[WIP] Add hypothesis to some str methods

[eliasdorneles]

This is a WIP and not ready to merge, but I've been playing with hypothesis lately with help from @damedebugger (hi!).

I don't know if Str.java is the best place to start, but I keep finding new bugs for the existing methods so it'll take a little while for this to be ready.
It might make sense to split this into separate PRs, but as I also want to find a good way of reusing the strategies for other functions I'll continue working on this branch for now.

I hope this will be handy for ensuring quality of the patches, I know I'll be glad to not have to think so hard to cover the edge cases. :)

[eliasdorneles]

After thinking more about this, trying out things with @mistermocha and talking briefly with @freakboy3742, I now believe it will make more sense to limit the hypothesis usage for development and not on the regular Travis builds, for two reasons:

• it would increase the build time greatly
• it would find many problems that we have no clear solution yet, like the discrepancy between different versions of Unicode data

For reference, here are the Unicode data versions of CPython:

$ $(pythonz locate 3.4.4) -c 'import unicodedata; print(unicodedata.unidata_version)'
6.3.0
$ $(pythonz locate 3.5.3) -c 'import unicodedata; print(unicodedata.unidata_version)'
8.0.0
$ $(pythonz locate 3.6.1) -c 'import unicodedata; print(unicodedata.unidata_version)'
9.0.0

In the JVM land, Java 7 uses 6.0.0, Java 8 uses 6.2.0, Java 9 will be 7.2.0.

These causes differences some subtle differences of behavior for methods like isupper/islower and I believe isdigit as well.

[Zac-HD]

While the issues noted above are real, I'd also like to note that you can choose how many examples Hypothesis will try - 2000 might be unacceptably slow, but what about 50? You can also use it for some types but not others - Unicode is always tricky, but numeric and simple collection types could benefit enormously.

As noted on #580, I'd be happy personally and as a maintainer of Hypothesis to assist anyone who wants to take this on.

[phildini]

Hi there! It looks like this PR might be dead, so we're closing it for now. Feel free to re-open it if you'd like to continue, or think about directing your efforts to https://github.com/beeware/briefcase or https://github.com/beeware/toga. Both of these have more active development right now. 😄