Feature/nav 181 initialize hashset size of marked stops

[clukas1]

Created new pull request, to allow you to review the request.

Change is that boolean[] arrays are used to mask routesToScan and markedStops instead of HashSets. Performance has been improved significantly.

Original:

With Hashset set to size of Stops/Routes

With Boolean Arrays

[munterfi]

Thanks @clukas1, looks good.

It is a bit difficult to read the code and follow in the debugger due to passing around all the arrays.

In this commit 9236013 on the branch feature/NAV-181-extract-marked-stops-class I extracted a MarkedStops class to encapsulate the logic of the stop mask. In performance there is not really a difference (some ms faster):

Before:

After:

Then i tried to get rid of the array copy / new allocation in 7e69a26, by using an int array instead boolean, tracking the rounds. But somehow the logic messes up...

Should we introduce a MarkedStops (or other name, e.g. StopMask) class, to track the marked stops? And possibly omit the rest on the branch...

[clukas1]

@munterfi I've moved the logic of markedStopsMasks to the Query State because they fit in there nicely. The int array will not work, because you will sometimes overwrite a to be checked mark to the new round potentially preventing finding ideal solutions.

benchmarking the both solutions (before and after refactoring) shows following results, probably just noise holding them apart.

#before refactoring

#after refactoring

[munterfi]

Thanks, looks good to merge.