On Jul 19, 2005, at 11:19 AM, Jeff Davis wrote:
Hi,
I'm seeing some strange behavior in the way the QueryParser handles
consecutive backslash characters. I know that backslash is the escape
character in Lucene, and so I would expect "\\\\" to match fields that
have two consecutive backslashes, but this does not seem to be the
case.
The fields I'm searching are UNC paths, e.g. "\\192.168.0.15\public".
The only way I can get my query to find the record containing that
value is to type "FieldName:\\\192.168.0.15\\public" (three slashes).
Why is the third backslash character not treated as an escape? Is it
just that any backslash that is preceded by a backslash is interpreted
as a literal backslash character, regardless of whether the "escape"
backslash was itself escaped?
I can code around this, but it seems inconsistent with the way that
escape characters usually work. Is this a bug, or is it intentional,
or am I missing something?
I've waited until I had a chance to experiment with this before
replying. I say that this is a bug. There is a private method in
QueryParser called discardEscapeChar (shown below). I copied it to a
JUnit test case and gave it this assert:
assertEquals("\\\\\\\\192.168.0.15\\\\public", discardEscapeChar
("\\\\192.168.0.15\\\\public"));
This test fails with:
Expected:\\\\192.168.0.15\\public
Actual :\192.168.0.15\public
Which is wrong in my opinion. (though my head hurts thinking about
metaescaping backslashes in Java code to make this a proper test)
The bug is isolated to the discardEscapeChar() method where it eats
too many backslashes. Could you have a shot at tweaking that method
to do the right thing and submit a patch?
private String discardEscapeChar(String input) {
char[] caSource = input.toCharArray();
char[] caDest = new char[caSource.length];
int j = 0;
for (int i = 0; i < caSource.length; i++) {
if ((caSource[i] != '\\') || (i > 0 && caSource[i-1] == '\\')) {
caDest[j++]=caSource[i];
}
}
return new String(caDest, 0, j);
}
Erik
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
