Faster gmt2mat

[TylerSagendorf]

@ivokwee @ESCRI11 Hello, I've written a faster version of gmt2mat() if you are interested in using it.

Changes

• Improve speed and memory usage of gmt2mat() by a factor of ~12.
• Warn when use.multicore = TRUE in gmt2mat() and set the default to FALSE.
• Replace parallel import with collapse.

Benchmarks

# Simulate gene sets
n_genes <- 1e4L
genes <- paste0("gene", seq_len(n_genes))

n_sets <- 5e4L
min_size <- 10L
max_size <- 1000L

set_sizes <- rep(min_size:max_size, length.out = n_sets)

gene_sets <- lapply(seq_len(n_sets), function(i) {
  set.seed(i)
  sample(x = genes, size = set_sizes[i])
})
names(gene_sets) <- paste0("set", seq_len(n_sets))

# gmt2mat_old contains the original gmt2mat code (not shown)
bench::mark(
  "old" = gmt2mat_old(gene_sets, use.multicore = FALSE),
  "new" = gmt2mat(gene_sets, use.multicore = FALSE),
  iterations = 10L,
  filter_gc = FALSE
)

  expression      min   median `itr/sec` mem_alloc `gc/sec` n_itr  n_gc total_time
  <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl> <int> <dbl>   <bch:tm>
1 old           17.4s    19.8s    0.0499   19.88GB    0.753    10   151      3.34m
2 new           1.43s     1.6s    0.603     1.69GB    0.784    10    13     16.59s