A
Austin Ziegler
It is with pleasure that I announce the latest versions of Diff::LCS,
a pure-Ruby differencing algorithm.
Version 1.0.4 is an incremental bugfix that improves the installer (it
is not entirely fixed for ri document generation; I believe that I
have done most of what I can do with this), fixes a problem of missing
left-side changes when comparing against an empty sequence, and fixes
a problem with ldiff output, especially for unified diff output.
Version 1.1.0 is a feature release that includes all bugfixes up to
and including 1.0.4. In addition, it adds a new standard callback
(ContextDiffCallback) that works similarly to SDiffCallback but
produces a smaller diff set.
More importantly, it is now possible to patch a sequence with a diff
set to produce the second sequence. If the patch is performed without
specifying direction, then it will be examined to see which direction
in which the patch will be performed. If the patch is performed by
specifying the direction, then the patch will be performed without
verification that the patch set applies to the source sequence.
Using version 1.0.4, I have produced a couple of patches today for
both rdoc and for Ruby's dir.c that look *exactly* like they do from
unix diff (at least the version included with cygwin).
Both versions can be found in the Ruwiki project filelist:
http://rubyforge.org/frs/?group_id=84
-austin
Diff::LCS README
================
Diff::LCS is a port of Algorithm:iff[1] that uses the McIlroy-Hunt
longest common subsequence (LCS) algorithm to compute intelligent
differences between two sequenced enumerable containers[2]. The
implementation is based on Mario I. Wolczko's[3] Smalltalk version
(1.2, 1993)[4] and Ned Konz's[5] Perl version (Algorithm:iff)[6].
This release is version 1.1.0 with the addition of new features,
including the ability to #patch and #unpatch changes as well as a new
contextual diff callback, Diff::LCS::ContextDiffCallbacks, that should
improve the context sensitivity of patching.
Using this module is quite simple. By default, Diff::LCS does not
extend objects with the Diff::LCS interface, but will be called as if
it were a function:
require 'diff/lcs'
seq1 = %w(a b c e h j l m n p)
seq2 = %w(b c d e f j k l m r s t)
lcs = Diff::LCS.LCS(seq1, seq2)
diffs = Diff::LCS.diff(seq1, seq2)
sdiff = Diff::LCS.sdiff(seq1, seq2)
seq = Diff::LCS.traverse_sequences(seq1, seq2, callback_obj)
bal = Diff::LCS.traverse_balanced(seq1, seq2, callback_obj)
seq2 == Diff::LCS.patch!(seq1, diffs)
seq1 == Diff::LCS.unpatch!(seq2, diffs)
seq2 == Diff::LCS.patch!(seq1, sdiff)
seq1 == Diff::LCS.unpatch!(seq2, sdiff)
Objects can be extended with Diff::LCS:
seq1.extend(Diff::LCS)
lcs = seq1.lcs(seq2)
diffs = seq1.diff(seq2)
sdiff = seq1.sdiff(seq2)
seq = seq1.traverse_sequences(seq2, callback_obj)
bal = seq1.traverse_balanced(seq2, callback_obj)
seq2 == seq1.patch!(diffs)
seq1 == seq2.unpatch!(diffs)
seq2 == seq1.patch!(sdiff)
seq1 == seq2.unpatch!(sdiff)
By requiring 'diff/lcs/array' or 'diff/lcs/string', Array or String
will be extended for use this way.
Copyright
=========
# Copyright 2004 Austin Ziegler <[email protected]>
# adapted from:
# Algorithm:iff (Perl) by Ned Konz <[email protected]>
# Smalltalk by Mario I. Wolczko <[email protected]>
# implements McIlroy-Hunt diff algorithm
#
# This program is free software. It may be redistributed and/or modified
# under the terms of the GPL version 2 (or later), the Perl Artistic
# licence, or the Ruby licence.
#
# $Id: README,v 1.5 2004/08/08 20:33:09 austin Exp $
Footnotes
=========
[1] This library is called Diff::LCS because there are multiple
Ruby libraries called Algorithm:iff maintained by other authors.
[2] By sequenced enumerable, I mean that the order of enumeration is
predictable and consistent for the same set of data. While it is
theoretically possible to generate a diff for unordereded hash, it
will only be meaningful if the enumeration of the hashes is
consistent. In general, this will mean that containers that behave
like String or Array will perform best.
[3] (e-mail address removed)
[4] ftp://st.cs.uiuc.edu/pub/Smalltalk/MANCHESTER/manchester/4.0/diff.st
[5] (e-mail address removed)
[6] http://search.cpan.org/~nedkonz/Algorithm-Diff-1.15/
a pure-Ruby differencing algorithm.
