HEAD");
+	}
+	&process();
+}
+
+sub process ()
+{
+	my(@vine);
+	my $refs = &get_refs();
+	my($fh, $fhc) = $Repo->command_output_pipe("log", "--date-order",
+	                "--pretty=format:<%H><%h><%P>%s", @ARGV);
+
+	while (defined(my $line = <$fh>)) {
+		chomp $line;
+		my($sha, $mini_sha, $parents, $msg) =
+			($line =~ /^<(.*?)><(.*?)><(.*?)>(.*)/s);
+		my @parents = split(" ", $parents);
+
+		&vine_branch(\@vine, $sha);
+		my $ra = &vine_commit(\@vine, $sha, \@parents);
+
+		if (exists($refs->{$sha})) {
+			print &vis_post(&vis_commit($ra,
+			      $Color{at}."─".$Color{default}));
+			&ref_print($refs->{$sha});
+		} else {
+			print &vis_post(&vis_commit($ra, " "));
+		}
+		if ($With_sha) {
+			print $msg, $Color{at}, "──(", $Color{sha}, $mini_sha,
+			      $Color{at}, ")", $Color{default}, "\n";
+		} else {
+			print $msg, "\n";
+		}
+
+		&vine_merge(\@vine, $sha, \@parents);
+	}
+	$Repo->command_close_pipe($fh, $fhc);
+}
+
+sub get_refs ()
+{
+	my($fh, $c) = $Repo->command_output_pipe("show-ref");
+	my $ret = {};
+
+	while (defined(my $ln = <$fh>)) {
+		chomp $ln;
+		if (length($ln) == 0) {
+			next;
+		}
+
+		my($sha, $name) = ($ln =~ /^(\S+)\s+(.*)/s);
+		if (!exists($ret->{$sha})) {
+			$ret->{$sha} = [];
+		}
+		push(@{$ret->{$sha}}, $name);
+		if ($name =~ m{^refs/tags/}) {
+			my $sub_sha = $Repo->command("log", "-1",
+			              "--pretty=format:%H", $name);
+			chomp $sub_sha;
+			if ($sha ne $sub_sha) {
+				push(@{$ret->{$sub_sha}}, $name);
+			}
+		}
+	}
+
+	$Repo->command_close_pipe($fh, $c);
+
+	my $rebase = -e $Repo->repo_path()."/.dotest-merge/git-rebase-todo" &&
+	             $Show_rebase;
+	if ($rebase) {
+		my $up  = $Repo->command("rev-parse", ".dotest-merge/upstream");
+		my $old = $Repo->command("rev-parse", ".dotest-merge/head");
+		chomp $up;
+		chomp $old;
+		unshift(@{$ret->{$up}}, "rebase/upstream");
+		unshift(@{$ret->{$old}}, "rebase/old-HEAD");
+	}
+
+	my $head = $Repo->command("rev-parse", "HEAD");
+	chomp $head;
+	if ($rebase) {
+		unshift(@{$ret->{$head}}, "rebase/inprogress");
+	}
+	unshift(@{$ret->{$head}}, "HEAD");
+
+	return $ret;
+}
+
+sub ref_print ($)
+{
+	foreach my $symbol (@{shift @_}) {
+		print $Color{at}, "[";
+		if ($symbol eq "HEAD" || $symbol =~ m{^rebase/}) {
+			print $Color{hhead}, $symbol;
+		} elsif ($symbol =~ m{^refs/(remotes/[^/]+)/(.*)}s) {
+			print $Color{remote}, $1, $Color{head}, "/$2";
+		} elsif ($symbol =~ m{^refs/heads/(.*)}s) {
+			print $Color{head}, $1;
+		} elsif ($symbol =~ m{^refs/tags/(.*)}s) {
+			print $Color{tag}, $1;
+		} elsif ($symbol =~ m{^refs/(.*)}s) {
+			print $Color{ref}, $1;
+		}
+		print $Color{at}, "]──", $Color{default};
+	}
+}
+
+sub vine_branch ($$)
+{
+	my($vine, $rev) = @_;
+	my $idx;
+
+	my $left = "â• ";
+	my $matched = 0;
+	my $ret;
+
+	for ($idx = 0; $idx < scalar(@$vine); ++$idx) {
+		if (!defined($vine->[$idx])) {
+			$ret .= "═";
+			next;
+		} elsif ($vine->[$idx] ne $rev) {
+			$ret .= "╪";
+			next;
+		}
+		if ($matched == 0) {
+			$ret .= "â• ";
+		} else {
+			$ret .= "â•©";
