# install.packages(c("readxl", "janitor"))
library(tidyverse)
library(readxl)
library(janitor)
d <- read_excel("data-raw/DATASET_srilanka.xlsx")
d |> glimpse()
# clean names
d <-
  d |> 
  janitor::clean_names()
d |> glimpse()
d$date
# so date as imported is all over the place. we have:
#  * something that looks like a human readable data, like "22/05/2019"
#    * it looks like most of those are of the format "DD/MM/YYYY"
#  * something that is just an integer, like "43471"
#    * these are actually Excel numerical version of a date
#  to get date in order in R we have to do things two ways:
d <- 
  d |> 
  rename(date_org = date) |> 
  mutate(date = case_when(!is.na(as.integer(date_org)) ~ excel_numeric_to_date(as.numeric(date_org)),
                          .default = dmy(date_org)))
# let's see what we got:
d |> knitr::kable()
# Given we know that the data entry was such that dates where supposed to be in 
#  chronological order we see that even though we have tried to convert values
#  to a date format that R understands. E.g. row where we have date "2019-05-31"
#  we get date "2019-01-06". This means that the date entered in Excel orginally
#  was wrong. So we need to do something else that what was tried above.
# Here we do:
#  Split the original date column into year, month and day
#  Where the date entry is wrong we will interpret day as month and month as day
#  Where right, we will just keep that value
# 
# Let's start from scatch
read_excel("data-raw/DATASET_srilanka.xlsx") |> 
  janitor::clean_names() |> 
  
d |> 
  



# get rid of redundunant variable (date_org)
d <- 
  d |> 
  select(date, boats = number_of_boats, catch = catch_number_of_individuals)
# Lets check boats vs catch
d |> 
  ggplot(aes(boats, catch)) +
  geom_point() +
  # fit a linear model:
  geom_smooth(method = "lm")
# lets calculate cpue (catch per boat) and then plot the histogram
d |> 
  mutate(cpue = catch / boats) |> 
  ggplot(aes(cpue)) +
  geom_histogram()
# lets look at the cpue over time
d |> 
  mutate(cpue = catch / boats) |> 
  ggplot(aes(date, cpue)) +
  geom_point() +
  geom_smooth()
# so most observations in june and july
# lets look at number of "observations" by month
d |> 
  mutate(date = floor_date(date, "month")) |> 
  group_by(date) |> 
  reframe(boats = sum(boats)) |> 
  ggplot(aes(date, boats)) +
  geom_col()