uni-ba-socialagenda/issuecomp-1-scraping.R

# PREPARATIONS ------------------------------------------------------------

require(jsonlite)
require(stringr)
require(devtools)
require(RTwitterAPI)

setwd("~/Dokumente/Uni/Aktuell/BA-Arbeit/uni-ba-issuecomp")

source("issuecomp-functions.R")


#acc_url <- "http://www.bundestwitter.de/api/politiker"
#acc_df <- fromJSON(acc_url)

acc_df <- read.csv("MdB-twitter.csv")

delrow <- NULL
for(r in 1:nrow(acc_df)) {
  acc <- as.character(acc_df$twitter_acc[r])
  if(!nzchar(acc)) {
    delrow <- c(delrow, r)
  }
}
acc_df <- acc_df[-delrow, ]
rm(delrow, r, acc)
acc_df$row.names <- NULL
row.names(acc_df) <- NULL


# COLLECT ALL TWEETS ------------------------------------------------------

# http://www.joyofdata.de/blog/twitters-rest-api-v1-1-with-r-for-linux-and-windows/
#   --> devtools::install_github("joyofdata/RTwitterAPI")
# https://dev.twitter.com/rest/reference/get/statuses/user_timeline
api_params <- c(
  "oauth_consumer_key"     = readLines("twitter-api-credentials.txt")[2], 
  "oauth_nonce"            = NA,
  "oauth_signature_method" = "HMAC-SHA1",
  "oauth_timestamp"        = NA,
  "oauth_token"            = readLines("twitter-api-credentials.txt")[4],
  "oauth_version"          = "1.0",
  "consumer_secret"        = readLines("twitter-api-credentials.txt")[3],
  "oauth_token_secret"     = readLines("twitter-api-credentials.txt")[5]
)

#api_url2 <- "https://api.twitter.com/1.1/statuses/show.json"
#id2="498492933922754560" # 499533113676931073(\" ), 325320073906622464(\\>), 498492933922754560(\"W)
#query2 <- c(id=id2, trim_user="true", include_entities="false")
#current2 <- twitter_api_call(api_url2, query2, api_params)



api_url <- "https://api.twitter.com/1.1/statuses/user_timeline.json";
max_count <- "200"
keep <- c("created_at", "id_str", "text", "retweet_count")

# tweets_complete: All tweets
# tweets_full: All tweets of current user
# tweets_temp: The current max 200 tweets of current user
tweets_full <- data.frame(user=character(), name=character(), created_at=character(), id_str=character(), text=character(), retweet_count=character())
tweets_complete <- tweets_full

for(a in 1:nrow(acc_df)) {
  user <- as.character(acc_df$twitter_acc[a])
  name <- as.character(acc_df$name[a])
  max_id <- "999999999999999999"
  loop <- 1
  error <- 0
  repeat {
    # Define specific search query
    query <- c(include_rts=1, exclude_replies="true", trim_user="true", include_entities="false", 
               screen_name=user, 
               count=max_count, 
               max_id=max_id);
    
    # At first, work with an temporary tweet-DB
    current <- twitter_api_call(api_url, query, api_params)
    rm(tweets_temp)
    tweets_temp <- fromJSON(correctJSON(current))
    
    ## START ERROR HANDLING ##   
    # Empty API output
    status <- errorEmptyAPI(tweets_temp)
    if(status == 1) { Sys.sleep(3);error <- error + 1;next}
    if(status == 2) {break}
    
    # Contains "error" column
    status <- errorErrorColumn(tweets_temp)
    if(status == 1) { Sys.sleep(3);error <- error + 1;next}
    if(status == 2) {break}
    
    # Check if error code exists
    code <- errorCheckCode(tweets_temp) # 0 if no error
    if(code == 34) {  # page does not exist
      status <- errorCode34()
      if(status == 1) { Sys.sleep(3);error <- error + 1;next}
      if(status == 2) {break}
    }
    if(code == 88) {  # rate limit exceeded
      wait <- errorCode88()
      Sys.sleep(wait)
      next
    }  
    ## END ERROR HANDLING ##
    
    
    # Delete unnecessary columns and add username and real name to dataframe
    tweets_temp <- tweets_temp[keep]
    tweets_temp <- cbind(user=user, name=name, tweets_temp)
    
    # Now sleep 3 second to dodge 300queries/15min limit
    cat("[",a,"/",nrow(acc_df),"] ", sep = "")
    cat("User: ",user," in loop: ",loop,". \n", sep = "")
    Sys.sleep(2)
    
    if(tweets_full$id_str[nrow(tweets_full)] == tweets_temp$id_str[nrow(tweets_temp)] && nrow(tweets_full) > 0) {
      cat("[INFO] Last tweet of temp is last tweet of full. Abort loop and begin with next user.\n")
      break
    }
    
    ## CHECK if we need another loop
    
    # Extract year of last tweet in tweets_temp
    year_last <- as.numeric(str_extract(tweets_temp$created_at[nrow(tweets_temp)], "\\d{4}$"))
    status <- year_last < 2014
    
    # Is last tweet earlier than 2014? So break the loop
    if (status) {
      
      # Is even the first tweet older than 2014?
      year_first <- as.numeric(str_extract(tweets_temp$created_at[1], "\\d{4}$"))
      status <- year_first < 2014
      if(status) {
        cat("[INFO] Timeline enhält keinen einzigen aus 2014\n")
      }
      
      tweets_full <- insertRow(tweets_full, tweets_temp)
      break  # End loop because 2013 is reached
    }
    
    # The last tweet is newer or equal 2014, so we need another loop
    else {
      # Setting max_id to gather next 200 tweets
      max_id <- tweets_temp$id_str[nrow(tweets_temp)]
      loop <- loop + 1  # just for stats
      
      tweets_full <- insertRow(tweets_full, tweets_temp)
    }
  } # /repeat
  
  tweets_complete <- insertRow(tweets_complete, tweets_full)
  tweets_full <- head(tweets_full, -nrow(tweets_full)) # Empty tweets_full
  cat("User:",user,"finished after",loop,"loops. Total Tweets now:",nrow(tweets_complete),"\n")
  write.csv(tweets_complete, "tweets_complete.csv")
  
  # Every tweet from 2014 or newer from user[a] is downloaded. Now next user in for-loop
}
rm(a, code, current, error, loop, max_id, max_count, year_first, year_last, name, query, status, user, wait, tweets_full, tweets_temp)


# CLEAR DATAFRAME ---------------------------------------------------------

save(tweets_complete, file="tweets_complete.RData")

# Remove duplicates
tweets <- tweets_complete[!duplicated(tweets_complete), ]
tweets <- na.omit(tweets)
rm(tweets_complete)

# Format dates in data frame
Sys.setlocale("LC_TIME", "C")

tweets$created_at <- as.POSIXct(tweets$created_at, format = "%a %b %d %H:%M:%S %z %Y")
tweets <- tweets[order(tweets$created_at), ]

# Finally delete every tweet not from 2014 (2013 or 2015)
delrow <- NULL
for(r in 1:nrow(tweets)) {
  if(format(tweets$created_at[r], "%Y") != "2014") {
    delrow <- c(delrow, r)
  }
  curtext <- as.character(tweets$text[r])
  curtext <- str_replace_all(curtext, "$", " ")
  curtext <- str_replace_all(curtext, "http://.+?\\s", "URL ")
  tweets$text[r] <- curtext
}
tweets <- tweets[-delrow, ]
rm(delrow, r)


# Convert dates to omit (unnecessary) time
tweets$created_at <- format(tweets$created_at, "%Y-%m-%d")

save(tweets, file="tweets_untagged.RData")