Convert pdf, doc and docx files to text by default

Converting PDF and  Word files to text before diffing them allows an
easier comparison between changed files. This reintroduces some
functionality of Git for Windows 1.x.

The pdftotext tool exists both in the xpdf package and the poppler
package; we opted to include the xpdf one because requires us to add
only its dependency libstdc++-6.dll in addition. Poppler's version would
require 23 additional dlls.

In Johannes' tests, this change increased the portable Git by a scant
701kB -- which he deems worth the benefit.

This fixes git-for-windows/git#355

[jes: re-wrapped commit message, avoided double list entry for
astextplain (in git-extra) and unzip.exe (dependency of docx2txt) that
7-Zip would complain about]

Signed-off-by: Matthias Aßhauer <[email protected]>
Signed-off-by: Johannes Schindelin <[email protected]>

Author: rimrul

git-extra/PKGBUILD

@@ -18,6 +18,8 @@ pkgver() {
 build() {
   test $startdir/$pkgname.install -nt $startdir/$pkgname.install.in ||
   sed -e "/^@@GITCONFIG@@$/r $startdir/gitconfig" -e "/^@@GITCONFIG@@$/d" \
+	-e "/^@@GITATTRIBUTES@@$/r $startdir/gitattributes" \
+	-e "/^@@GITATTRIBUTES@@$/d" \
 	<$startdir/$pkgname.install.in >$startdir/$pkgname.install
 
   gcc -o create-shortcut.exe $startdir/create-shortcut.c -luuid -lole32
@@ -50,4 +52,5 @@ package() {
   install -m755 $startdir/bash_profile.sh $pkgdir/etc/profile.d
   install -m644 $startdir/msys2-32.ico $pkgdir/usr/share/git
   install -m644 $startdir/99-post-install-cleanup.post $pkgdir/etc/post-install
+  install -m755 $startdir/astextplain $pkgdir/$mingwdir/bin
 }

git-extra/astextplain

@@ -0,0 +1,30 @@
+#!/bin/sh -e
+# minimalistic replacement for `run-mailcap --action=cat <file>`
+
+if test "$#" != 1 ; then
+	echo "Usage: astextplain <file>" 1>&2
+	exit 1
+fi
+
+# XXX output encoding (UTF-8) hardcoded
+case "$1" in
+	*.doc | *.DOC | *.dot | *.DOT)
+		antiword -m UTF-8 "$1" | sed "s/\^M$//" || cat "$1"
+		;;
+	*.docx | *.DOCX)
+		docx2txt.pl "$1" -
+		;;
+	*.pdf | *.PDF)
+		pdftotext -layout "$1" -enc UTF-8 - | sed "s/(\^M$)|(^\^L)//"
+		;;
+	# TODO add rtf support
+	*.rtf | *.RTF)
+		cat "$1"
+		;;
+	*)
+		echo "E: unsupported filetype $1" 1>&2
+		exit 1
+		;;
+esac
+
+exit 0

git-extra/git-extra.install.in

@@ -8,6 +8,12 @@ post_install () {
 		cat > /$dir/etc/gitconfig <<\GITCONFIG
 @@GITCONFIG@@
 GITCONFIG
+		test ! -d /$dir ||
+		test -f /$dir/etc/gitattributes ||
+		cat > /$dir/etc/gitattributes <<\GITATTRIBUTES
+@@GITATTRIBUTES@@
+
+GITATTRIBUTES
 	done
 
 	grep -q '^db_home: env windows' /etc/nsswitch.conf ||

git-extra/gitattributes

@@ -0,0 +1,10 @@
+*.doc	diff=astextplain
+*.DOC	diff=astextplain
+*.docx	diff=astextplain
+*.DOCX	diff=astextplain
+*.dot	diff=astextplain
+*.DOT	diff=astextplain
+*.pdf	diff=astextplain
+*.PDF	diff=astextplain
+*.rtf	diff=astextplain
+*.RTF	diff=astextplain

make-file-list.sh

@@ -28,14 +28,15 @@ pacman_list () {
 }
 
 # Packages that have been added after Git SDK 1.0.0 was released...
-pacman -S --needed --noconfirm mingw-w64-$ARCH-connect git-flow >&2 ||
+pacman -S --needed --noconfirm mingw-w64-$ARCH-connect git-flow unzip docx2txt \
+	mingw-w64-$ARCH-antiword mingw-w64-$ARCH-xpdf >&2 ||
 die "Could not install required packages"
 
 pacman_list mingw-w64-$ARCH-git mingw-w64-$ARCH-git-doc-html \
 	git-extra ncurses mintty vim openssh winpty \
 	sed awk less grep gnupg tar findutils coreutils diffutils patch \
 	dos2unix which subversion mingw-w64-$ARCH-tk \
-	mingw-w64-$ARCH-connect git-flow "$@" |
+	mingw-w64-$ARCH-connect git-flow docx2txt mingw-w64-$ARCH-antiword "$@" |
 grep -v -e '\.[acho]$' -e '\.l[ao]$' -e '/aclocal/' \
 	-e '/man/' -e '/pkgconfig/' -e '/emacs/' \
 	-e '^/usr/lib/python' -e '^/usr/lib/ruby' \
@@ -71,7 +72,8 @@ grep --perl-regexp -v -e '^/usr/(lib|share)/terminfo/(?!.*/(cygwin|dumb|xterm.*)
 sed 's/^\///'
 
 test -z "$PACKAGE_VERSIONS_FILE" ||
-pacman -Q filesystem dash rebase util-linux >>"$PACKAGE_VERSIONS_FILE"
+pacman -Q filesystem dash rebase util-linux unzip \
+	mingw-w64-$ARCH-xpdf >>"$PACKAGE_VERSIONS_FILE"
 
 cat <<EOF
 etc/profile
@@ -89,4 +91,7 @@ usr/bin/dash.exe
 usr/bin/rebase.exe
 usr/bin/rebaseall
 usr/bin/getopt.exe
+mingw$BITNESS/etc/gitattributes
+mingw$BITNESS/bin/pdftotext.exe
+mingw$BITNESS/bin/libstdc++-6.dll
 EOF