Experiments for the graph golf challenge.

Although this is (was) just a personal playground, feel free to join in!

To configure this tool according to your hardware, please see the bottom of the file lib/enhancers.py.

• see how far we can get with an implementation that runs on PyPy3
  • no C modules etc.
• semantically
  • if beneficial for performance (measure!)
    • implement as much domain knowledge as possible
    • do invariant and sanity checks using assert statements
      • so we can easily skip them with the "-O" option
      • also, we skip some debug statements using assert None is debug(...)
        • nasty, sorry
    • classes/instances do only what is inherently needed
      • more responsibility for callers
        • callers often can act more precise/targeted since they tend to have more information available
      • e.g. a function expects a list with unique items as argument; instead of just making the items in the list unique, the function requires the caller to supply a list with only unique items; the caller - with its potential additional knowledge -might be able to avoid duplicate items in the first place
• adhere to Python performance tips
  • keep in mind that with PyPy (w/ JIT etc.) things might be a bit different
    • but it usually won't get worse adhering to the tips, thought
• data structures
  • i.e. mainly sets vs. lists vs. tuples vs. dicts etc.
  • identify required properties
    • order?
    • uniqueness?
    • mutability?
  • identify predominant operation
    • iterate
    • add/append
    • remove
    • member checks
    • ...
  • choose data structure by performance measurements of predominant operations
    • e.g. w/ module timeit