From ee66e089c1d4162765a3cec79fd5406e515b0db1 Mon Sep 17 00:00:00 2001
From: Paul Mackerras <paulus@samba.org>
Date: Fri, 9 May 2008 10:14:07 +1000
Subject: [PATCH] gitk: Make updates go faster

This goes back to the method of doing updates where we translate the
revisions we're given to SHA1 ids and then remove the ones we've asked
for before or that we've already come across.  This speeds up updates
enormously in most cases since it means git log doesn't have to traverse
large parts of the tree.  We used to do this, but it had bugs, and commit
468bcaedbb1589f16955e63b6bfba01c2f53e433 (gitk: Don't filter view
arguments through git rev-parse) went to the slower method to avoid the
bugs.

In order to do this properly, we have to parse the command line and
understand all the flag arguments.  So this adds a parser that checks
all the flag arguments.  If there are any we don't know about, we
disable the optimization and just pass the whole lot to git log
(except for -d/--date-order, which we remove from the list).

With this we can then use git rev-parse on the non-flag arguments to
work out exactly what SHA1 ids are included and excluded in the list,
which then enables us to ask for just the new ones when updating.
One wrinkle is that we have to turn symmetric diff arguments (of the
form a...b) back into symmetric diff form so that --left-right still
works, as git rev parse turns a...b into a b ^merge_base(a,b).

This also updates a couple of copyright notices.

Signed-off-by: Paul Mackerras <paulus@samba.org>
---
 gitk | 225 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-----
 1 file changed, 210 insertions(+), 15 deletions(-)

diff --git a/gitk b/gitk
index 4f83977070..5f27c6ac4f 100755
--- a/gitk
+++ b/gitk
@@ -2,7 +2,7 @@
 # Tcl ignores the next line -*- tcl -*- \
 exec wish "$0" -- "$@"
 
-# Copyright (C) 2005-2006 Paul Mackerras.  All rights reserved.
+# Copyright © 2005-2008 Paul Mackerras.  All rights reserved.
 # This program is free software; it may be used, copied, modified
 # and distributed under the terms of the GNU General Public Licence,
 # either version 2, or (at your option) any later version.
@@ -117,38 +117,190 @@ proc unmerged_files {files} {
 }
 
 proc parseviewargs {n arglist} {
-    global viewargs vdatemode vmergeonly
+    global vdatemode vmergeonly vflags vdflags vrevs vfiltered vorigargs
 