+			$vine->[$idx] = undef;
+		}
+		++$matched;
+	}
+
+	if ($matched < 2) {
+		return;
+	}
+
+	while (!defined($vine->[$#$vine])) {
+		pop(@$vine);
+	}
+
+	print &vis_post(&vis_branch($ret)), "\n";
+}
+
+sub vine_commit ($$$)
+{
+	my($vine, $rev, $parents) = @_;
+	my $ret;
+
+	for (my $i = 0; $i <= $#$vine; ++$i) {
+		if (!defined($vine->[$i])) {
+			$ret .= " ";
+		} elsif ($vine->[$i] eq $rev) {
+			$ret .= "â•Ÿ";
+		} else {
+			$ret .= "â•‘";
+		}
+	}
+
+	if ($ret !~ /â•Ÿ/) {
+		# Not having produced a â•Ÿ before means this is a HEAD
+		$ret .= "â•“";
+		push(@$vine, $rev);
+	}
+
+	while (scalar(@$vine) > 0 && !defined($vine->[$#$vine])) {
+		pop(@$vine);
+	}
+
+	if (scalar(@$parents) == 0) {
+		# tree root
+		$ret =~ s/â•Ÿ/â•™/g;
+	}
+
+	return $ret;
+}
+
+#
+#	Generate vine graphics for a merge
+#
+sub vine_merge ($$$)
+{
+	my($vine, $rev, $parents) = @_;
+	my $orig_vine = -1;
+	my @slot;
+	my($ret, $max);
+
+	for (my $i = 0; $i <= $#$vine; ++$i) {
+		if ($vine->[$i] eq $rev) {
+			$orig_vine = $i;
+			last;
+		}
+	}
+
+	if ($orig_vine == -1) {
+		die "vine_commit() did not add this vine.";
+	}
+
+	if (scalar(@$parents) <= 1) {
+		#
+		# A single parent does not need a visual. Update and return.
+		#
+		$vine->[$orig_vine] = $parents->[0];
+
+		while (scalar(@$vine) > 0 && !defined($vine->[$#$vine])) {
+			pop(@$vine);
+		}
+		return;
+	}
+
+	#
+	# Find some good spots to split out into.
+	#
+	push(@slot, $orig_vine);
+	my $parent = 0;
+
+	for (my $seeker = 2; $parent < $#$parents &&
+	    $seeker < 2 + 2 * $#$vine; ++$seeker)
+	{
+		my $idx = ($seeker % 2 == 0) ? -1 : 1;
+		$idx   *= int($seeker / 2);
+		$idx   += $orig_vine;
+
+		if ($idx >= 0 && $idx <= $#$vine && !defined($vine->[$idx])) {
+			push(@slot, $idx);
+			++$parent;
+		}
+	}
+	for (my $idx = $orig_vine + 1; $parent < $#$parents; ++$idx) {
+		if (!defined($vine->[$idx])) {
+			push(@slot, $idx);
+			++$parent;
+		}
+	}
+
+	if (scalar(@slot) != scalar(@$parents)) {
+		die "Serious internal problem";
+	}
+
+	@slot = sort { $a <=> $b } @slot;
+	$max  = scalar(@$vine) + scalar(@slot);
+
+	for (my $i = 0; $i < $max; ++$i) {
+		if ($#slot >= 0 && $i == $slot[0]) {
+			shift @slot;
+			$vine->[$i] = shift @$parents;
+			$ret .= ($i == $orig_vine) ? "S" : "s";
+		} elsif (defined($vine->[$i])) {
+			$ret .= "â•‘";
+		} else {
+			$ret .= " ";
+		}
+
+	}
+
+	$ret =~ s/ +$//gs;
+	print &vis_post(&vis_merge($ret)), "\n";
+}
+
+sub vis_branch ($)
+{
+	# Sample input: ╬═╠╬╬╬╩╬═╬╬╬╬╬╬╩╬╩═╬╬
+	my $ra = shift @_;
+	my $i;
+
+	$ra =~ s{^(.+?)â• }{
+		$_ = $1;
+		$_ =~ tr/╪═/║ /;
+		$_ =~ s/(.)/$1 /gs;
+		$_ .= 'â• ';
+	}es;
+	$ra =~ s{(â• .*)â•©}{
+		$_ = $1;
+		$_ =~ s/(.)/$1═/gs;
+		$_ .= '╝';
+	}es;
+	$ra =~ s{╝(.*)$}{
+		$_ = $1;
+		$_ =~ tr/╪═/║ /;
+		$_ =~ s/(.)/$1 /gs;
+		$_ = "╝ $_";
+	}es;
+	return $ra;
+}
+
+sub vis_commit ($$)
+{
+	my($ra, $sep) = @_;
+	my($l, $r) = ($ra =~ /^(.*?)([â•Ÿâ•“â•™].*)/);
+	$l =~ s/(.)/$1 /gs;
+	$r =~ s/(.)/$1 /gs;
+	$r =~ s/ /$sep/gs;
+	return $l.$r;
+}
+
+sub vis_merge ($)
+{
+	my $s = shift @_;
+
