goss_importer.R

library(tidyverse)

#### clear workspace ####
rm(list=ls())

#### read in the html file and clean up the data ####
msgs <- readLines("~/Documents/gosschat.html")
msgs <- gsub(" ", msgs, replacement="__")
msgs <- gsub("\"", msgs, replacement="##")
msgs <- gsub("\'", msgs, replacement="##")
msgs <- gsub("<div__class=##message_header##>", msgs, replacement="#>>#gosshead")
msgs <- gsub("<div__class=##message##>", msgs, replacement="#>>#newgoss")
msgs <- gsub("<span__class=##user##>", msgs, replacement="#>>#gosser")
msgs <- gsub("<span__class=##meta##>", msgs, replacement="#>>#gossdate")
msgs <- gsub("<ul__class=##meta##>", msgs, replacement="#>>#gossreact")
msgs <- gsub("</div>", msgs, replacement="")
msgs <- gsub("<div>", msgs, replacement="")
msgs <- gsub("</p>", msgs, replacement="")
msgs <- gsub("<p>", msgs, replacement="")
msgs <- gsub("</span>", msgs, replacement="")
msgs <- gsub("<span>", msgs, replacement="")
msgs <- gsub("&#064;", msgs, replacement="@")
msgs <- gsub("&#039;", msgs, replacement="'")
msgs <- gsub("\\s", msgs, replacement="")
msgs <- gsub("CDT", msgs, replacement="CDT#>>#")

msg_split <- str_split(msgs, "#>>#newgoss#>>#gosshead") %>% 
  flatten() %>% 
  map(~str_split(., "#>>#") %>% flatten())

msg_reformat <- transpose(msg_split)[2:4] %>% 
  set_names("gosser", "gossdatetime", "goss") %>% 
  as.tibble()

msg_reformat_clean <- msg_reformat %>% 
  mutate(gosser = gsub("gosser", "", gosser),
         gosser = gsub("__", " ", gosser),
         gossdatetime = gsub("_", " ", gossdatetime),
         gossdatetime = gsub("gossdate", "", gossdatetime),
         goss = gsub("_", " ", goss))