Test problematic variants

[iskandr]

• added unit test module for "problematic" variants (anything which raises an exception when annotating variants in the wild, e.g. from TCGA MAF files)
• Variant now trims any shared suffix or prefix strings from ref and alt and adjusts the variant position accordingly. The position field has been renamed to start. The original variant fields are preserved as Variant.original_ref, Variant.original_alt, Variant.original_start. For example, if a VCF contains "chr1 10 CAG CATG", then the corresponding Variant object will have ref = "", alt = "T", start = 11, original_ref = "CAG", original_alt = "CATG", original_start = 10.
• Renamed top_priority_transcript_effect to top_priority_effect (since we now have effects like Intergenic which aren't tied to a particular transcript)
• Added ExonLoss as an effect
• Reorganized the effect annotation logic in Variant to deal with variants which span feature boundaries (e.g. deletions which affect both an exon and its neighboring intron).
• renamed VariantCollection.variant_effects to just VariantCollection.effects
• renamed VariantCollection.variant_summary to VariantCollection.summary_string
• added unit tests for misc. VariantCollection methods

[timodonnell]

will this code actually run? will nose call next() automatically on tests that return generators?

[timodonnell]

nvm, I googled, and the answer is yes! good to know http://nose.readthedocs.org/en/latest/writing_tests.html

[timodonnell]

I didn't check all the logic here (really unit tests are the only way), and sounds like much of this will change if you move to the ref_position_before and ref_position_after way of localizing variants, but from looking at what you have so far I don't see any major flaws

[iskandr]

Did anything stand out as particularly in need of more/any unit tests?