[
https://issues.apache.org/jira/browse/DIGESTER-120?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12579019#action_12579019
]
Simon Kitching commented on DIGESTER-120:
-----------------------------------------
Thanks very much for the bugreport, and your great analysis of the problem.
I think you are right that the NodeBuilder.characters method is being too eager
to strip whitespace.
However I am unable to actually trigger the bug using a unit test case. The
existing unit test code for digester has something very similar to your
example, but with all xml in one line. I therefore added linefeeds as you
suggested, but the unit test still passes.
The unit tests I'm talking about can be found here:
http://svn.apache.org/repos/asf/commons/proper/digester/trunk/src/test/org/apache/commons/digester/NodeCreateRuleTestCase.java
If you could manage to create a unit test for Digester that reproduces the
problem that would be really helpful. I'm not sure it is going to be easy
though; I think the cause of the problem is that the SAX api says that
characters(..) can be invoked by the xml parser as many times as the parser
wants. Digester is currently assuming (as you point out) that it is invoked
only once with the complete code block. But adding linefeeds to the input
doesn't appear to force the xml parser i'm using to call characters() with
partial strings.
In short, I think that this bug is very much dependent on which xml parser
implementation is used, and what its internal parsing buffer size happens to be
set to etc. That will make it difficult to get a unit test to reproduce this
bug.
Which xml parser are you using?
I've created a patch anyway, which looks ok to me and passes the existing
digester test cases. Could you try applying this patch and seeing if it fixes
your problem?
> digesting xml content with NodeCreateRule swallows spaces.
> ----------------------------------------------------------
>
> Key: DIGESTER-120
> URL: https://issues.apache.org/jira/browse/DIGESTER-120
> Project: Commons Digester
> Issue Type: Bug
> Affects Versions: 1.8
> Environment: jdk 1.4.2_08, digester 1.8
> Reporter: Nguyen Thanh Son Daniel
>
> i need to process an xml file that contains entities: ie:
> <?xml version="1.0" encoding="UTF-8"?>
> <top>
> <body>A A</body>
> </top>
> i'm using digester as follows:
> Digester digester = new Digester ();
> digester.addRule ("top", new ObjectCreateRule (MyContent.class));
> digester.addRule ("top/body", new NodeCreateRule ());
> digester.addSetNext ("top/body", "setBody");
> then
> ...
> digester.parse (file);
> MyContent class transforms the node into text as follows:
> public class MyContent
> {
> public void setBody (Element node)
> {
> String content = serializeNode (node);
> System.out.println (content);
> }
> ...
> }
> the content displayed is in this case: <body>AA</body>
> if the body was encoded in the xml file as: <top><body>A A</body></top>, the
> content would then be correctly displayed as:
> <body>A A</body>
> looking at the NodeCreateRule.NodeBuilder.characters () implementation, the
> following code generates the problem:
> String str = new String(ch, start, length);
> if (str.trim().length() > 0) {
> top.appendChild(doc.createTextNode(str));
> when entities are being used; the characters () method is called for 'A', ' '
> and 'A' in the first case. in the second case, it is called once with 'A A'.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
