implement noLP in sparse algorithm

Member LocARNA::AlignerP::compute_arcmatch_probabilities ()

check: why is that long double? do we need it? should we rather use pf_t?

constructors ensure name uniqueness; if constructing as alignment of alignments, names are made unique if needed.

because this class does not know whether it contains RNA, it would be useful to have a derived class RNAMultipleAlignment. This class could guarantee that its sequences are normalized RNA sequences. Consequently, we could enforce by the type system that RnaEnsemble is generated only from RNAMultipleAlignment etc.

automatically adapt output width for sequence names to longest name

Member LocARNA::MultipleAlignment::set_annotation (const AnnoType::type &annotype, const SequenceAnnotation &annotation)

check that annotation is valid for multiple alignment; throw failure if annotation is not valid

Member LocARNA::RnaData::RnaData (const std::string &filename, double p_bpcut, double max_bps_length_ratio, const PFoldParams &pfoldparams)

consider to allow reading from istream; use istream::seekg(0) to reset stream to beginning (needed for format autodetect.) Is there a problem due to fail on eofbit in C++98?

Class LocARNA::RnaEnsemble

support constraints for in loop probabilities

split up RnaEnsemble into two classes; one with and one without in-loop probabilities (like RnaData and ExtRnaData)

Class LocARNA::RnaStructure

read structure in ct format from stream

Class LocARNA::Scoring

check proper use of unpaired penalty; e.g. is correction needed for arc match (like for lambda)??

Class LocARNA::SparseVector< T >

this code is basically a stripped down version of the SparseMatrix code; likely one could reduce redundancy

Member main (int argc, char **argv)

currently not a command line option of locarna_p