Getting rid of `new File(URL.getPath()).toURI()`

[mkarg]

Getting rid of `new File(URL.getPath()).toURI()`

[jansupol]

The reason for filing the issue was not performance related at the beginning, the problem was that Jersey did not build on Windows. The issue showed the reason, Paths.get(MyTest.class.getResource("").getPath()); failed there and new File() fixes that.

The performance degradation comes from the additional operations the new File performs. getResource("").getPath() starts with /<drive>: and the linux-root slash at the beginning is normalized by new File(). But the NIO does not understand that linux-root slash on Windows.

[AlanBateman]

The reason for filing the issue was not performance related at the beginning, the problem was that Jersey did not build on Windows. The issue showed the reason, Paths.get(MyTest.class.getResource("").getPath()); failed there and new File() fixes that.

Hopefully it is clear now that MyTest.class.getResource("").getPath() returns the path component of the URL, it's not a file path.

The performance degradation comes from the additional operations the new File performs. getResource("").getPath() starts with /: and the linux-root slash at the beginning is normalized by new File(). But the NIO does not understand that linux-root slash on Windows.

On Windows, the new implementation will of course convert forward slashes to back slashes, the issue I think you are running into seems to be "/C:/" which comes about from treating the URL path component as a file path.

[mkarg]

@jansupol I would propose that I set up a test environment so I can check Alan's ideas. Regarding performance I do not see where this is in the critical path. Do you have a pointer for me where performance plays a role here, so I keep checking that special case then?