white TO talyn

Rewrote the secondary structure handling in Alignment. Also changed first parameter in writeFasta to be a PrintStream not an OutputStream

Added printing of pairwise alignment when checkSequnces fail for target sequence

GraphAverager: got rid of redundant code in countVotes. Changed signature of one of the constructors: now requires target sequence.
averageGraph: adapted accordingly
Alignment: new method removeSequence (untested), fixed some docs

added a comment line

Fixed bug: now PhiPsiAverager behaves like GraphAverager:  consensus is on templates agreeing out of potential templates (non-gaps) instead of out of total number of templates.

Made voteThreshold and angleInterval global.
Improved comments

Changed the type used for the set of template graphs/pdbs from arrays to TemplateList. Now the loading of pdb data and getting the graphs is done by TemplateList.

Fixed bugs in averageGraph:
- output of benchmark statistics and pairwise comparison was overlapping
- benchmark mode was broken after introduction of align mode (when no sequence file given we were always creating dummy sequence in both align or benchmark modes)

Improved:
- better handling of modes through mode variable that takes 1 of 3 states
- better input check
- in align mode now calling another GraphAverager constructor instead of creating dummy seq/tag within averageGraph
- changed constructor in GraphAverager: now sequence not a parameter, instead taken from alignment. Check for input moved down to averageGraph
- now allowing alignments that contain other sequences than the target and the templates
- new exception GraphAveragerError thrown in GraphAverager constructor when sequences of alignment and templates don't match

Fixed bug: PhiPsiAverager was not working at all for single template case. 
Fixed bug: in ConsensusInterval changed MARGIN value to 2, for cases where max=min to make sure we pass an Interval bigger than 0 to tinker.

added boolean parameter 'showSequence' to Alignment.writeSecStructMatching()

Important change: now final interval taken for restraints is not anymore of size angleInterval, but simply goes from min to max.

Introduced some more useful debugging output in the main method: phi/psi of target (if one is avalable from a reconstruction) is also printed.

Fixed (hopefuly) the issue of wrapping angles (at -180,180). Tested minimally and seems fine.

Fixed bug: algorithm to find consensus was not working correctly. We weren't taking in account the 2-dimensionality of the problem. Changed the approach completely, new algorithm.
Made the ConsensusInterval class a separate class and added new class ConsensusSquare to represent a phi/psi angle consensus region.

Also adding caspParents member to graph classes, value will be kept from Pdb if present; adding method to GraphAverager to retrieve the most frequently used parents specified in the templates

added new mode of averaging without consensus threshold but specifiying the number of contacts to be predicted and picking starting from highest consensus down

IntervalCandidate not Comparable anymore, finding max is now done with a specific Comparator. Additionally it now extends Interval so we can use the Interval type for the final public output.

Fixed bug: recenterInterval was not working: was not taking min/max correctly.
New method to get the consensus mapped onto a target sequence.

New class PhiPsiAverager to get consensus of phi/psi angles from a TemplateList and an Alignment. Tested with a few examples and seems to work. The wrapping of angles at 180/-180 is not yet taking into account, i.e. if an interval falls in the region just below 180 and just above -180, no consensus will be found.
Pdb: added some checks to methods getPhi/Psi so that it doesn't fail when there's no coordinates.
Changed yet again the design of TemplateList/Template. Now loading of PDB data happens upon call of the loadPDBdata method. Changed dependencies accordingly.

added option to GraphAverager and averageGraph to create dummy sequence with length of alignment to view the overlaid graphs without mapping to a target sequence

Changed behaviour of graphAverager: Now evaluating consensus threshold based only on edges that could really exist (i.e. both ends map to non-gaps) rather than simple count of templates. The old behaviour can be restored (for comparison) by setting legacyMode=true in getAverageGraph() and getConsensusGraph().

Added some more error output

New feature in GraphAverager: method to write the voters together with the edges; new constructor taking a RIGEnsemble
Fixed bug in RIGEnsemble: was not working for cif files.
Added a graph averaging output to main() in RIGEnsemble

Fixed another bug in GraphAverager. Was still not working correctly: was putting edges twice in the final graph. 
Also now handling correctly graph averaging for directed or undirected graphs.

Fixed GraphAverager: wasn't working since we introduced JUNG. 
Also added some better output in checkSequences() when sequence don't match between alignment/graphs. Now we take the 2 sequences calculate a pairwise alignment (using jaligner) and output it. This is useful for picking up unobserves and similars.

Changed Alignment class so that both alignment and sequence indexing are starting at 1 (before alignment indices were starting at 0). Also mapping is now done through arrays not maps.
Changed all other classes using Alignment to accommodate this. 
NOTE: graph averaging hasn't been tested after the change

Copied the aglappe-jung branch into trunk.

added methods to calculate and filter by consensus scores

moved GraphAverager to subpackage, added removeVertex function to RIGraph and AIGraph, added getEdgeFromSerials and containsEdgeIJ functions to ProtStructGraph