library(googlesheets4)
library(skimr)
library(tidyverse)
library(knitr)
library(textTinyR)
library(textstem)
library(tidytext)
library(tm)
library(wordcloud)
library(stringr)
library(patchwork)

sheet_key <- "1fjcQMWOd-nlz3OUZ3f3MdCYAkJXJWN-2fsVSDu0bB_U" 
# sheet_key is a static reference to the sheet

sheet <- gs4_get(sheet_key)

# Get Sheet 1 - regular episodes
datareg <- read_sheet(sheet, sheet = 1) %>%
  mutate(type = 'Regular') 

# Get Sheet 2 - bonus episodes
databonus <- read_sheet(sheet, sheet = 2) %>%
  mutate(type = 'Bonus') %>%
  rename(`...1` = `Episode Number`,
         `...5` = `Brandi's Case`) # renaming for consistency to allow merge of datasets

# Combine into one dataset - 'stacks' both on top of each other
data <- bind_rows(datareg, databonus) 

data_clean <- data %>%
  select(-`Inside Joke Origin`, 
         -`Discord Questions Answered`, 
         -`Interesting Facts/Anecdotes`, 
         -`Release Date`) %>%
  rename(episode_num = `...1`,
         title = `Episode Title`,
         summary = `Episode summary/Show notes`,
         b_case = `...5`,
         b_disc = `Brandi's Disclaimers`,
         k_case = `Kristin's Case`,
         k_disc = `Kristin's Disclaimers`,
         after_dark = `After Dark?`,
         fun_facts = `Tim Pound's Fun Facts`,
         not_fun_facts = `Kristin's Not-So-Fun Fun Facts`,
         guests = Guests,
         guest_case = `Guest's Case`,
         songs = `Songs Sung`,
         k_source = `Kristin's Sources`,
         b_source = `Brandi's Sources`,
         g_source = `Guest's Sources`) %>%
  mutate(episode_num = as.numeric(sub("#", "", episode_num))) %>%
  filter(!is.na(title))

skim(data_clean)

b_disclaimers <- data_clean %>%
  select(episode_num, title, b_case, b_disc, type) %>%
  rename(disclaimer = b_disc,
         case = b_case) %>%
  mutate(name = 'Brandi') %>%
  filter(!is.na(disclaimer))

k_disclaimers <- data_clean %>%
  select(episode_num, title, k_case, k_disc, type) %>%
  rename(disclaimer = k_disc,
         case = k_case) %>%
  mutate(name = "Kristin") %>%
  filter(!is.na(disclaimer))

disclaimers <- bind_rows(b_disclaimers, k_disclaimers) %>%
  separate_rows(disclaimer, sep = " and ") %>%
  arrange(episode_num) %>%
  mutate(disclaimer = case_when(
    disclaimer == 'old Timey Disclaimer' ~ 'Old Timey Disclaimer',
    disclaimer == 'Old timey case, but no proper disclaimer' ~ 'Old Timey Disclaimer',
    disclaimer == 'International Case Disclaimer' ~ 'International Disclaimer',
    TRUE ~ disclaimer
  ))

kable(disclaimers, col.names = c("Ep Number", "Title", "Case", "Disclaimers", "Episode Type", "Name"))

summary_disc <- disclaimers %>%
  count(name, disclaimer) %>%
  rename(count = n) 

# Set custom colours
colours <- RColorBrewer::brewer.pal(5, "Accent")
disc_colours <- c(colours[1], colours[5])


# Plot
ggplot(summary_disc, aes(x = disclaimer, y = count, fill = name)) +
  geom_bar(stat = "identity") +
  scale_fill_manual(values = disc_colours) +
  labs(title = "Disclaimer Counts by Host",
       y = "Count",
       fill = "Name") +
  theme_minimal() + 
  theme(legend.position = "right", 
        legend.title = element_blank(), 
        axis.text.x = element_text(angle = 45, hjust = 1), 
        axis.title.x = element_blank(),
        axis.title.y = element_blank())

remove <- "WARNING: The audio in this episode is rough\\. What can we say\\? We were young\\(ish\\), dumb, and thought we’d save a little money by sharing one microphone\\. Yeah\\. The audio quality improves drastically after episode 9\\."