+	if ($s =~ s/(s.*)S(.*s)/&vis_merge3($1, $2)/es) {
+		;
+	} elsif ($s =~ /(?:s.*)S/s) {
+		while ($s =~ s/(s.*)║(.*S)/$1╪$2/s) {
+			;
+		}
+		$s =~ s/(s.*)S/&vis_merge2L($1)."â•£"/es;
+	} elsif ($s =~ /S(?:.*s)/s) {
+		while ($s =~ s/(S.*)║(.*s)/$1╪$2/s) {
+			;
+		}
+		$s =~ /S(.*s)/;
+		$s =~ s/S(.*s)/"â• ".&vis_merge2R($1)/es;
+	} else {
+		# $s =~ s/S/â•‘/s;
+		die "Should not come here";
+	}
+	$s =~ s{(.)}{&vis_merge1($1)}egs;
+	return $s;
+}
+
+sub vis_merge1 ($)
+{
+	if ($_[0] eq "╔" || $_[0] eq "╦" || $_[0] eq "╠" || $_[0] eq "╪") {
+		return $_[0]."═";
+	} else {
+		return $_[0]." ";
+	}
+}
+
+sub vis_merge2L ($)
+{
+	my $l = shift @_;
+	$l =~ s/^s/â•”/;
+	$l =~ s/s/╦/g;
+	return $l;
+}
+
+sub vis_merge2R ($)
+{
+	my $r = shift @_;
+	$r =~ s/s$/â•—/;
+	$r =~ s/s/╦/g;
+	return $r;
+}
+
+sub vis_merge3 ($$)
+{
+	my($l, $r) = shift @_;
+	$l =~ s/^s/â•”/;
+	$l =~ s/s/╦/g;
+	$r =~ s/s$/â•—/;
+	$r =~ s/s/╦/g;
+	return "$l╪$r";
+}
+
+#
+#	post-process vine graphic
+#
+sub vis_post ($)
+{
+	my $s = shift @_;
+
+	if ($Style == 1) {
+		$s =~ tr/╔╦╗╠╬╣╚╩╝║═╟╓╙╪/┌┬┐├┼┤└┴┘│─├┌└┼/;
+	} elsif ($Style == 2) {
+		$s =~ tr/╪/╬/;
+	} elsif ($Style == 3) {
+		$s =~ tr/╔╦╗╠╬╣╚╩╝║╟╓╙/╒╤╕╞╪╡╘╧╛│├┌└/;
+	} elsif ($Style == 4) {
+		$s =~ tr/╔╦╗╠╬╣╚╩╝╪═/╓╥╖╟╫╢╙╨╜╫─/;
+	}
+
+	if ($Color{default} ne "") {
+		$s =~ s{\Q$Color{default}\E}{$&$Color{tree}}g;
+	}
+	return $Color{tree}, $s, $Color{default};
+}
diff --git a/contrib/git-forest/git-forest.txt b/contrib/git-forest/git-forest.txt
new file mode 100644
index 0000000..5dfd478
--- /dev/null
+++ b/contrib/git-forest/git-forest.txt
@@ -0,0 +1,53 @@
+
+Usage: git-forest [OPTIONS...] [REVLISTOPTIONS] [REFSPEC...]
+
+--no-rebase
+	Do not show rebase/ pseudo-refs
+--style=1
+	Use single-line visuals
+--style=2
+	Use double-line visuals (default)
+--style=3
+	Use single-line visuals in vertical direction,
+	and double-line ones in horizontal direction.
+--style=4
+	Use double-line visuals in vertical direction,
+	and single-line ones in horizontal direction.
+--sha
+	Display SHAs for each commit
+
+All other options and arguments are passed down to git-log.
+Commonly useful options are --all and --topo-order.
+
+Example:
+	git-forest --all | less -RS
+	git-forest origin/master | less -RS
+
+This tool does not try to minimize the empty space between branches
+like gitk does. Take it as a feature.
+
+Notes on interpretation:
+
+Connections with four "legs" ('╬' or variants thereof, like '╪') are
+meant to be interpreted as being connected only in the horizontal and
+vertical direction, NOT around-the-corner.
+
+Connections with three or less "legs" ('╠', '╩', '╦', etc.) imply
+a connection to all connected directions.
+
+Consider the following example history. The merge at 070 has merged
+together the branches with commits f21, 886 and 8f8, but NOT d3e.
+
+	╓─[HEAD]──[master]──-──(0c6)
+	╠═╗ 
+	║ ╟ -──(710)
+	╟ ║ -──(070)
+	╠═╬═╦═╗
+	║ ╟ ║ ║ -──(d3e)
+	║ ║ ╟ ║ -──(886)
+	║ ║ ║ ╟ -──(8f8)
+	║ ╠═╝ ║
+	║ ╟   ║ -──(199)
+	╟ ║   ║ -──(f21)
+	╠═╩═══╝ 
+	╙ -──(15f)
-- 
1.5.4.4

