missing(float) is returning true for float_vector_ends as well

[MauricioCarneiro]

it should differentiate between the two.

[MauricioCarneiro]

this should be addressed following #120

[droazen]

See the following test program I wrote:

https://gist.github.com/droazen/7eabfc2cd9054ba9886a

The output of this program is:

missing_value:
val is a NaN
val converted to int is missing

end_of_vector_value:
val is a NaN
val converted to int is vector end

What this shows is that:

• signaling NaNs can be safely passed around by value without getting converted to quiet NaN
• as Karthik pointed out, direct == comparisons won't work

• we can still distinguish between the missing value and end of vector value by converting to int using the union trick in reverse and then comparing using ==. Eg.,

  uint32_t bcf_float_missing = 0x7F800001; uint32_t bcf_float_vector_end = 0x7F800002; union { uint32_t i; float f; } u; u.f = val; if ( u.i == bcf_float_missing ) { ... else if ( u.i == bcf_float_vector_end ) { ...

I will do a pull request in a bit to implement this.

[kgururaj]

You can use functions bcf_float_is_missing and bcf_float_is_vector_end from htslib (htslib/vcf.h)