Create 'fcalib' Github Project

[JanFK]

This is how I imagine the course of action:

• Merge fcalib & fcaapi into just fcalib and create the respective Github project in the fcatools organization
• Copy everything that is reusable & fca-related from our project into fcalib (i.e. also the association related code, as well as the burmeister code etc.)
• Export this new fcalib as a jar (or better put it in the maven central repository), use it as a dependency in ConExp-NG & remove the code that we moved to fcalib
• Initiate a discussion on what the API currently provides and what not (e.g. is the complexity burden of generics worth it?)

This is as much as I would do as I don't have time anymore. But after the above steps are done there is still a lot of things to do. I would create the following two milestones with the according issues in the fcalib project if Baris agrees:

• Milestone 1 - "Freeze the semantics / Make the current functionality testable"
  • Make our <String,String>-specialised code generic (solely a Sisyphean task I imagine)
  • Create (a lot of) unit tests for all the functionality fcalib provides
  • Create a thorough benchmark suite to be able to measure performance improvements/regressions
  • Improve the documentation
• Milestone 2 - "Make the code harder, better, faster"
  • Improve the efficiency of the current code by e.g.
    • changing the API a bit and rewriting our inefficent implementations to take advantage of these changes
    • looking at Colibri
    • reading through Fast FCA
    • and, of course, measuring and documenting the improvements with the benchmark suite without breaking the semantics using the unit tests as an assistance
  • Try to improve the API further by e.g.
    • thinking about the benefits/disadvantages of generics (find use cases and if there aren't any remove them)
    • looking at best practices in Java
    • watching How To Design A Good API and Why it Matters
    • getting input from users (also asking Baris Sertkaya)
    • looking at other "fcalib"-projects in possibly different programming languages
    • and, of course, trying to use the api e.g. by creating small tutorials / user documentation
  • Try to add useful new fca methods/code by e.g.
    • looking at other "fcalib"-projects in possibly different programming languages, e.g. conexp-clj
    • looking at some FCA papers

I (@eugenkiss) feel this would be a perfect task that Mr.Jäschke (@rjoberon) could write out for someone to work on. I'd actually love to do it myself but I wouldn't be able to stark the task before summer 2014 (before that I'll be very busy and I haven't even participated in Mr.Jäschke's FCA lecture!). Of course, as the author Baris Sertkaya needs to agree with the plan as well as Mr.Jäschke. Anyway, I just wanted to write down my ideas.

[eugenkiss]

This won't happen during our remaining development cycle. What a pity, but it can still be finished in the future.