Add euclidean option and missing correction to bitwise.dist

[zkamvar]

To aid in the move towards getting genlight objects supported for AMOVA, I've implemented a Euclidean distance option as well as a correction for missing data, which mimics the default distance by dist().

The benefit here is that the speed and amount of memory used are greatly reduced with bitwise.dist() as compared to dist() due to the fact that it doesn't require conversion to an integer matrix:

suppressPackageStartupMessages(library("poppr"))
library("lineprof")
options(width = 100)

# Creating 50 samples with 100k SNPs with ~6% missing data
x <- new("genlight", matrix(sample(c(0:2, NA), 5000000, replace = TRUE, prob = c(1, 1, 1, 0.2)), nrow = 50), ploidy = 2)
x
#>  /// GENLIGHT OBJECT /////////
#> 
#>  // 50 genotypes,  100,000 binary SNPs, size: 2.5 Mb
#>  312468 (6.25 %) missing data
#> 
#>  // Basic content
#>    @gen: list of 50 SNPbin
#>    @ploidy: ploidy of each individual  (range: 2-2)
#> 
#>  // Optional content
#>    @other: a list containing: elements without names

# Memory --------------------------
lineprof(bd <- bitwise.dist(x, euclidean=TRUE, scale_missing=TRUE))
#> Reducing depth to 2 (from 5)
#>    time alloc release dups                             ref                      src
#> 1 0.001 0.113       0   53                 internal.r#2669 fix_uneven_diploid/%>%  
#> 2 0.247 0.028       0   21                         ".Call" .Call                   
#> 3 0.001 0.027       0    3 c("NA.posi", "standardGeneric") NA.posi/standardGeneric 
#> 4 0.001 0.017       0    2               "lazyLoadDBfetch" lazyLoadDBfetch         
#> 5 0.722 0.060       0    1                   bitwise.r#229 missing_correction/.Call
#> 6 0.001 0.000       0    3 c("as.dist", "as.dist.default") as.dist/as.dist.default
lineprof(dd <- dist(x))
#> Reducing depth to 2 (from 58)
#>    time  alloc release dups                                  ref                          src
#> 1 0.223 18.285  13.639 1605 c("as.matrix", "as.matrix.genlight") as.matrix/as.matrix.genlight
#> 2 0.736  4.768   2.392    6                         character(0)

# Speed ---------------------------
system.time(bd <- bitwise.dist(x, euclidean=TRUE, scale_missing=TRUE))
#>    user  system elapsed 
#>   1.534   0.028   0.484
system.time(dd <- dist(x))
#>    user  system elapsed 
#>   2.297   0.082   2.399

Created on 2018-04-01 by the reprex package (v0.2.0).

Things that still need to be done

• change warnings_are_errors back to true in Travis