broadinstitute / gamgee / 1180 / 2
93%
master: %

LAST BUILD BRANCH: jt_rbs_vb2
DEFAULT BRANCH: master
Ran
Files 95
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 93.265% (+0.9%) from 92.362%
1180.2

push

travis-ci

droazen 
VariantBuilder

VariantBuilder code review round 2 update

-Addressed round 1 code review comments

-Enable move semantics when setting individual fields
 (allows for zero-copy bulk setting of a field for
 all samples at once)

-As a side effect of the above, setting individual fields
 by sample instead of in bulk is now supported. Uses
 short-value optimization to keep the per-sample values
 as contiguous as possible, but this option is still less
 efficient than setting in bulk.

-Significant refactoring to support the new move semantics
 and per-sample field setting:

    -split VariantBuilderDataField into VariantBuilderSharedField
     and VariantBuilderIndividualField

    -individual fields are now encoded at build time instead of
     at field set time; shared fields are still encoded upfront

    -to allow setting individual fields by nested vectors and
     by sample to compete with setting it by flattened vector,
     had to write a lot of low-level encoding routines to replace
     the former strategy of converting to flattened vector then
     handing off to htslib

Remaining TODOs:

-Fix issues with float missing/end of vector values (these are
 proving very problematic)

-Build-by-merge support (Joel indicated that this would be helpful
 for his use case, after all)

-More tests (now that Coveralls is working again, I can see the
 coverage gaps clearly)

-Doxygen comments / documentation

Fix bug with float missing / end of vector values

Apparently we can't use the htslib bcf_float_missing and bcf_float_vector_end
values directly -- need to copy them into local variables before use.

Fix bug in validation of vector lengths when setting individual fields

Previous validation routine would incorrectly allow nested vectors
whose first dimension was greater than but divisible by the number of
samples. Now validation of vector<vector> is separated from validation
of flattened vectors.

Improve test coverage, and fix issues found by new tests

refactoring for readability and safety

2008 of 2153 relevant lines covered (93.27%)

630.32 hits per line

Source Files on job 1180.2