NSW fuel data -cyclic recoding

# We also want to identify services stations which are the same... 
# but where the address is written differently
# in a way not corrected for, above...


unique_stations %>%
  distinct(full_name, new_address) %>%
  group_by(full_name) %>%
  mutate(rank = row_number()) %>%
  spread(rank, new_address) %>%
  mutate(response = NA_character_) %>%
  select(full_name, response, everything()) %>%
  openxlsx::write.xlsx("are these dupliate names.xlsx", asTable = T)


# if there are duplicates... 
# How should we idenitify them?
# and what should we do? 
# If they are the same identify the Rank of the one that you will assign all entries


recoded_addresses <- readxl::read_excel("reference/are these dupliate names.xlsx", col_types = "text") %>%
  filter(!is.na(response)) %>%
  gather(key, new_address, -full_name, -response) %>%
  filter(!is.na(new_address))


addresses_ready <- recoded_addresses %>%
  left_join(recoded_addresses %>%
              filter(!is.na(as.numeric(response)) & as.numeric(response) == key) %>%
              select(-key) %>%
              rename(replacement_address = new_address),
            by = c("full_name", "response")) %>%
  mutate(replacement_address = if_else(is.na(replacement_address), new_address, replacement_address)) %>%
  select(-response, -key)


# take the replacements into account

stations_we_are_using <- unique_stations %>%
  inner_join(addresses_ready, by = c("full_name", "new_address")) %>%
  select(-new_address) %>%
  mutate(unique_store_id = paste0(full_name, replacement_address, sep = "|")) %>%
  semi_join(valid_stores, by = c("unique_store_id" = "Unique Store Id"))