     set vdatemode($n) 0
     set vmergeonly($n) 0
-    set glargs {}
-    foreach arg $viewargs($n) {
+    set glflags {}
+    set diffargs {}
+    set nextisval 0
+    set revargs {}
+    set origargs $arglist
+    set allknown 1
+    set filtered 0
+    set i -1
+    foreach arg $arglist {
+	incr i
+	if {$nextisval} {
+	    lappend glflags $arg
+	    set nextisval 0
+	    continue
+	}
 	switch -glob -- $arg {
 	    "-d" -
 	    "--date-order" {
 		set vdatemode($n) 1
+		# remove from origargs in case we hit an unknown option
+		set origargs [lreplace $origargs $i $i]
+		incr i -1
+	    }
+	    # These request or affect diff output, which we don't want.
+	    # Some could be used to set our defaults for diff display.
+	    "-[puabwcrRBMC]" -
+	    "--no-renames" - "--full-index" - "--binary" - "--abbrev=*" -
+	    "--find-copies-harder" - "-l*" - "--ext-diff" - "--no-ext-diff" -
+	    "--src-prefix=*" - "--dst-prefix=*" - "--no-prefix" -
+	    "-O*" - "--text" - "--full-diff" - "--ignore-space-at-eol" -
+	    "--ignore-space-change" - "-U*" - "--unified=*" {
+		lappend diffargs $arg
+	    }
+	    # These cause our parsing of git log's output to fail, or else
+	    # they're options we want to set ourselves, so ignore them.
+	    "--raw" - "--patch-with-raw" - "--patch-with-stat" -
+	    "--name-only" - "--name-status" - "--color" - "--color-words" -
+	    "--log-size" - "--pretty=*" - "--decorate" - "--abbrev-commit" -
+	    "--cc" - "-z" - "--header" - "--parents" - "--boundary" -
+	    "--no-color" - "-g" - "--walk-reflogs" - "--no-walk" -
+	    "--timestamp" - "relative-date" - "--date=*" - "--stdin" -
+	    "--objects" - "--objects-edge" - "--reverse" {
+	    }
+	    # These are harmless, and some are even useful
+	    "--stat=*" - "--numstat" - "--shortstat" - "--summary" -
+	    "--check" - "--exit-code" - "--quiet" - "--topo-order" -
+	    "--full-history" - "--dense" - "--sparse" -
+	    "--follow" - "--left-right" - "--encoding=*" {
+		lappend glflags $arg
+	    }
+	    # These mean that we get a subset of the commits
+	    "--diff-filter=*" - "--no-merges" - "--unpacked" -
+	    "--max-count=*" - "--skip=*" - "--since=*" - "--after=*" -
+	    "--until=*" - "--before=*" - "--max-age=*" - "--min-age=*" -
+	    "--author=*" - "--committer=*" - "--grep=*" - "-[iE]" -
+	    "--remove-empty" - "--first-parent" - "--cherry-pick" -
+	    "-S*" - "--pickaxe-all" - "--pickaxe-regex" - {
+		set filtered 1
+		lappend glflags $arg
+	    }
+	    # This appears to be the only one that has a value as a
+	    # separate word following it
+	    "-n" {
+		set filtered 1
+		set nextisval 1
+		lappend glflags $arg
+	    }
+	    "--not" {
+		set notflag [expr {!$notflag}]
+		lappend revargs $arg
+	    }
+	    "--all" {
+		lappend revargs $arg
 	    }
 	    "--merge" {
 		set vmergeonly($n) 1
-		lappend glargs $arg
+		# git rev-parse doesn't understand --merge
+		lappend revargs --gitk-symmetric-diff-marker MERGE_HEAD...HEAD
 	    }
+	    # Other flag arguments including -<n>
+	    "-*" {
+		if {[string is digit -strict [string range $arg 1 end]]} {
+		    set filtered 1
+		} else {
+		    # a flag argument that we don't recognize;
+		    # that means we can't optimize
+		    set allknown 0
+		}
+		lappend glflags $arg
+	    }
+	    # Non-flag arguments specify commits or ranges of commits
 	    default {
-		lappend glargs $arg
+		if {[string match "*...*" $arg]} {
+		    lappend revargs --gitk-symmetric-diff-marker
+		}
+		lappend revargs $arg
+	    }
+	}
+    }
+    set vdflags($n) $diffargs
+    set vflags($n) $glflags
+    set vrevs($n) $revargs
+    set vfiltered($n) $filtered
+    set vorigargs($n) $origargs
+    return $allknown
+}
+
+proc parseviewrevs {view revs} {
+    global vposids vnegids
+
+    if {$revs eq {}} {
+	set revs HEAD
+    }
+    if {[catch {set ids [eval exec git rev-parse $revs]} err]} {
+	# we get stdout followed by stderr in $err
+	# for an unknown rev, git rev-parse echoes it and then errors out
+	set errlines [split $err "\n"]
+	set badrev {}
+	for {set l 0} {$l < [llength $errlines]} {incr l} {
+	    set line [lindex $errlines $l]
+	    if {!([string length $line] == 40 && [string is xdigit $line])} {
+		if {[string match "fatal:*" $line]} {
+		    if {[string match "fatal: ambiguous argument*" $line]
+			&& $badrev ne {}} {
+			if {[llength $badrev] == 1} {
+			    set err "unknown revision $badrev"
+			} else {
+			    set err "unknown revisions: [join $badrev ", "]"
+			}
+		    } else {
+			set err [join [lrange $errlines $l end] "\n"]
+		    }
+		    break
+		}
+		lappend badrev $line
+	    }
+	}		    
+	error_popup "Error parsing revisions: $err"
+	return {}
+    }
+    set ret {}
+    set pos {}
+    set neg {}
+    set sdm 0
+    foreach id [split $ids "\n"] {
+	if {$id eq "--gitk-symmetric-diff-marker"} {
+	    set sdm 4
+	} elseif {[string match "^*" $id]} {
+	    if {$sdm != 1} {
+		lappend ret $id
+		if {$sdm == 3} {
+		    set sdm 0
+		}
 	    }
+	    lappend neg [string range $id 1 end]
+	} else {
+	    if {$sdm != 2} {
+		lappend ret $id
+	    } else {
+		lset ret end [lindex $ret end]...$id
+	    }
+	    lappend pos $id
 	}
+	incr sdm -1
     }
-    return $glargs
+    set vposids($view) $pos
+    set vnegids($view) $neg
+    return $ret
 }
 
