parse transaction details

library(data.table)
library(pipeR)
library(stringr)
library(iterators)
library(foreach)
library(doParallel)

# generate data
set.seed(100) # for reproducing
numCustomer <- 1e4L
numPurchase <- sample.int(10L, numCustomer, TRUE)

out <- data.table(id = 1L:numCustomer, payment = sample(LETTERS[1:3], numCustomer, TRUE),
                  readTime = Sys.time() - sample.int(86400*10, numCustomer, TRUE),
                  items = lapply(numPurchase, function(x){
                    data.table(PN = sample(letters, x), price = abs(rnorm(x, 100, 20)),
                               number = sample.int(100, x))
                  }))

if (file.exists("transRecord.txt"))
  unlink("transRecord.txt")

st <- proc.time()
fileConn <- file("transRecord.txt", "w+")
foreach(it = iter(out, by = "row")) %do%
{
  timeStr <- format(it$readTime, "%Y%m%d%H%M%S")
  cat(sprintf("customerid: %i\npayment: %s\nReadTime: %s\n",
                it$id, it$payment, timeStr), file = fileConn, append = TRUE)
  foreach(it2 = iter(it$items, by = "row")) %do%
  {
    cat(sprintf("%sPruductNumber: %s: NT%010.4fNumber: %i\n",
                  timeStr, it2$PN, it2$price, it2$number), file = fileConn, append = TRUE)
  }
} %>>% invisible
close(fileConn)
proc.time() - st
#  user  system elapsed
# 39.50    0.14   40.11
file.size("transRecord.txt") / 2^20 # 3.508 MB

# start parsing
cl <- makeCluster(detectCores() - 1L)
registerDoParallel(cl)

lines <- readLines("transRecord.txt")
basicInfoColNames <- c("customerid", "payment", "ReadTime")
transInfoColNames <- c("transTime", "PruductNumber", "price", "number")

st <- proc.time()
it <- isplit(lines, cumsum(str_detect(lines, "^customerid")))
resDT <- foreach(it = it, .combine = rbind, .multicombine = TRUE,
        .packages = c("data.table", "stringr", "pipeR")) %dopar%
{
  idx <- which(str_detect(it$value, "^ReadTime"))
  part1 <- str_match(it$value[1:idx], "(customerid|payment|ReadTime): (.*)")
  part2 <- str_match(it$value[(idx + 1):length(it$value)],
                     "(\\d{14})PruductNumber: ([a-zA-Z0-9]+): NT(\\d{5}\\.\\d{4})Number: (\\d+)")
  outDT <- data.table(t(part1[match(part1[ , 2], basicInfoColNames), 3, drop = FALSE])) %>>%
    setnames(basicInfoColNames)
  cbind(outDT, data.table(part2[ , 2:5, drop = FALSE]) %>>% setnames(transInfoColNames))
}
proc.time() - st
# user  system elapsed
# 9.33    3.14   16.86

stopCluster(cl)

head(resDT, 10)
#     customerid payment       ReadTime      transTime PruductNumber      price number
#  1:          1       B 20170410235836 20170410235836             g 00087.0582     84
#  2:          1       B 20170410235836 20170410235836             z 00109.8510     62
#  3:          1       B 20170410235836 20170410235836             a 00113.2259     13
#  4:          1       B 20170410235836 20170410235836             p 00124.6366     44
#  5:          2       A 20170416151637 20170416151637             a 00127.0563     49
#  6:          2       A 20170416151637 20170416151637             e 00102.5638     40
#  7:          2       A 20170416151637 20170416151637             y 00117.9419      3
#  8:          3       B 20170415064513 20170415064513             p 00053.0424     19
#  9:          3       B 20170415064513 20170415064513             h 00059.5930     31
# 10:          3       B 20170415064513 20170415064513             s 00087.3829     50