word_df <- data_clean %>%
  mutate(
    summary = str_replace_all(summary, remove, ""),
    summary = str_replace_all(summary, "(?<!\\d)-(?!\\d)", " ") %>%
    str_replace_all("[^A-Za-z\\s]", "")
  ) %>%
  unnest_tokens(word, summary) %>%
  mutate(word = lemmatize_words(word)) %>%
  anti_join(stop_words) %>%
  filter(!is.na(word)) %>%
  filter(!word %in% c("kristin", "brandi"))

word_counts <- word_df %>%
  group_by(word) %>%
  summarise(count = n()) %>%
  arrange(desc(count))

top_words <- word_counts %>%
  top_n(n=20, wt=count)

ggplot(top_words, aes(x = count, y = fct_reorder(word, count))) +
  geom_bar(stat = "identity", fill = disc_colours[1], alpha = 0.8) +
  labs(title = "Top Words Used in Episode Summary", x = "Count", y = "") +
  theme_minimal() +
  theme(axis.title.y = element_blank())

wordcloud(
  words = word_df$word,
  scale = c(3, 0.5),
  min.freq = 15, 
  random.order = FALSE, 
  colors = brewer.pal(8, "Dark2")
)

guests <- data_clean %>%
  select(type, episode_num, title, guests, guest_case) %>%
  filter(!is.na(guests)) %>%
  mutate(guest = case_when(
    grepl("norm|norman", guests, ignore.case = TRUE) ~ "Norman",
    grepl("dp|daryl", guests, ignore.case = TRUE) ~ "DP",
    TRUE ~ guests
  )) %>%
  mutate(Notes = str_replace_all(guests, c("Norman" = "", "Norm" = "", "Caruso" = "", "DP" = "", "Daryl" = ""))) %>%
  select(-guests) %>%
  mutate(Notes = str_trim(str_replace(Notes, "^[[:punct:][:space:]]+", "")),
         guest_case = coalesce(guest_case, "None")) %>%
  select(type, episode_num, title, guest, guest_case, Notes) %>%
  arrange(type, episode_num) 

kable(guests, col.names = c("Episode Type", "Episode Num", "Title", "Guest", "Case", "Notes"))

guest_summary <- guests %>%
  group_by(guest) %>%
  summarise(count = n())

ggplot(guest_summary, aes(x = guest, y = count, fill = guest)) +
  geom_bar(stat = "identity") +
  labs(title="Visitor Counts",
       y="Count") +
  theme_minimal() + 
  theme(axis.title.x = element_blank(),
        legend.position = "none")

songs <- data_clean %>%
  select(type, episode_num, title, songs) %>%
  filter(!is.na(songs))

song_type <- songs %>%
  group_by(type) %>%
  summarise(count = n()) %>%
  mutate(percentage = count / sum(count) * 100)

g1 <- ggplot(song_type, aes(x = "", y = count, fill = type)) +
  geom_bar(width = 1, stat = "identity") +
  geom_text(aes(label = paste0(round(percentage), "%")), 
            position = position_stack(vjust = 0.5)) +
  coord_polar(theta = "y") +
  theme_minimal() +
  theme(axis.line = element_blank(),
        axis.text = element_blank(),
        axis.ticks = element_blank(),
        panel.grid = element_blank(),
        plot.margin = unit(rep(0, 4), "cm"),
        plot.background = element_blank()) +  # Remove plot background
  labs(title = "Episodes with Songs",
       fill = "Episode Type",
       y = "Count") +
  theme_void() +
  scale_fill_manual(values = disc_colours)

# Compare to overall episodes
overall_count <- data_clean %>%
  group_by(type) %>%
  summarise(count = n()) %>%
  mutate(percentage = count / sum(count) * 100)

g2 <- ggplot(overall_count, aes(x = "", y = count, fill = type)) +
  geom_bar(width = 1, stat = "identity") +
  geom_text(aes(label = paste0(round(percentage), "%")), 
            position = position_stack(vjust = 0.5)) +
  coord_polar(theta = "y") +
  theme_minimal() +
  theme(axis.line = element_blank(),    # Remove the axis line
        axis.text = element_blank(),    # Remove axis text
        axis.ticks = element_blank(),   # Remove axis ticks
        panel.grid = element_blank(),   # Remove grid lines
        plot.margin = unit(rep(0, 4), "cm"),
        plot.background = element_blank()) +  # Remove plot background
  labs(title = "Episodes Overall",
       fill = "Episode Type",
       y = "Count") +
  theme_void() +
  scale_fill_manual(values = disc_colours)

(g1 | g2) + plot_layout(guides = "collect")

tim_pound <- data_clean %>%
  select(episode_num, title, fun_facts) %>%
  filter(!is.na(fun_facts)) %>%
  arrange(episode_num)

kable(tim_pound, col.names = c("Episode Num", "Episde", "Tim Pound's Fun Facts"))

not_fun_facts <- data_clean %>%
  select(episode_num, title, not_fun_facts) %>%
  filter(!is.na(not_fun_facts)) %>%
  arrange(episode_num)

kable(not_fun_facts, col.names = c("Episode Num", "Episode", "Kristin's Not-So-Fun Fun Facts"))