$Id: CHANGES,v 1.37 2003/12/29 02:22:51 dfs Exp $
Version 2.0.8
o examples moved to an examples package and com.oroinc migration tool
moved to tools package.
o Fixed bug whereby compiling an expression with
Perl5Compiler.MULTILINE_MASK wasn't always having the proper effect
with respect to the matching of $ even though
Perl5Matcher.setMultiline(true) exhibited the proper behavior. For
example, the following input
" aaa bbb \n ccc ddd \n eee fff "
should produce "bbb ", "ddd ", and "fff " as matches for both the
patterns "\S+\s*$" and "\S+ *$" when compiled with MULTILINE_MASK.
Perl5Matcher was only producing the correct matches for the second
pattern, producing only "fff " as a match for the first pattern
unless setMultiline(true) had been called. This has now been fixed.
o Fixed embarrassing bug whereby an expression like (A)(B)((C)(D))+
when matched against input like ABCDE would produce matching groups
of: "A" "B" "" null "D" instead of "A" "B" "CD" "C" "D".
Version 2.0.7
o Made Perl5Util.toString() return null if no match exists in keeping
with the MatchResult interface. Previously, a NullPointerException
was being thrown.
o Fixed problem whereby an AwkMatchResult resulting from a match on an
AwkStreamInput source (AwkMatcher.contains(AwkStreamInput, Pattern);)
would contain offsets relative to the buffered input rather than to where
the input stream was first read from (usually the beginning of the stream).
o Added support for negating modifiers. For example (?-i)foo. Note,
we still do not support Perl 5.005+ group-local modifiers such as
(?imsx-imsx:pattern).
Version 2.0.7-dev-1
o Fixed a problem whereby offset information for captured groups was
considered out of bounds when capturing parentheses occurred in a
lookahead assertion that was not part of the actual match (occurring
immediately after the match).
o Changed processMatches to take Reader and Writer arguments. Reimplemented
old method in terms of new method. Added a version of old method that
allows you to specify input encoding. Changes suggested by Harald Kuhn.
o Changed behavior of Perl5Util.split() to match Perl's behavior, where
"leading empty fields are preserved, and empty trailing one are
deleted." Util.split() is left unchanged.
Version 2.0.6
o Added $& as a valid interpolation variable for Perl5Substitution.
The behavior of $0 was changed from undefined to the same as $&,
but it should be avoided since $0 no longer means the same thing
as $& in Perl. Only use $& if possible.
o Removed some leftover references to OROMatcher in the Perl5Util javadocs.
o Added an int substitute(...) method to Perl5Util to correspond to
the similar method added to org.apache.oro.text.regex.Util in v2.0.3
o Removed ant and support jars from distribution and moved build.xml to
top level directory. From now on, you must have ant installed on your
system to build the source.
o Added a strings.java example to the awk examples, better demonstrating
the character encoding issues associated with AwkMatcher's 8-bit
character set limitation.
Version 2.0.5
o Fixed [[:upper:]] so it would match lower case characters during case
insensitive matches.
o Fixed [[:punct:]] (which also affected [[:print:]] and [[:graph:]])
to conform to Single Unix Specification (some characters had been
omitted).
o Fixed bug whereby a - in a Perl expression would be ignored when
it followed a builtin character class like \w. In other words,
[\w-] behavig like \w instead of like [-\w]. The regression was
introduced with the unicode character class patch from version 2.0.2.
Version 2.0.4
o Deprecated Vector Perl5Util.split(String input) and replaced with
void Perl5Util.split(Collection results, String input). This
should have been done in an earlier release along with the other
split methods, but this method slipped through the cracks.
o Fixed bug in AwkMatcher that didn't properly handle PatternMatcherInput
matches when PatternMatcherInput had a non-zero begin offset.
o Added code to Perl5Matcher to handle bytecode generated for [[:alpha:]].
[[:alpha:]] would compile but not be interpreted.
o Fixed problem whereby MatchActionProcessor would never set
MatchActionInfo.pattern before calling MatchAction.processAction.
Also changed MatchActionInfo.fields from Vector to List.
o Added semicolon.txt input file for the semicolon.java example program.
o Fixed bug where RegexFilenameFilter() constructor with options argument
would not use the options in compiling the expression.
o Added READ_ONLY_MASK compilation option to GlobCompiler
Version 2.0.3
o Changed version information in javadocs to correspond to release
version rather than CVS version.
o Changed the backing store for GenericCache from Hashtable to Hashmap.
o Fixed a problem in Perl5Debug.printProgram where it wouldn't handle
the printing of the unicode character class bytecode correctly.
o Added links to Jakarta ORO home page in javadocs and converted
web site documentation to jakarta-site2 xml docs stored in xdocs
directory.
o Applied Mark Murphy's patch to add case modification support to
Perl5Substitution. The \u\U\l\L\E escapes from Perl5 double-quoted
string handling are now recognized in a substitution expression.
o Made a backwards-incompatible change in the Substitution interface.
