cmusphinx / sphinx4

Pure Java speech recognition library
cmusphinx.sourceforge.net
Other
1.4k stars 587 forks source link

sphinx-data doesn't work in modular Java apps #86

Open Thorn1089 opened 5 years ago

Thorn1089 commented 5 years ago

Because of changes made to how Class.getResource(String) works in Java 9 and above, simply including the sphinx-data JAR does not work to load the English acoustic model, dictionary, etc.

The problem is here:

public static URL resourceToURL(String location) throws MalformedURLException {
    Matcher jarMatcher = jarPattern.matcher(location);
    if (jarMatcher.matches()) {
      String resourceName = jarMatcher.group(1);
      return ConfigurationManagerUtils.class.getResource(resourceName);
    } else {
      if (location.indexOf(58) == -1) {
        location = "file:" + location;
      }

      return new URL(location);
    }
  }

Using a location such as resource:/edu/cmu/sphinx/models/en-us/en-us will no longer work. This is because Class.getResource will delegate to the classloader and pass the module name (in this case derived from the JAR file as "sphinx.core"). The problem is that the resources are in a different JAR, which gets the automatic module name "sphinx.data". Since the classloader will only look within the "sphinx.core" module, it won't find the resources from "sphinx.data."

FriedaSmith commented 4 years ago

@TomRK1089 Hello. Have you solved the problem? And how to solve it?

Thorn1089 commented 4 years ago

I had to copy the data files out of the JAR, repackage them in my own JAR, and extract them to the filesystem at startup. There was no simple workaround.

FriedaSmith commented 4 years ago

Thank you very much.