java.util.zip.ZipException: invalid code lengths set with SentenceModel

Question

I'm using OpenNLP for syntactic analysis. I wrote these lines to load the Sentence detector model:

    InputStream is = getClass().getClassLoader().getResourceAsStream("models/en-sent.bin") ;            
    SentenceModel sModel = new SentenceModel(is);

the above line throws the exception: java.util.zip.ZipException: invalid code lengths set.

The "models/en-sent.bin" is located under "src/main/resources/models/en- token.bin".

When I try to print the file path using:

   System.out.println(getClass().getClassLoader().getResource("models/en-sent.bin"));           

The path shows as : "file:/C:/Users/aaa/workspace/qa/target/classes/models/en-sent.bin". if I load the model using:

    InputStream is =  new FileInputStream("src/main/resources/models/en-sent.bin") ;
    SentenceModel sModel = new SentenceModel(is);

It works fine, but this method will not work when I build the file into a runnable jar. Any help why this Exception is thrown?

The full stack exception is here:

java.util.zip.ZipException: invalid code lengths set
at java.util.zip.InflaterInputStream.read(Unknown Source)
at java.util.zip.ZipInputStream.read(Unknown Source)
at java.io.FilterInputStream.read(Unknown Source)
at java.util.Properties$LineReader.readLine(Unknown Source)
at java.util.Properties.load0(Unknown Source)
at java.util.Properties.load(Unknown Source)
at opennlp.tools.util.model.PropertiesSerializer.create(PropertiesSerializer.java:31)
at opennlp.tools.util.model.PropertiesSerializer.create(PropertiesSerializer.java:27)
at opennlp.tools.util.model.BaseModel.loadModel(BaseModel.java:224)
at opennlp.tools.util.model.BaseModel.<init>(BaseModel.java:173)
at opennlp.tools.sentdetect.SentenceModel.<init>(SentenceModel.java:91)
at com.elsevier.sherpath.syntactic_analysis.QuestionNLPAnalysis.<init>(QuestionNLPAnalysis.java:88)
at com.elsevier.sherpath.main.QuestionProcessing.syntacticAnalysis(QuestionProcessing.java:255)
at com.elsevier.sherpath.main.BatchProcess.main(BatchProcess.java:23)

Thanks

Solution 1

You can set resource filtering false and include that bin file, so that it will not be affected in mvn install.

<resource>
    <directory>your resources directory</directory>
    <filtering>false</filtering>
        <includes>
            <include>your-bin.bin</include>
        </includes>
</resource>

The cause of the exception is maven is trying to recreate your bin file using pom.xml properties and, as a result of that, the bin file in the target directory will not be same as in your resource folder. Therefore set resource filtering false for the bin file.

Solution 2

It works fine, but this method will not work when I build the file into a runnable jar. Any help why this Exception is thrown?

   InputStream is =  new FileInputStream("src/main/resources/models/en-sent.bin") ;

uses a system-dependent file name.
When you are not in a JAR, it uses as base path the folder of you project.
When you are in a JAR, it uses as base path the JAR itself.
So it cannot work.

You have two ways to solve your problem :

• either you use a classpath resource loading by specifying the resource that is in the classpath. For example InputStream is = YourClass.class.getResourceAsStream("/models/en-sent.bin");
  In your tries, I think that you have forgotten the "/" at the beginning of the path.
  Without the "/" you use a path relative to the package where the method is invoked.
  

• or you move the resource outside the JAR and you access it with an absolute path. For example : InputStream is = new FileInputStream("file:///D:/models/en-sent.bin");.

Comments

[Elshaimaa Ali]

Thanks for the response. I used InputStream is = loader.getClass().getResourceAsStream("/models/en-sent.bin")‌​; but SentenceModel sModel = new SentenceModel(is);

[davidxxx]

You are welcome :) Instead of loader, try with a class of your application.

[Elshaimaa Ali]

the path using this line: System.out.println(loader.getClass().getResource("/models/en‌​-sent.bin").getPath(‌​)); the result is : /C:/Users/alishaim/workspace/question-assessment/target/clas‌​ses/models/en-sent.b‌​in the path using this line: System.out.println(loader.getClass().getResource("/models/en‌​-sent.bin").toURI().‌​toString(); the result is : file:/C:/Users/alishaim/workspace/question-assessment/target‌​/classes/models/en-s‌​ent.bin Does the format of the path is the problem. I can' use absolute paths, that won't work when running this from the Jar

[Elshaimaa Ali]

I tried with ClassName.Class().getResourceAsStream but still gettin the same problem.

[wcolen]

What is your class package? The first "/" is important. Without it the path is relative. Use getResourceAsStream. The getResource will not work outside the IDE.

[Elshaimaa Ali]

So I added the "/" When I use InputStream is = getClass().getResourceAsStream("/models/en-sent.bin") ; This line : sModel = new SentenceModel(is); throws a java.util.zip.ZipException: invalid code lengths set, when I run it from eclipse When I use: InputStream is = new FileInputStream ("absolute path/en-token.bin"); It works fine through the IDE but of course won't work from the jar because of the path.

[Rajeev Rathor]

I faced this issue and struggled 1 days. Tried many plugin, solution suggested on stackoverflow in POM.xml. But it was waste. Finally I removed all the contents from .m2 directory and tried to build jar from command "mvn clean package" . It perfectly working for me.

[Jonas_Hess]

I also just recompiled the project and the error was gone