Facebook w R

library(Rfacebook)
library(tidyverse)
library(lubridate)
library(glue)

fb_oauth <- fbOAuth(app_id = "xxxxxxxxxxx",
                    app_secret = "yyyyyyyyyyyyyyyy")

save(fb_oauth, file = "fb_oauth.rda")

load("fb_oauth.rda")

getFPPosts <- function(page_name, n_posts = 50) {
  # pobieramy posty z fanpage'a
  fb_page <- getPage(page = page_name, token = fb_oauth, n = n_posts)

  # zostawiamy sobie tylko potrzebne informacje
  fb_page <- fb_page %>%
    select(id, created_time, message, likes_count, comments_count) %>%
    mutate(created_time = ymd_hms(created_time))

  return(fb_page)
}

getPostComments <- function(post_id, post_coments = 100) {
  
  # pobieramy komentarze z wybranego postu, max 100 sztuk
  fb_post <- getPost(post_id, token = fb_oauth, n = max(post_coments, 1000), comments = TRUE)

  # tabela z komciami - tylko potrzebne dane
  comments <- fb_post$comments %>%
    select(created_time, message, likes_count) %>%
    mutate(created_time = ymd_hms(created_time)) %>%
    arrange(created_time)

  return(comments)
}

page_name <- "daneanalizy"

fb_page_table <- getFPPosts(page_name, 1000)

fb_page_table %>%
  mutate(h = hour(created_time),
         d = wday(created_time, week_start = 1, label = TRUE, locale = "pl_PL.UTF8")) %>%
  count(d, h) %>%
  ggplot() +
  geom_tile(aes(d, h, fill = n), color = "gray60") +
  scale_y_reverse() +
  scale_fill_distiller(palette = "Reds", direction = 1) +
  labs(title = "Kiedy publikowane są posty na fanpage'u Dane i Analizy?",
       x = "", y = "Godzina")

fb_page_table %>%
    mutate(date = floor_date(created_time, unit = "months")) %>%
    group_by(date) %>%
    summarise(n = n(),
              l = sum(likes_count),
              c = sum(comments_count)) %>%
    ungroup() %>%
    ggplot() +
    geom_line(aes(date, n), color = "blue") + # posts
    geom_line(aes(date, l), color = "red") +  # likes
    geom_line(aes(date, c), color = "green") + # comments
    annotate(geom = "label",
             x = min(fb_page_table$created_time), y = 400,
             label = "Liczba like'ów",
             fill = "red", color = "black", hjust = 0) +
    annotate(geom = "label",
             x = min(fb_page_table$created_time), y = 350,
             label = "Liczba postów",
             fill = "blue", color = "white", hjust = 0) +
    annotate(geom = "label",
             x = min(fb_page_table$created_time), y = 300, 
             label = "Liczba komentarzy",
             fill = "green", color = "black", hjust = 0) +
    labs(title = "Liczba postów, like'ów i komentarzy na fanpage'u Dane i Analizy",
         x = "", y = "")

fb_page_table %>%
    mutate(date = floor_date(created_time, unit = "weeks")) %>%
    group_by(date) %>%
    summarise(n = n(),
              l = sum(likes_count),
              c = sum(comments_count)) %>%
    ungroup() %>%
    ggplot() +
    geom_smooth(aes(date, l/n), color = "blue") + # likes/post
    geom_smooth(aes(date, c/n), color = "green") + # comments/post
    annotate(geom = "label",
             x = min(fb_page_table$created_time), y = 12,
             label = "Liczba like'ów/post",
             fill = "blue", color = "white", hjust = 0) +
    annotate(geom = "label",
             x = min(fb_page_table$created_time), y = 10, 
             label = "Liczba komentarzy/post",
             fill = "green", color = "black", hjust = 0) +
    labs(title = "Liczba like'ów i komentarzy na post - na fanpage'u Dane i Analizy",
         x = "", y = "")

# najbradziej komentowany post
post_id <- fb_page_table %>% filter(comments_count == max(comments_count)) %>% pull(id) %>% .[[1]]

# weźmy wszystko co o nim powie Facebook:
post_details <- getPost(post_id, token = fb_oauth)

post_details$post %>% kable() %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"))

post_details$comments %>% kable() %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"))

comments_table <- getPostComments(post_id)

comments_table %>%
  kable() %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"))

comments <- getPostComments("180340042008871_2128002970575892")

library(tidytext)

ludzie <- comments %>%
  select(message) %>%
  mutate(message = str_replace_all(message, "\\.", " ")) %>%
  unnest_tokens(word, message, token = "words") %>%
  mutate(word = case_when(
    # czasem ludzie wpisują jedno lub drugie - ujednolicamy
    word == "jackowska" ~ "kora",
    word == "kazik" ~ "staszewski",
    word == "muniek" ~ "staszczyk",
    TRUE ~ word)) %>%
  count(word, sort = T) %>%
  filter(nchar(word) > 2)

ludzie %>%
  as.data.frame() %>%
  head(60) %>%
  kable() %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"))

stop_w <- c("krystyna", "anna", "janusz", "jerzy", "zbigniew",
            "hanka", "irena", "jan",  "czesław",  "nie", "artur",
            "ale", "beata", "danuta", "eugeniusz", "krzysztof", "franciszek",
            "marek", "piotr", "pola", "tomasz", "andrzej", "wojciech",
            "joanna", "agnieszka", "daniel", "roman", "ewa", "grzegorz", "gustaw", "katarzyna",
            "się", "maryla", "jeszcze", "aktorzy", "kalina",
            "panowie","maria", "mieczysław", "jest", "adolf", "edyta",
            "urszula", "jacek", "kabaret", "tadeusz", "tylko",  "może", "pan", "pani",
            "tak", "agata", "violetta")

ludzie <- filter(ludzie, !word %in% stop_w)

# ile razy występuje 50 najpopularniejsze słowo?
n_limit <- ludzie %>% arrange(desc(n)) %>% pull(n) %>% .[[50]]

# zostawmy tylko te, które wystąpują więcej razy
ludzie <- ludzie %>% filter(n >= n_limit)

# ile mamy wszystkich komentarzy?
n_comm <- nrow(comments)

# rysujemy wykres słupkowy z wynikiem ankiety
ludzie %>%
  arrange(n, word) %>%
  mutate(word = fct_inorder(str_to_title(word))) %>%
  ggplot() +
  geom_col(aes(word, n), color = "gray40", fill = "lightgreen") +
  coord_flip() +
  labs(title = "Kto Waszym zdaniem powinien znaleźć się w pierwszej dwudziestce\ngwiazd polskiej rozrywki na 100-lecie niepodległości Polski?",
       subtitle = glue("Zestawienie na podstawie {n_comm} komentarzy pod postem Karoliny Korwin-Piotrowskiej"),
       x = "", y = "Ile razy nazwisko padło w komentarzu?")