broadinstitute / gamgee / 1428

Pull #378

travis-ci

VariantBuilderMultiSampleVector: simple class to make it easier to prepare multi-sample data for VariantBuilder

Made the efficient one-dimensional vector option for setting individual fields MUCH easier
to use by providing a VariantBuilderMultiSampleVector class that handles the work of
setting missing values and padding the vector to the maximum field width.

To use, first determine the number of samples and the maximum number of values per sample
for the field, then get a pre-initialized vector from the builder. Eg.,

auto multi_sample_vector = builder.get_integer_multi_sample_vector(num_samples, max_values_per_sample);

This vector will have missing values for all samples, with appropriate padding to the maximum field width.

Then, fill in the values for each non-missing sample by invoking the set_sample_value() and/or
set_sample_values() functions on your multi-sample vector (set_sample_value() is more efficient
than set_sample_values() since it doesn't require a vector construction/destruction for each call).
You don't have to worry about samples with no values, since all samples start out with missing values.

Finally, pass your multi-sample vector to the builder (ideally by move):

builder.set_integer_individual_field(field_index, std::move(multi_sample_vector));

-Works with integer and float individual fields, as well as the GT field
 (string fields are still set by passing in a one-dimensional vector of strings).

-Removed the ability to pass in a raw one-dimensional vector directly for int/float
 fields -- you must use a VariantBuilderMultiSampleVector if you want to work
 with flattened multi-sample data.

-Updated tests as necessary.

Resolves #325

45 of 45 new or added lines in 2 files covered. (100.0%)

2686 of 2913 relevant lines covered (92.21%)

1247.02 hits per line