Re: [PATCH 1/1] Import git-forest

[Junio C Hamano]

Jan Engelhardt <jengelh@computergmbh.de> writes:


    Content-Type: TEXT/PLAIN; charset=ISO-8859-1

> +#	git-??????

Hmmmm...

> +&main();
> +
> +sub main ()
> +{
> +	&Getopt::Long::Configure(qw(bundling pass_through));
> +
> +	&GetOptions(
> +		"all"      => \$Show_all,
> +		"no-color" => sub { %Color = (); },
> +		"no-rebase" => sub { $Show_rebase = 0; },
> +		"style=i"  => \$Style,
> +		"sha"      => \$With_sha,
> +	);
> +	if ($Show_all) {
> +		unshift(@ARGV, "--all", "HEAD");
> +	}
> +	&process();
> +}

This is a Perl script with a curious style.  Your subroutines force the
"funny" parameter semantics to their callers by using prototypes, but
avoid the potential surprises and inconveniences prototypes may cause by
using the &name(...) form when they themselves make subroutine calls,
i.e. not eating their own dog food.

Is it fashionable to make text based tree render these days?  How does
this compare with the git-graph in the nearby thread?  Can they borrow
from each other?

Re: [PATCH] Import git-forest (20080403)

[Jan Engelhardt]

On Wednesday 2008-04-02 23:49, Junio C Hamano wrote:
> Jan Engelhardt <jengelh@computergmbh.de> writes:
>
>    Content-Type: TEXT/PLAIN; charset=ISO-8859-1
>
>> +#	git-??????
>
> Hmmmm...

A git-send-email bug it seems. It does not add any Content-Type header
thereby effectively trashing UTF-8. How wonderful. Explains a many 
corrumped patches on LKML (and subsequently, the git history).

> This is a Perl script with a curious style.  Your subroutines force the
> "funny" parameter semantics to their callers by using prototypes, but
> avoid the potential surprises and inconveniences prototypes may cause by
> using the &name(...) form when they themselves make subroutine calls,
> i.e. not eating their own dog food.

Right, I removed that. I would have probably used PHP (perl is touted as 
a write-once language), but I could not be sure everyone has _that_ 
installed. Perl, most people have it.

> Is it fashionable to make text based tree render these days?

 	Same arguments as last time :>
Definitely. Do not assume everyone has an X display ready at any given 
time. Also, gitk does a Follows/Precedes scan which would not really be 
necessary if you just wanted to have the commit forest.

