extract from Whatsapp history to data.table

read_whatsapp.R

library(data.table)
whatsapp_raw = readLines('~/Downloads/WhatsApp Chat with PhDelphia.txt')

#have to deal with multi-line messages :\
idx = grepl('^1?[0-9]/', whatsapp_raw)

idxrle = rle(idx)
bdpts = cumsum(idxrle$lengths)
for (ii in seq_len(length(idxrle$values))) {
  if (!idxrle$values[ii]) {
    whatsapp_raw[bdpts[ii - 1L]] = 
      paste(whatsapp_raw[bdpts[ii - 1L]:bdpts[ii]], collapse = '\n')
  }
}

whatsapp_raw = whatsapp_raw[idx]

whatsapp = setDT(tstrsplit(whatsapp_raw, '(?<=[AP]M) - ', perl = TRUE))
whatsapp[ , time := 
            as.POSIXct(V1, tz = 'EDT', format = '%m/%d/%y, %I:%M %p')]
whatsapp[ , V1 := NULL]
whatsapp[ , is_text := grepl(':', V2, fixed = TRUE)]

alerts = copy(whatsapp[(!is_text)])
alerts[ , is_text := NULL]
setnames(alerts, 'V2', 'alert_text')
whatsapp = whatsapp[(is_text)]
whatsapp[ , is_text := NULL]

whatsapp[ , c('name', 'message') := 
            .(gsub('([^:]+):.*', '\\1', V2),
              gsub('[^:]+:(.*)', '\\1', V2))]
whatsapp[ , V2 := NULL]

#fun
whatsapp[ , has_media := grepl('<Media omitted>', message, fixed = TRUE)]
whatsapp[ , has_emoji := grepl('[\U{1F300}-\U{1F6FF}]', message)]
whatsapp[ , emoji_only := !nzchar(gsub('[\U{1F300}-\U{1F6FF} ]', '', message))]
whatsapp[ , has_url := grepl('http|[.](edu|org|com)', message)]

#sample plot: messages by day, and a trend line with
#  a 60-day moving average (easily adjustable window)
whatsapp[ , .N, keyby = .(t = as.Date(time))
          ][CJ(seq.Date(min(t), max(t), by = 'day')), 
            on = c(t = 'V1')
            ][is.na(N), N := 0
              ][ , {
                plot(t, N, main = 'Messages by Day\nWith 60-day Windowed Average',
                     xlab = '', ylab = '#', cex = .6, xaxt = 'n')
                dts = which(mday(t) == 1L)
                axis(side = 1L, at = t[dts], las = 2L,
                     labels = format(t[dts], '%b-%y'))
                mavg = rowMeans(sapply(-30:30, function(n) {
                  if (n < 0) shift(N, abs(n))
                  else { if (n == 0) N else shift(N, n, type = 'lead') }}),
                  na.rm = TRUE)
                lines(t, mavg, lwd = 3L, col = 'red')}]
                                       
#sample plot: Emoji per message
#  and most frequent emoji
whatsapp[ , {
  emoji = na.omit(unlist(lapply(strsplit(message, ''), function(x) 
    x[grepl('[\U{1F300}-\U{1F6FF}]', x)])))
  #U1F4FB is a modifier (white skin)
  #  the rest just wouldn't display on the plot
  modif = c('\U{1F389}', '\U{1F3FB}', '\U{1F44F}', '\U{1F37B}',
            '\U{1F483}', '\U{1F3B1}', '\U{1F44D}', '\U{1F4A3}',
            '\U{1F4A9}', '\U{1F37E}', '\U{1F38A}')
  omit_modif = emoji[!emoji %in% modif]
  .(freq = names(table2(omit_modif, ord = 'dec'))[1L],
    avg = length(omit_modif)/.N)}, by = name
  ][avg>0][order(-avg), {
    par(mar = c(9.1, 4.1, 4.1, 2.1))
    x = barplot(avg, col = 'yellow', names.arg = name, las = 2,
                main = 'Emoji per Message\nAnd Most Frequent Emoji', cex.names = .8,
                ylim = c(0, 1.2*max(avg)))
    text(x, avg, freq, pos = 3)
    }]

Nice! You should make this a small package!