Input trait

[m4rw3r]

See #44

• Investigate removal of State::Incomplete

  Simplifies all combinators which otherwise had to special-case the incomplete states. Does not seem to actually improve performance in any way despite decreasing the number of possible branches, guess that goes to show how good branch-prediction is in a modern CPU.

  Conclusion: removed State::Incomplete.

• Implement Input for &str

• Implement Input for an Iterator wrapper

  Moved to separate PR Implementation of Input wrapper for Iterator types #49

• Implement Input for Tendril

• Replace the inner State type with Result since it is just a success/fail state and makes for easier conversions and simpler code in general.

• Implement a numbering wrapper for any Input.

• Move input::InputBuf to buffer module and replace the mode property with a boolean flag.

• Update tests to accommodate for Incomplete removal and restructuring of the Input.

  It is probably suitable to investigate if Quickcheck can be used here (Maybe use quickcheck to simplify some tests #33)

• Rework buffer module

  Make sure lifetimes are removed (now part of Input implementation), simplify.

• Update doctests

• Support for optimized skip_while in Input

[m4rw3r]

This will probably need a rework of the buffer module to better fit the new Input traits.

[m4rw3r]

Would be interesting to investigate if the State::Incomplete could be removed in favor of something like attoparsec's continuations (ie. ask the type implementing Input for more data, fail or succeed if no more data could be provided depending on the parser).

[coveralls]

Coverage decreased (-0.9%) to 95.508% when pulling 359e909 on feature/input_trait into 456c5d4 on master.

[coveralls]

Coverage decreased (-0.7%) to 95.686% when pulling 39f5838 on feature/input_trait into 456c5d4 on master.