The input parameter is now a PatternMatcherInput instance instead
of a String. A new substitute method was added to Util to allow
programmers to reduce the number of string copies in the existing
substitute() implementation. This required that the Substitution
interface be changed. The existing Substitution implementing classes
do not use the input parameter and it is very easy for anyone
implementing the Substitution interface in a custom class to update
their class to the new interface. A deprecation phase was skipped
because it would have still required implementors of the interface
to implement a method with the new signature.
Version 2.0.2
o Fixed default behavior of '.' in awk package. Previously it wouldn't
match newlines, which isn't how AWK behaves. The default behavior now
matches any character, but a compilation mask (MULTILINE_MASK) has been
added to AwkCompiler to enable the old behavior.
o Replaced the use of Vector with ArrayList in Perl5Substitution.
o Replaced the use of deprecated Perl5Util split method in printPasswd
example with newer method. Also updated splitExample to use
ArrayList instead of Vector.
o Updated RegexFilenameFilter to also implement the Java 1.1 FileFilter
interface.
o Applied a follow up to Takashi Okamoto's unicode/posix patch that
implemented negative posix character classes (e.g., [[:^digit:]])
Version 2.0.2-dev-2
o Applied a modified version of Takashi Okamoto's unicode/posix patch.
It adds unicode support to character classes and adds partial support
for posix classes (it supports things like [:digit:] and [:print:], but
not [:^digit:] and [:^print:]). It will be improved/optimized later, but
gives people the functionality they need today.
Version 2.0.2-dev-1
o Removed commented out code and changed OpCode._isWordCharacter() to
use Character.isLetterOrDigit()
o Some documentation fixes.
o Perl5Matcher was not properly tracking the current offset in
__interpret so PatternMatcherInput would not have its current offset
properly updated. This bug was introduced when fixing the
PatternMatcherInput anchor bug. When the currentOffset parameter
was added to __interpret, not all of the necessary
__currentOffset = beginOffset assignments were changed to
__currentOffset = currentOffset assignments. This has been
fixed by updating the remaining assignments.
o The org.apache.oro.text.regex.Util.split() method was further
generalized to accept a Collection reference argument to store the result.
Version 2.0.1
o Jeff ? (jlb@houseofdistraction.com) and Peter Kronenberg
(pkronenberg@predictivetechnologies.com) identified a bug in the
behavior of PatternMatcherInput matching methods with respect to
anchors (^). Matches were always being performed interpreting the
beginning of the string from either a 0 or a current offset rather
than from the beginOffset. Essentially, the beginning of the string
for purposes of matching ^ wasn't being associated with beginOffset.
This problem has been fixed.
o A problem with multiline matches that was introduced in the
transition from OROMatcher/Perltools/etc. to Jakarta-ORO
was fixed.
o The org.apache.oro.text.regex.Util.split() method was generalized to
accept a List reference argument to store the result.
Version 2.0
Many small, but important, changes have been made to this software
between its last release as ORO software and its first release as part
of the Apache Jakarta project. The last versions of the ORO software
were:
OROMatcher 1.1 (com.oroinc.text.regex)
PerlTools 1.2 (com.oroinc.text.perl)
AwkTools 1.0 (com.oroinc.text.awk)
TextTools 1.1 (com.oroinc.io, com.oroinc.text, com.oroinc.util)
OROMatcher 1.1 was fully compatible with Perl 5.003 regular expressions.
Many changes have been made to the regular expression behavior in
successive versions of Perl. The goal is to update org.apache.oro.text.regex
to the latest version of Perl (5.6 at this time), but only some of this
work has been done.
o Guaranteed compatibility with JDK 1.1 is discontinued. JDK 1.2 features
may be used indiscriminately, even though they may not have been used
at this time.
o Perl5StreamInput and methods manipulating Perl5StreamInput have been
removed. For the technical reasons behind this decision see
"On the Use of Regular Expressions for Searching Text", Clark and Cormack,
ACM Transactions on Programming Languages and Systems, Vol 19, No. 3,
pp 413-426.
o Default behavior of '^' and '$' as matched by Perl5Matcher has
changed to match as though in single line mode, which appears to be
the Perl default. Previously if you did not specify single line or
multi-line mode, '^' and '$' would act as though in multi-line mode.
Now they act as though in single line mode, unless you specify otherwise
with setMultiline() or by compiling the expression with an appropriate mask.
However, '.' acts as in multi-line mode by default and its behavior
can only be altered by compiling an expression with SINGLELINE_MASK.
setMultiline() emulates the Perl $* variable, whereas the MASK variables
emulate the ismx modifiers.
o All deprecated methods have been removed. This includes various
substitute() convenience methods from com.oroinc.text.regex.Util
o Javadoc comments have been updated to use 1.2 standard doclet features.
