speed can be improved

[alexyfyf]

Especially in doco counting, maybe the aggregate function can be replaced with something else.

longhaul/R/blessy-createDoCoCount.R

Line 89 in 357a211

aggregated_df <- aggregate(

Use data.table package will be faster, but will add dependency.

Here is some test on a count table with 174k transcripts and 44k DoCo.

dim(merged_df)
# [1] 174278   3853
length(unique(merged_df$DoCo))
# [1] 44099

# use merge
system.time(aggregated_df <- aggregate(
  merged_df[, sample_columns],
  by = list(DoCo = merged_df$DoCo),
  FUN = sum
))
# user  system elapsed 
# 938.484  21.817 950.460 

# use dplyr
system.time(aggregated_df2 <- merged_df %>%
              group_by(DoCo) %>%
              summarise(across(all_of(sample_columns), sum)))
# user  system elapsed 
# 457.294  25.740 482.254 

# use data.table
system.time(dt <- as.data.table(merged_df))
# user  system elapsed 
# 1.238   0.028   1.234 
system.time(aggregated_dt <- dt[, lapply(.SD, sum), by = DoCo, .SDcols = sample_columns])
# user  system elapsed 
# 17.612   0.084   2.258
system.time(aggregated_df3 <- data.frame(aggregated_dt))
# user  system elapsed 
# 0.471   0.000   0.472