Oscar Best Film Nomination Wide Release Month (2000-2020)

knitr::kable(last20, align='c')

library(rvest)
library(jsonlite)

url <- "https://en.wikipedia.org/wiki/Academy_Award_for_Best_Picture"

htmlBase <- url %>% read_html() 
tables <- htmlBase %>% html_nodeas("table")

last20 <- NULL
reqYears <- sprintf('%s', seq(2000,2020))
colNames <- c("Year", "Film", "Producer(s)")
bgWinner <- "background:#FAEB86"
for (k in 1:length(tables)) {
  tmp <- html_table(tables[[k]], fill=TRUE)
  if (all(names(tmp) %in% colNames)) {
    tmp$Year <- gsub("\\s*\\([^\\)]+\\)","", as.character(tmp$Year))
    if (any(tmp$Year %in% reqYears)) {
      # Set up winners
      tmp$winner <- FALSE
      winRes <- tables[[k]] %>% html_nodes("tr") %>% html_attr("style")
      tmp$winner[winRes[2:length(winRes)] == bgWinner] <- T
      # Remove NAs
      tmp <- tmp[!is.na(tmp$Film),]
      
      # get wiki Links
      tmp$link <- NA
      links <- tables[[k]] %>%  html_nodes("a") %>% html_attr("href")
      for (j in 1:nrow(tmp)) {
        idx <- grep(gsub(" ", "_", tmp$Film[j]), links)
        # try url encode
        if (length(idx) == 0) {
          idx <- grep(URLencode(gsub(" ", "_", tmp$Film[j]), reserved=T), links)
        }
        if (length(idx) == 1) {
          tmp$link[j] <- links[idx]
        } 
      }
      # combine
      last20 <- rbind(last20, tmp)
    }
  }
}

last20$Film[is.na(last20$link)]

# Update these names
last20$Film[last20$Film == "Precious: Based on the Novel 'Push' by Sapphire"] <- "Precious"
last20$Film[last20$Film == "Birdman or (The Unexpected Virtue of Ignorance)"] <- "Birdman"

# Search by OMDB list to get release date and imdb id for it
last20$imdbId <- NA
last20$releaseDate <- NA
last20$rated <- NA
for (k in 47:nrow(last20)) {
  tmp <- fromJSON(paste0('http://www.omdbapi.com/?apikey=',Sys.getenv('API_OMDB_KEY'), 
                         '&t=', gsub(" ", "+",gsub("[,|:]","",last20$Film[k])), 
                         '&y=', if(last20$Film[k] == "The Hurt Locker") { "2008" } else { last20$Year[k] }), flatten=T)
  if (!("error" %in% tolower(names(tmp)))) {
    last20$imdbId[k] <- tmp$imdbID
    last20$releaseDate[k] <- tmp$Released
    last20$rated[k] <- tmp$Rated
  }
  Sys.sleep(0.1)
}

# Convert release date to date function
last20$releaseDate <- as.Date(last20$releaseDate, "%d %B %Y")
# Get month
# last20$releaseMonth <- as.numeric(format(last20$releaseDate, "%m"))
last20$releaseMonth <- format(last20$releaseDate, "%m")


library(ggplot2)
# Missing april, need to add it in as a limit
monLimit <- c(paste0("0", 1:9),10:12)
ggplot(last20, aes(releaseMonth)) + 
  geom_bar() + 
  xlab("Month") + 
  ylab("Number of Oscar Best Picture Nominations (2000-2020") + 
  ggtitle("Month vs Number of Oscar Best Picture Nominations (2000-2020)") +
  scale_x_discrete(labels=month.abb, limits=monLimit)

last20Winners <- last20[last20$winner,]

ggplot(last20Winners, aes(releaseMonth)) + 
  geom_bar() + 
  xlab("Month") + 
  ylab("Number of Oscar Best Picture Winners (2000-2020") + 
  ggtitle("Month vs Number of Oscar Best Picture Winners (2000-2020)") +
  scale_x_discrete(labels=month.abb, limits=monLimit)