# Setup ------------------------------------------------------------------------

test_input <- c(
  "mxmxvkd kfcds sqjhc nhms (contains dairy, fish)",
  "trh fvjkl sbzzf mxmxvkd (contains dairy)",
  "sqjhc fvjkl (contains soy)",
  "sqjhc mxmxvkd sbzzf (contains fish)"
)
real_input <- readLines('input.txt')

library(dplyr)
library(purrr)


# Functions --------------------------------------------------------------------

# Helper function, given a line number `line_no` and a list where the first
# element is a space-separated vector of ingredients and the second element
# is a space-separated vector of allergens `split_line`, returns a data frame
# with one row for each combination of `line_no` (as 'recipe'), ingredient, 
# and allergen.
line_to_df <- function(line_no, split_line) {
  parts <- strsplit(split_line, ' ')
  expand.grid(
    recipe = line_no, 
    ingredient = parts[[1]], 
    allergen = parts[[2]], 
    stringsAsFactors = F
  )
}

# Helper function, given a list of lines from the puzzle input, returns a 
# data frame containing one row per line number (recipe), ingredient, and
# allergen
parse_input <- function(input) {
  line_count <- length(input)
  only_words <- gsub('[\\(\\),]', '', input)
  split_lines <- strsplit(only_words, 'contains ')
  dfs <- mapply(line_to_df, 1:line_count, split_lines, SIMPLIFY = F)
  do.call(rbind, dfs)
}

# Processing -------------------------------------------------------------------

recipe_list <- parse_input(real_input)  # Parse the input

# For each type of allergen, return a data frame consisting of that allergen
# and all ingredients that appear in all recipes containing that allergen.
possible_allergens <- recipe_list %>% 
  group_by(allergen, recipe) %>% 
  summarise(ingredients = list(ingredient)) %>% 
  summarise(ingredient = reduce(ingredients, intersect))

# Remove all rows containing ingredients that definitely contain an allergen
# (though we're not sure which one), then count the number of ingredient
# and recipe combinations remaining
answer1 <- recipe_list %>% 
  filter(!(ingredient %in% possible_allergens$ingredient)) %>% 
  distinct(recipe, ingredient) %>% 
  nrow()

# Setup ------------------------------------------------------------------------

source('exercise_1.R')


# Processing -------------------------------------------------------------------

# Setup, count the number of ingredients that contain an allergen and create an
# empty data frame to add allergen/ingredient combinations to as we confirm
num_allergen_ingredients <- length(unique(possible_allergens$ingredient))
confirmed_allergens <- data.frame(
  allergen = character(0), 
  ingredient = character(0)
)

# Until the number of confirmed allergens is as large as the number of allergens...
while (nrow(confirmed_allergens) < num_allergen_ingredients) {
  
  # Remove all the previously confirmed allergens from the table of possible
  # allergens, then identify which ingredients are now associated with a 
  # single allergen
  newly_confirmed_allergens <- possible_allergens %>% 
    filter(!(allergen %in% confirmed_allergens$allergen)) %>% 
    group_by(ingredient) %>% 
    mutate(n = n()) %>% 
    filter(n == 1) %>% 
    select(-n)
  
  # Add the newly confirmed allergens to the ongoing list of confirmed 
  # allergens
  confirmed_allergens <- bind_rows(
    confirmed_allergens, 
    newly_confirmed_allergens
  )
}

# Sort the table of confirmed allergen/ingredient combinations by allergen name,
# then collapse the ingredient names with commas between
answer2 <- confirmed_allergens %>% 
  arrange(allergen) %>% 
  pull(ingredient) %>% 
  paste(collapse = ',')