 # Start off a git log process and arrange to read its output
 proc start_rev_list {view} {
     global startmsecs commitidx viewcomplete
     global commfd leftover tclencoding
-    global viewargs viewargscmd vactualargs viewfiles vfilelimit
+    global viewargs viewargscmd viewfiles vfilelimit
     global showlocalchanges commitinterest mainheadid
     global progressdirn progresscoords proglastnc curview
     global viewactive loginstance viewinstances vmergeonly
     global pending_select mainheadid
+    global vcanopt vflags vrevs vorigargs
 
     set startmsecs [clock clicks -milliseconds]
     set commitidx($view) 0
@@ -167,8 +319,7 @@ proc start_rev_list {view} {
 	}
 	set args [concat $args [split $str "\n"]]
     }
-    set args [parseviewargs $view $args]
-    set vactualargs($view) $args
+    set vcanopt($view) [parseviewargs $view $args]
 
     set files $viewfiles($view)
     if {$vmergeonly($view)} {
@@ -187,6 +338,16 @@ proc start_rev_list {view} {
     }
     set vfilelimit($view) $files
 
+    if {$vcanopt($view)} {
+	set revs [parseviewrevs $view $vrevs($view)]
+	if {$revs eq {}} {
+	    return 0
+	}
+	set args [concat $vflags($view) $revs]
+    } else {
+	set args $vorigargs($view)
+    }
+
     if {[catch {
 	set fd [open [concat | git log --no-color -z --pretty=raw --parents \
 			 --boundary $args "--" $files] r]
@@ -248,11 +409,12 @@ proc getcommits {} {
 }
 
 proc updatecommits {} {
-    global curview vactualargs vfilelimit viewinstances
+    global curview vcanopt vorigargs vfilelimit viewinstances
     global viewactive viewcomplete loginstance tclencoding mainheadid
     global startmsecs commfd showneartags showlocalchanges leftover
     global mainheadid pending_select
     global isworktree
+    global varcid vposids vnegids vflags vrevs
 
     set isworktree [expr {[exec git rev-parse --is-inside-work-tree] == "true"}]
     set oldmainid $mainheadid
@@ -266,13 +428,46 @@ proc updatecommits {} {
 	}
     }
     set view $curview
+    if {$vcanopt($view)} {
+	set oldpos $vposids($view)
+	set oldneg $vnegids($view)
+	set revs [parseviewrevs $view $vrevs($view)]
+	if {$revs eq {}} {
+	    return
+	}
+	# note: getting the delta when negative refs change is hard,
+	# and could require multiple git log invocations, so in that
+	# case we ask git log for all the commits (not just the delta)
+	if {$oldneg eq $vnegids($view)} {
+	    set newrevs {}
+	    set npos 0
+	    # take out positive refs that we asked for before or
+	    # that we have already seen
+	    foreach rev $revs {
+		if {[string length $rev] == 40} {
+		    if {[lsearch -exact $oldpos $rev] < 0
+			&& ![info exists varcid($view,$rev)]} {
+			lappend newrevs $rev
+			incr npos
+		    }
+		} else {
+		    lappend $newrevs $rev
+		}
+	    }
+	    if {$npos == 0} return
+	    set revs $newrevs
+	    set vposids($view) [lsort -unique [concat $oldpos $vposids($view)]]
+	}
+	set args [concat $vflags($view) $revs --not $oldpos]
+    } else {
+	set args $vorigargs($view)
+    }
     if {[catch {
 	set fd [open [concat | git log --no-color -z --pretty=raw --parents \
-			  --boundary $vactualargs($view) --not [seeds $view] \
-			  "--" $vfilelimit($view)] r]
+			  --boundary $args "--" $vfilelimit($view)] r]
     } err]} {
 	error_popup "Error executing git log: $err"
-	exit 1
+	return
     }
     if {$viewactive($view) == 0} {
 	set startmsecs [clock clicks -milliseconds]
@@ -2217,7 +2412,7 @@ proc about {} {
     message $w.m -text [mc "
 Gitk - a commit viewer for git
 
-Copyright © 2005-2006 Paul Mackerras
+Copyright © 2005-2008 Paul Mackerras
 
 Use and redistribute under the terms of the GNU General Public License"] \
 	    -justify center -aspect 400 -border 2 -bg white -relief groove
-- 
2.11.4.GIT