Question

I am using XSLT to transform HTML to XSL-FO. The fact that we are transforming from HTML to XSL-FO is probably irrelevant to answering this question.

For several of my values which use concat for font size and height, the output of concat is randomly duplicated for a single call.

Example: where the value of $lineheight-td is 13 for the following code:

<fo:block line-height="{concat($lineheight-td, 'pt')}">

this is the expected correct output that I usually receive:

<fo:block line-height="13pt">

this is incorrect output that I receive ~1% of the time:

<fo:block line-height="13pt13pt">

this is incorrect output that I receive < 1% of the time:

<fo:block line-height="13pt&#0;&#0;&#0;&#0;">

(note how unexpected output is 4 characters ('1', '3', 'p', 't') and the unexpected '' is produced 4 times).

These duplicate outputs generate from the concat call in different places from the same input HTML file on different iterations.

My current workaround is to re-transform if we get a bad output; so far this solves the problem, and proves this isn't data-related, but the duplicate output shouldn't be happening in the first place. I'm running under a multi-threaded Java environment, but I'm using a new Transformer for each individual transformation call, from a shared Template initialized one time at application start-up.

Why is this happening with concat, and how do I fix it?

The application uses an older Xalan library (2.7.0). I'm going to look into Xalan bugs and upgrading to Xalan 2.7.1 or higher.

Environment:

JBoss 7.2.0
JRE 1.7.0_45
xalan-2.7.0
xml-apis-1.3.04
xml-apis-ext-1.3.04

Java code:

String inputData = ...; // OUR HTML

Templates template = (Templates)templatesMap.get("HTML2FO");
Transformer transformer = template.newTransformer();
StreamSource streamSource = new StreamSource(new StringReader(data));
StringWriter writer = new StringWriter();
transformer.transform(streamSource, new StreamResult(writer));

String outputData = writer.toString();  // OUR FO

Answer 1

该应用程序使用较旧的Xalan库（2.7.0）。 我将调查Xalan错误并升级到Xalan 2.7.1或更高版本。

Xalan版本发行说明仅列出没有描述的错误号，并且在发布Xalan 2.7.1之后的某个时刻，Xalan开发切换到另一个错误跟踪器。关于他们旧的，＆＃34;固定＆＃34;的信息。虫子已经消失了。

没有任何东西＆＃34;证明＆＃34;任何给定的库都存在问题，我最终将所有与XML和转换相关的JAR文件更新为以下版本：

JAR FILE NAME           VERSION
serializer.jar          2.7.2
xalan.jar               2.7.2
xerxesImpl.jar          2.11.0
xml-apis.jar            1.4.01
xml-apis-ext.jar        1.4.01

使用原始XML库进行~8,000次转换时，我的成功率达到99％;失败的文件在第二次迭代时成功重新处理，这表明问题不在于我的数据。

对于使用更新的XML库进行约11,500次转换，我获得了100％的成功率。

我会继续测试，但是我可以得出结论，更新所有库似乎已经解决了问题。

Why is concat randomly duplicating output in this XSLT transformation?

1 个答案: