Re: [Ant-contrib-developers] propertyregex: '\' within patterns

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 454-5900

OK, an update: if I write the property to a file and
use <replaceregexp> the behaviour is the same: I must
*quadruple* the backslashes in the replace string.
Since this appears universal in Ant regexps, I guess
it is here to stay.

Is it to do with escaping for a pass of interpretation
by ant (which halves the number of '\'), then escaping
for a pass of interpretation by regexp (again halving
the number of '\')?

Anyway, since this appears canonical behaviour, I think
it is a good idea to document it in ant replaceregexp
and cpp-contrib propertyregex.

Rutger

P.S. One seriously minor nibble: why is the nomenclature
      of the global flag different in propertyregex and
      replaceregexp? global="true" <-> flags="g".

Rutger Hofman wrote:
> Thanks for your effort, to start with!
> 
> Inger, Matthew wrote:
> 
>> You will need to upgrade to ant 1.5.3-1
>> I don't the exact version where it's fixed, but there
>> was apparantly a problem with the way substititions were
>> being done, where the code was pulling a character beyond
>> the end of the string, giving an exception.  (The exception
>> happens when i run within IntelliJ, which uses it's own
>> ANT, but not when i use the 1.5.3-1 version).
> 
> 
> I run either ant 1.5.4 or ant 1.6alpha, last night's build,
> and now I also tried our previously installed version 1.5.3-1.
> Same behaviour everywhere.
> Besides, does propertyregex really depend on ant i.s.o. the
> regexp stuff underlying both ant and ant-contrib, or on its
> own ant-contrib stuff? Maybe I have some wrong regexp
> package -- though replaceregexp works fine for me.
> 
>> Note:  That the "regexp" attribute needs you to double
>> the backslash character, as specified by the perl re syntax.
>> However, the replacement should be done as is, since it
>> is interperated as a literal string, with special processing
>> done for the backreference insertion.  ANT is not using the
>> native substite methods (such as org.apache.regexp.RE.substitute)
> 
> 
> This does not really match what I see... Also, I cannot believe that
> a '\' to be used in replace must not be escaped, since \0, \1 etc
> (and I do not know how many more special strings, e.g. \n) are
> treated specially. Otherwise, how could replace distinguish between
> \0 as "the first ()-ed match" and \0 as the literal string '\0', or
> a newline and the string '\n'?
> 
> Rutger
> 
>>         <property name="p-in" value="\\abc\\def" />
>>         <echo message="${p-in}" />
>>         <propertyregex
>>             property="p-out"
>>             input="${p-in}"
>>             regexp="\\\\"
>>             replace="\"
>>             global="true" />
> 
> 
> (BTW, this implements the reversed substitution from what I intended: it
> does "\\" -> "\" i.s.o. "\" -> "\\".)
> 
> This example gives me, for all of ant-1.5.3-1, ant-1.5.4 and ant-1.6alpha:
> 
> init:
>     [echo] \\abc\\def
> 
> BUILD FAILED
> 
> java.lang.StringIndexOutOfBoundsException: String index out of range: 1
>        at java.lang.String.charAt(String.java(Compiled Code))
>        at java.util.regex.Matcher.appendReplacement(Matcher.java:543)
>        at java.util.regex.Matcher.replaceAll(Matcher.java:647)
>        at 
> org.apache.tools.ant.util.regexp.Jdk14RegexpRegexp.substitute(Jdk14RegexpRegexp.java:112) 
> 
>        at net.sf.antcontrib.property.RegexTask.doReplace(Unknown Source)
>        at net.sf.antcontrib.property.RegexTask.execute(Unknown Source)
>        at 
> org.apache.tools.ant.UnknownElement.execute(UnknownElement.java:193)
>        at org.apache.tools.ant.Task.perform(Task.java:341)
>        at org.apache.tools.ant.Target.execute(Target.java:309)
>        at org.apache.tools.ant.Target.performTasks(Target.java:336)
>        at org.apache.tools.ant.Project.executeTarget(Project.java:1339)
>        at org.apache.tools.ant.Project.executeTargets(Project.java:1255)
>        at org.apache.tools.ant.Main.runBuild(Main.java:609)
>        at org.apache.tools.ant.Main.start(Main.java:196)
>        at org.apache.tools.ant.Main.main(Main.java:235)
>                                                                                
> The intended effect is achieved for regexp="\\\\" replace="\\\\",
> so in replace the backslash must be *quadrupled*.
> Is this different from the behaviour you see?
> My system: Linux, RedHat 7.4 (or near that), JAVA_HOME is IBMJava2-14
