OPENNLP-936: Add thread-safe versions of POSTaggerME, SentenceDetecto…

[twgoetz]

…rME and TokenizerME. Include test case as well.

I'm open to changing the names of the classes, if anybody has a better idea.

[kottmann]

Thanks!

• Indent by 2 spaces (not 4), checkstyle is giving many warnings about that (click the green check mark to see the travis-ci output)
• Each new file needs to have AL 2.0 license header
• Commit messages subject line is limited to 72 chars, please make it short and also use the body of it
• We strongly recommend to set up Eclipse or Intellij with the style files from our code convention page (https://opennlp.apache.org/code-conventions.html)
• Update the commit above e.g. via git commit --amend or squash and then force push (git push -f) to update the PR

[twgoetz]

Ok done.

[linekin]

POSModel is not thread-safe because GISModel uses HashMap which is not thread-safe.

[danielruss]

There is no need to change the values in the map after the model is create. It should be a Collections.unmodifiableMap()... which would be Thread.Safe().

[mawiesne]

This issue (OPENNLP-936) is still relevant to many users. Is there any form of consensus - in 2022 - on how to progress with the current code base?

[jzonthemtn]

The comments on this PR seem fairly trivial. I have no objections to getting those comments addressed and merging this. I will go through the PR again and see if there's anything since then that should be addressed.

Thanks @mawiesne for bringing attention to this one.

[bachelarius]

Hi all, looks like this thread has broadly reached consensus back in 2022, but not yet been merged - what needs to happen to get it looked at again?

I've verified that the current version of the SentenceDetectorME indeed suffer from this issue and implementing the solution originally proposed by @twgoetz fixes it.

[rzo1]

@bachelarius Think it boils down to (a) rename the classes, (b) add the @threadsafe annotation for tracking and (c) rebase against latest changes from main.

[rzo1]

That being said: PRs are welcome.

[rzo1]

I rebased this PR, so we get CI feedback.

[rzo1]

https://ci-builds.apache.org/job/OpenNLP/job/eval-tests-configurable/24/

[mawiesne]

Thx @twgoetz, @rzo1 and other feedback providers!

[mawiesne]

Eval build no. 26 passed: https://ci-builds.apache.org/job/OpenNLP/job/eval-tests-configurable/26/