Validator should check for impossible zero-length branches

[ihh]

Branches with length zero will give a zero likelihood if there are any indels on the branch (or any mismatches). We can easily check for this situation on the client before posting.

[ihh]

Posting a history with zero-length indel branches currently gives the error "Special numeric value NaN is not allowed"

[ihh]

See #5 for files to reproduce the problem

[ShreyaKakhandiki]

This issue is now resolved.

I've added a validation rule to validator.js that checks for zero-length branches containing indels (I,D, or 'X') in their CIGAR strings. These are now rejected before reaching the scoring step, as such configurations yield invalid likelihoods.

The validation logic was added to the traverseTree() function and throws an error like:

Node at <root>.child[0] has zero-length branch but contains indels in CIGAR: '...'

Example test case

I tested the updated validator by submitting a dummy tree with zero-length branches and the following CIGAR strings. The dummy tree and alignment file used to post can be found in #5. Example string with branch length 0:

11M6D5M1D9M1D18M1D16M9D7M6D2M2D24M4D3M1D3M4D15M1D4M56D4M7D11M1D6M1D1M3D8M10D3M1D1M3D3M4D1M1D4M3D1M2D4M11D4M1D5M2D3M1D9M5D8M3D30M2D3M4D2M6D1M1D4M4D1M3D5M1D12M

This produced the expected validation error:

Invalid CIGAR tree
Node at <root>.child[0] has zero-length branch but contains indels in CIGAR: '...'

These changes are now live in Lambda and included in PR #10.