Collect (via Twitter's stream API) tweets from 2020 Nevada Caucus

nevada-caucus-tweets-2020.R

## load rtweet
library(rtweet)

## store bounding box coordinates for state of nevada
nv <- c(-120.00647, 35.00186, -114.03965, 42.00221)

## initialize output vector and Midwestern 'ope' counter
s <- list()
ope <- 0L

## specify stop time as *noon tomorrow* (you can manually override too)
stop_time <- as.POSIXct(format(Sys.Date(), "%Y-%m-%d 12:00:00"))

## keep streaming tweets until noon (UTC) tomorrow
while (Sys.time() < stop_time) {
  ## connect to stream API for one hour at a time
  s[[length(s) + 1L]] <- tryCatch(
    stream_tweets(nv, timeout = 60 * 60),
    error = function(e) data.frame())
  
  ## if no tweets (probably due to error) then add to 'ope' counter
  ## otherwise print running total of collected tweets
  if (NROW(s[[length(s)]]) == 0L) {
    cat("Ope!\n")
    Sys.sleep(1)
    ope <- ope + 1L
  } else {
    cat("Collected ", sum(dapr::vap_int(s, NROW)), " tweets!\n")
  }
  
  ## this is a safety valve; if your stream fails 101 times,
  ## you should probably shut it down and figure out why
  if (ope > 100L) {
    stop("Shut it down. That's too many 'opes'!")
  }
}

## combine into single data frame
s <- do.call("rbind", s)

## save as CSV (and how to read it back in)
save_as_csv(s, "nevada-caucus-tweets-2020.csv")
#s <- rtweet::read_as_csv("nevada-caucus-tweets-2020.csv")

## save as serialized R data (and how to read it back in)
saveRDS(s, "nevada-caucus-tweets-2020.rds")
#s <- readRDS("nevada-caucus-tweets-2020.rds")