> How does this compare with the git-graph in the nearby thread?

git-forest does have...
- unicode/line-drawing graphics, colors
- display of signs like gitk, that is, ref names like  ──[v2.6.25-rc7],
   ──[remotes/linus/master], ──[original/filtered-branch] etc.
- display detached HEAD and its parents when --all is given
- track in-progress rebase

differences:
- graph does space reduction between vines
- git-forest keeps a vine in its column right now - this way
   you can point the mouse on it, scroll down until its end and
   keep track of it. (Diagonal art would require adjusting the
   mouse position, which is error prone in complex history
   like linux.)

> Can they borrow from each other?

Mh, I have uptaken some suggestions from the git-graph code: support
for --pretty and --reverse. Did not take very long (and if you look
at the git log for hxtools you can see it's criminally small
patches :-)

-- 
make boldconfig -- to boldly select what no one has selected before

Re: [PATCH] Import git-forest (20080403)

[Junio C Hamano]

Jan Engelhardt <jengelh@computergmbh.de> writes:

> On Wednesday 2008-04-02 23:49, Junio C Hamano wrote:
>> Jan Engelhardt <jengelh@computergmbh.de> writes:
>>
>>    Content-Type: TEXT/PLAIN; charset=ISO-8859-1
>>
>>> +#	git-??????
>>
>> Hmmmm...
>
> A git-send-email bug it seems.

Do we know if it has been fixed?  I see you used 1.5.4.4.

By the way, this time you are speaking Russian ;-)

    Content-Type: TEXT/PLAIN; charset=KOI8-R; format=flowed

>> Can they borrow from each other?
>
> Mh, I have uptaken some suggestions from the git-graph code: support
> for --pretty and --reverse.

Good.  Let's see who survives ;-)

Re: [PATCH] Import git-forest (20080403)

[Jan Engelhardt]

On Thursday 2008-04-03 10:39, Junio C Hamano wrote:
>>>    Content-Type: TEXT/PLAIN; charset=ISO-8859-1
>>>> +#	git-??????
>>> Hmmmm...
>> A git-send-email bug it seems.
> Do we know if it has been fixed? I see you used 1.5.4.4

I hope you know, you are the git maintainer. :)
Yes, 1.5.4.4 from opensuse.
>
> By the way, this time you are speaking Russian ;-)
>
>    Content-Type: TEXT/PLAIN; charset=KOI8-R; format=flowed

curses to alpine... but that still does not change the
fact that send-email/1.5.4.4 completely forget the header.


-- 
make boldconfig -- to boldly select what no one has selected before

Re: [PATCH] Import git-forest (20080403)

[Jeff King]

On Thu, Apr 03, 2008 at 01:39:08AM -0700, Junio C Hamano wrote:

> >>    Content-Type: TEXT/PLAIN; charset=ISO-8859-1
> >>
> >>> +#	git-??????
> >>
> >> Hmmmm...
> >
> > A git-send-email bug it seems.
> 
> Do we know if it has been fixed?  I see you used 1.5.4.4.

I think the problem is that the characters are in the _patch_. We do the
right thing for the commit message, SoB, and all headers, but we have
always just assumed the patch was OK, since we have no encoding
information on the files (and worse, if we have a _different_ encoding
for the commit message versus the diff content, we will have to
re-encode something or make it an attachment).

-Peff

Re: [announce+patch] git-forest 20080402

[Jeff King]

On Wed, Apr 02, 2008 at 10:58:14PM +0200, Jan Engelhardt wrote:

> N.B.: This was a ridiculous command to get it right...
> can't this be done easier?
> `git-format-patch -C -M --stat=72 --summary -p --thread -n HEAD^..HEAD &&
> git-send-email --suppress-from --no-chain --compose --to git@xxx 00*`

Most of the time you don't need -C -M (unless you are actually renaming
stuff). If you always want it, try:

  git config --global diff.renames copies

format-patch already defaults to "--stat --summary -p" if you give it no
format options. The default stat is 80, I believe; for format-patch it
should probably be bumped to 72 anyway. format-patch will also default
to ..HEAD if just given HEAD^.

So you should be able to do:

  git format-patch --thread -n HEAD^

For your send-email invocation, doing:

  git config --global sendemail.suppressfrom true
  git config --global sendemail.chainreplyto false
  # the --global makes the others user-wide; this
  # one should just be specific to your git repo
  git config sendemail.to git@vger.kernel.org

should allow:

  git sendemail --compose 00*

though if you are just sending a single patch, it is common to put the
cover letter material (if it is short) after the '---' but before the
diffstat.

HTH,
-Peff