rm(list=ls(all=TRUE))
library(pipeR)       # %>>%
library(data.table)  # fread
library(Matrix)      # sparseMatrix

top_p <- 10
dt <- 600
colNames <- c("group", "second", "id")
txtFiles <- list.files(".", pattern="\\.txt")
inputDT <- fread(txtFiles[1]) %>>% setnames(colNames) %>>%
  `[`(, orderSecond := frankv(second, ties.method = "first"), by = group)

st = proc.time()
maxLen <- inputDT[ , list(len = length(second)), by = group] %>>% (max(.$len))
outDT <- lapply(1:(maxLen-1), function(j){
  inputDT[ , dataID := id[j], by = group] %>>%
    `[`( , secondDiff := second - second[j], by = group) %>>%
    `[`(orderSecond - j > 0) %>>%
    `[`(!is.na(secondDiff) & secondDiff < dt)
}) %>>% rbindlist
m11 <- outDT %>>% `[`( , .(dataID, id))
m21 <- outDT %>>% `[`(secondDiff == 0 , .(dataID, id))
time_method1 <- proc.time() - st


st = proc.time()
u = unique(inputDT$group)
m1 = array(dim=c(0,4))
m2 = array(dim=c(0,4))
for (i in 1:length(u)){
  y = inputDT[inputDT$group==u[i],]
  for(j in 1:(nrow(y)-1)){
    y$d = c(rep(NA,j), tail(y$second,-j)-y$second[j])
    k1 = which(y$d < dt)
    k2 = which(y$d == 0)
    if(length(k1)>0) {
      m1 = rbind(m1, cbind(u[i], j, id=y$id[j], y$id[k1]))
    }
    if(length(k2)>0) {
      m2 = rbind(m2, cbind(u[i], j, y$id[k2], pid=y$id[j]))
    }
  }
}
time_method2 <- proc.time() - st