Scripts should output the "discovered" value after conducting a binary search

[joshuaehill]

It would be helpful to output the value that was found in the binary search. Log is a 1-1 function, this can nominally be recovered, but it would be generally useful to just output the value after it is calculated.

[dj-on-github]

My half implemented plan is to pull the binary chop into a shared function used by the tests. I'll add that in with the refactoring.

[joshuaehill]

That's a good approach, as there are lots of corner cases for the binary search code to catch, and its obnoxious trying to keep a bunch of delicate separate implementations in sync.

I submitted some binary search to the NIST python tool; it deals with the problems that I'm aware of.
usnistgov/SP800-90B_EntropyAssessment#19

[dj-on-github]

I hope you don't mind if I steal that binary chop code you submitted against the NIST code?

[joshuaehill]

That's fine.

[joshuaehill]

As an aside, for biased outputs, the current binary search code seems to be missing the value it is looking for.

I've put some data files (and corresponding results, with extensive debugging information) here:
https://www.untruth.org/~josh/sp80090b/sp800-90b-noniid-testcases-20180920.zip