Version 1.0.4 is an incremental bugfix that improves the installer (it
is not entirely fixed for ri document generation; I believe that I
have done most of what I can do with this), fixes a problem of missing
left-side changes when comparing against an empty sequence, and fixes
a problem with ldiff output, especially for unified diff output.
Version 1.1.0 is a feature release that includes all bugfixes up to
and including 1.0.4. In addition, it adds a new standard callback
(ContextDiffCallback) that works similarly to SDiffCallback but
produces a smaller diff set.
More importantly, it is now possible to patch a sequence with a diff
set to produce the second sequence. If the patch is performed without
specifying direction, then it will be examined to see which direction
in which the patch will be performed. If the patch is performed by
specifying the direction, then the patch will be performed without
verification that the patch set applies to the source sequence.
Using version 1.0.4, I have produced a couple of patches today for
both rdoc and for Ruby's dir.c that look *exactly* like they do from
unix diff (at least the version included with cygwin).
Both versions can be found in the Ruwiki project filelist:
http://rubyforge.org/frs/?group_id=84
-austin
Diff::LCS README
================
Diff::LCS is a port of Algorithm:iff[1] that uses the McIlroy-Hunt
longest common subsequence (LCS) algorithm to compute intelligent
differences between two sequenced enumerable containers[2]. The
implementation is based on Mario I. Wolczko's[3] Smalltalk version
(1.2, 1993)[4] and Ned Konz's[5] Perl version (Algorithm:iff)[6].
This release is version 1.1.0 with the addition of new features,
including the ability to #patch and #unpatch changes as well as a new
contextual diff callback, Diff::LCS::ContextDiffCallbacks, that should
improve the context sensitivity of patching.
Using this module is quite simple. By default, Diff::LCS does not
extend objects with the Diff::LCS interface, but will be called as if
it were a function:
require 'diff/lcs'
seq1 = %w(a b c e h j l m n p)
seq2 = %w(b c d e f j k l m r s t)
lcs = Diff::LCS.LCS(seq1, seq2)
diffs = Diff::LCS.diff(seq1, seq2)
sdiff = Diff::LCS.sdiff(seq1, seq2)
seq = Diff::LCS.traverse_sequences(seq1, seq2, callback_obj)
bal = Diff::LCS.traverse_balanced(seq1, seq2, callback_obj)
seq2 == Diff::LCS.patch!(seq1, diffs)
seq1 == Diff::LCS.unpatch!(seq2, diffs)
seq2 == Diff::LCS.patch!(seq1, sdiff)
seq1 == Diff::LCS.unpatch!(seq2, sdiff)
Objects can be extended with Diff::LCS:
seq1.extend(Diff::LCS)
lcs = seq1.lcs(seq2)
diffs = seq1.diff(seq2)
sdiff = seq1.sdiff(seq2)
seq = seq1.traverse_sequences(seq2, callback_obj)
bal = seq1.traverse_balanced(seq2, callback_obj)
seq2 == seq1.patch!(diffs)
seq1 == seq2.unpatch!(diffs)
seq2 == seq1.patch!(sdiff)
seq1 == seq2.unpatch!(sdiff)
By requiring 'diff/lcs/array' or 'diff/lcs/string', Array or String
will be extended for use this way.
Copyright
=========
# Copyright 2004 Austin Ziegler <[email protected]>
# adapted from:
# Algorithm:iff (Perl) by Ned Konz <[email protected]>
# Smalltalk by Mario I. Wolczko <[email protected]>
# implements McIlroy-Hunt diff algorithm
#
# This program is free software. It may be redistributed and/or modified
# under the terms of the GPL version 2 (or later), the Perl Artistic
# licence, or the Ruby licence.
#
# $Id: README,v 1.5 2004/08/08 20:33:09 austin Exp $
Footnotes
=========
[1] This library is called Diff::LCS because there are multiple
Ruby libraries called Algorithm:iff maintained by other authors.
[2] By sequenced enumerable, I mean that the order of enumeration is
predictable and consistent for the same set of data. While it is
theoretically possible to generate a diff for unordereded hash, it
will only be meaningful if the enumeration of the hashes is
consistent. In general, this will mean that containers that behave
like String or Array will perform best.
[3] (e-mail address removed)
[4] ftp://st.cs.uiuc.edu/pub/Smalltalk/MANCHESTER/manchester/4.0/diff.st
[5] (e-mail address removed)
[6] http://search.cpan.org/~nedkonz/Algorithm-Diff-1.15/