yankrasny / repair-partitioning Goto Github PK

See printWordFragments for code that can be adopted to store the fragments that were found. Then scan the wordIDs for each fragment and record its frequency.

That ^ seems like it would be O(n^2). Ok, think about this some more.

We get a seg fault in the middle of calcOffsets, where our current node is at an invalid mem location. I think the tree might be really huge, and calcOffsets recurses twice, so perhaps we have a stack overflow.

Calling the code as a library using the runRepairPartitioning() function circumvents the functionality in main(). The wordID will not be set properly, but I'm not sure if that's really a problem for repair.

1. Investigate
2. Adapt main() to work with wordIDs instead of text

Add numLevelsDown or maxRecursiveCalls.
Remove old ones you're not using.

Also refactor jinru's code to take your params

In the getBestSubset function, If we hit maxNumCalls then set a variable to true. So if a recursive call seta that variable to true, that also sets yours to true. At that point just return.

The calling code checks that variable and if it's true, reruns the getBestSubset function with different params to try and reduce numCalls