> or sun j2sdk1.4.2, ant-contrib 0.4 (else, no propertyregex of course)
> downloaded today, ant 1.5.3-1, 1.5.4 or 1.6alpha. For completeness, I can
> also test under Windows XP (but I will believe in different behaviour
> only when I see it :-).
> 
> (BTW, this implements the reversed substitution that I intended: it does
> "\\" -> "\" i.s.o. "\" -> "\\".)
> 
> Maybe I should take a look at the code, after all... No problem for sure.
> As a first step, I will grab a cvs source distro of cpp-tasks.
> 
> Rutger Hofman
> VU Amsterdam
> 
>> ----------------
>> Matthew Inger [in...@sy...]
>> Software Developer
>> Synygy, Inc
>> 610-664-7433 x 7770
>> "Man who jump off cliff, leap to conclusions." - Confucious
>>
>>
>> -----Original Message-----
>> From: Rutger Hofman [mailto:ru...@cs...]
>> Sent: Thursday, September 04, 2003 1:29 PM
>> To: ant...@li...
>> Subject: Re: [Ant-contrib-developers] propertyregex: '\' within patterns
>>
>>
>> Why do I have to quadruplate the backslash? Why in replace and
>> not in regexp?
>> I would expect, to replace '\' with '\\':
>>    regexp="\\" replace="\\\\"             (2 and 4)
>> or if I must quadruplate, then I would expect
>>    regexp="\\\\" replace="\\\\\\\\"       (4 and 8)
>>
>> It is like none of these. Specifying 8 '\' in replace also
>> gives the correct result. So, 2 and 8 is OK, as is 2 and 7.
>> However, 4 and 8 gives a failure to match the pattern:
>> 4 apparently matches 2 '\' in the input string.
>>
>> (In fact, I would not have been surprised if the seventh
>> backslash would have escaped the following end-quote, but ant
>> life seems not so simple.)
>>
>> Greetz,
>>
>> Rutger Hofman
>> VU Amsterdam
>>
>> Inger, Matthew wrote:> Backslashes should be in increments of two if 
>> you're
>>  > actually trying to represent the backslash character.
>>  > So, you should probably have 8 backslashes.
>>  >
>>  > ----------------
>>  > Matthew Inger [in...@sy...]
>>  > Software Developer
>>  > Synygy, Inc
>>  > 610-664-7433 x 7770
>>  > "Man who jump off cliff, leap to conclusions." - Confucious
>>  >
>>  >
>>  > -----Original Message-----
>>  > From: Rutger Hofman [mailto:ru...@cs...]
>>  > Sent: Thursday, September 04, 2003 12:13 PM
>>  > To: ant...@li...
>>  > Subject: [Ant-contrib-developers] propertyregex: '\' within patterns
>>  >
>>  >
>>  > Another question about propertyregex.
>>  >
>>  > I want to basically set property p-out from replacing
>>  > "\" with "\\" in some property, p-in. Now, '\' is of
>>  > course escape character, so I must replace it. I would
>>  > presume that regexp="\\" and replace="\\\\" would do
>>  > the job.
>>  >
>>  > However, I must do:
>>  >
>>  > <propertyregex
>>  >       property="p-out"
>>  >       input="${p-in}"
>>  >       regexp="\\"
>>  >       replace="\\\\\\\"
>>  >       global="true" />
>>  >
>>  > This is, yes, seven(!) backslashes in replace,
>>  > where I would expect four.
>>  >
>>  > Other numbers of backslashes in replace lead to
>>  >
>>  > java.lang.StringIndexOutOfBoundsException: String index out of 
>> range: 3
>>  >          at java.lang.String.charAt(String.java:460)
>>  >          at 
>> java.util.regex.Matcher.appendReplacement(Matcher.java:544)
>>  >
>>  > or other errors.
>>  >
>>  > Any clue? Is this going to be required (and documented) behaviour,
>>  > or is it a bug?
>>  > If you want, I can spend some time to hunt down the bug (if my 
>> professor
>>  > allows it :-)
>>  >
>>  > Rutger
>>  >
>>  >
>>  >
>>  >
>>  > -------------------------------------------------------
>>  > This sf.net email is sponsored by:ThinkGeek
>>  > Welcome to geek heaven.
>>  > http://thinkgeek.com/sf
>>  > _______________________________________________
>>  > Ant-contrib-developers mailing list
>>  > Ant...@li...
>>  > https://lists.sourceforge.net/lists/listinfo/ant-contrib-developers
>>  >
>>
>>
>>
>>
>>
>> -------------------------------------------------------
>> This sf.net email is sponsored by:ThinkGeek
>> Welcome to geek heaven.
>> http://thinkgeek.com/sf
>> _______________________________________________
>> Ant-contrib-developers mailing list
>> Ant...@li...
>> https://lists.sourceforge.net/lists/listinfo/ant-contrib-developers
>>
> 
> 
> 
> -------------------------------------------------------
> This sf.net email is sponsored by:ThinkGeek
> Welcome to geek heaven.
> http://thinkgeek.com/sf
> _______________________________________________
> Ant-contrib-developers mailing list
> Ant...@li...
> https://lists.sourceforge.net/lists/listinfo/ant-contrib-developers