library('tidyverse')
library('lubridate')

read_weechatlog <- function(file_name, parse_dates=TRUE, messages_only=TRUE,
                            sanitize_names=TRUE) {
  pattern <- "(\\d{4}-\\d{2}-\\d{2} \\d{2}:\\d{2}:\\d{2})\t(.+)\t(.+)"

  lines <- readLines(file_name) %>%
    str_match_all(pattern)

  # using do.call(...) is WAY faster than calling
  # rbind() for some reason...?
  data <- do.call("rbind", lines) %>%
    as.data.frame(stringsAsFactors=FALSE) %>%
    transmute(date = V2, name = V3, message = V4)

  if (parse_dates) {
    date_format = "%Y-%m-%d %H:%M:%S"

    data$date = parse_datetime(data$date, date_format)
  }

  if (messages_only) {
    weechat_operators = c("--", "<--", "-->", " *")

    data <- data %>%
      filter(!name %in% weechat_operators)
  }

  if (sanitize_names) {
    irc_usermodes = "[~&@%]"

    data <- data %>%
      mutate(name = str_replace_all(name, irc_usermodes, ""))
  }

  return(data)
}

log <- read_weechatlog("#trollhour.weechatlog")

selected_day = as.POSIXct("2018-7-20")

data <- log %>%
  filter(year(date) == 2018 & month(date) == 7 & day(date) == 20) %>%
  mutate(date = hour(date))

ggplot(data, aes(x=as.factor(date), group=name, color=name)) +
  labs(x="Hour", y="Message count",
       title=paste("Activity in #trollhour on", selected_day)) +
  geom_smooth(stat="count", show.legend=FALSE) +
  geom_point(stat="count")