Changelog for ADPfusion-0.5.2.0

0.5.2.0

table filling fully inlined in the forward algorithm
experimental PrimBool operations
note: these optimizations are mostly of interest for linear languages, where is rule (or function call) is comparatively expensive
major re-arrangement of modules: import ADP.Fusion.Core for development of novel DP systems. Import ADP.Fusion.Point if you want to write a sequence alignment algorithm

complete re-design of Inside / Outside / Complement handling based on phantom types
very liberal combination of multi-tape grammars
simplified index generation system (both faster, and easier to write new symbol now)

bugfix in Multitape Subword Index calculations (A.F.S.Indices.hs) [this one is quite spurious. I needed quickcheck to find a suitable minimal example where Pseudoknot.hs fails]

initial support for multi-context free grammars
mcfgs allow for interleaved syntactic variables
applications include: natural language modelling and pseudoknotted structures in RNA
the simplest formal language that requires this is: a^i b^j a^i b^j
the GenussFold library gives a simple example grammar

travic-ci integration
forward phase now operates on immutable tables that are internally thawed
resembles the behavior of Data.Vector.Generic.constructN
Empty needs to be bound to input. We require this as certain index structures have no natural notion of and empty index -- unless one provides additional information in the index

simplified boundary checking: sometimes gives performance gain (!) due to one loop variable less
optimized loop variable design following "The HERMIT in the Stream" (Farmer et al, 2014)
somewhat nicer programmer interfaces
automatic filling and freezing of tables
multiple example algorithms (build with -fexamples switch):
- Needleman-Wunsch global alignment
- RNA secondary structure prediction using simple base pair maximization
updated Table code to handle single-dim Subwords in a better way.
simplified backtracking