Huffman Minimization Mark List Initialization Issue

[Versifex]

The initialization of the List called "mark" in MinimizationOperations.MinimizeHuffman is incomplete, and consequently an Index Out of Bounds error is produced for "mark[n1][n2] = true;" (This is of course only when Automaton.Minimization is set to Automaton.MinimizeHuffman).

[moodmosaic]

Thank you for reporting this – The reason is mainly behind the methodology that has been used in to port the code from dk.brics.automaton and xeger.

For each test in the RegexPatternTestCases class we were porting the smallest possible amount of code in order to make the test pass.

However, if we have to port further we need to start with a failing test case in the NotSupportedRegexPatternTestCases class.

I think it would be great if you provide a concrete test case :)

[Versifex]

Hi Nikos,

Near as I can tell, the MinimizationOperations.MinimizeHuffman method will fail with any input whatsoever for the reason I described... it's just a matter of switching the hard-coded Automaton.Minimization to Automaton.Huffman.

My need for this fix has passed however... I was experiencing the documented .* issue with the default minimization (where the program would never return while running Xeger against a regex like ._G._V.X.), and just wanted to see if the other minimization algorithms handled it better. I've since addressed the issue by stripping off any leading or trailing .* (which I made work for my purposes), and now the default minimization runs just fine.

Otherwise, while I have some experience writing unit tests, it is limited and not very recent. So I fear I would only embarrass myself trying to assist in this area.

[moodmosaic]

If you have modified the code to suit your needs you can always send a Pull Request so we have something concrete to discuss.

Pull Requests are more than welcome :)