options(max.print = 1e3, scipen = 999, width = 1e2)
options(stringsAsFactors = F)

suppressPackageStartupMessages({
  library(dplyr)
  library(tidyr)
  
  library(lubridate)
  library(stringr)
  
  library(gt)
  library(tidyverse)
  library(glue)
  
  library(ggplot2)
  
  library(azuremlsdk)
})

.azureml_dataset_name <- "Currencies"

ws <- azuremlsdk::load_workspace_from_config()
sprintf(
  "%s workspace located in %s region", ws$name, ws$location
)

currencies_ds <- azuremlsdk::get_dataset_by_name(ws, name = .azureml_dataset_name)

sprintf(
  "Dataset name: %s. %s",
  currencies_ds$name, 
  currencies_ds$description
)

quotes_df <- currencies_ds$to_pandas_dataframe()

# ~ 20 years, 150 currencies and 1.5M rows

quotes_df %>%
  filter(slug == "USD/RUB") %>%
  select(-slug) %>% 
  top_n(10) %>% 
  
  gt() %>%
  tab_header(
    title = "USD/RUB Rate",
    subtitle = glue("{min(quotes_df$date)} to {max(quotes_df$date)}")
  ) %>%
  fmt_date(
    columns = date,
    date_style = 6
  ) %>%
  fmt_number(
    columns = c(open, high, low, close)
  )

quotes_df %<>% 
  transmute(
    symbol = slug,
    price = close,
    date
  ) %>% 
  
  filter(
    str_detect(symbol, "USD/") &
    date > max(date) - lubridate::years(10)
  ) %>% 
  
  filter(!(symbol == "USD/RUB" & price < 1)) %>% 
  
  arrange(date) %>% 
  group_by(symbol) %>%
  
  mutate(
    return = c(NA_real_, diff(price))/lag(price),
    log_return = log(1 + return)
  ) %>% 
  na.omit

quotes_stats <- quotes_df %>% 

  summarise(
    max_price = max(price), 
    min_price = min(price),
    last_price = last(price),
    max_min_rate = max(price)/min(price),
    volatility = sd(log_return)
  )

quotes_stats %>% 
  mutate(
    `100x Volatility` = volatility*100
  ) %>% 
  arrange(volatility) %>% 
  select(-volatility) %>% 
  
  gt() %>% 
  tab_header(
    title = "The Least and The Most Volatile Currencies",
    subtitle = glue("{min(quotes_df$date)} to {max(quotes_df$date)}")
  ) %>%
  fmt_number(
    columns = c(max_price, min_price, max_min_rate, last_price, `100x Volatility`)
  )

symbols  <- c(
  'RUB', 
  'EUR', 'GBP', 'CHF', 'CNY', 'HKD', 'JPY', 'SEK', 'SGD', 'AUD',  
  'AED', 'KZT', 'BYN', 'TRY', 'MXN'
)

symbols <- str_c("USD", symbols, sep = "/")


quotes_stats %>% 
  filter(
    symbol %in% symbols
  ) %>% 
  mutate(
    `100x Volatility` = volatility*100
  ) %>% 
  arrange(volatility) %>% 
  select(-volatility) %>% 
  
  gt() %>% 
  tab_header(
    title = "The Most Promised Currencies",
    subtitle = glue("{min(quotes_df$date)} to {max(quotes_df$date)}")
  ) %>%
  fmt_number(
    columns = c(max_price, min_price, last_price, max_min_rate, `100x Volatility`)
  )

usdrub_vol <- quotes_stats %>% filter(symbol == "USD/RUB") %>% pull(volatility)

low_risk_symbols <- quotes_stats %>% 
  filter(
    symbol %in% symbols &
    volatility <= usdrub_vol
  ) %>% 
  pull(symbol) %>% 
  unique

cat(
  sprintf(
    "['%s']",
    paste(low_risk_symbols, collapse = "', '")
))

jumper_symbols <- quotes_stats %>% filter(max_min_rate > 2) %>% pull(symbol)

quotes_df %>% 
  filter(symbol %in% low_risk_symbols) %>% 
  mutate(
    jumper = if_else(symbol %in% jumper_symbols, "High risk currencies", "Low risk currencies")
  ) %>% 
  group_by(symbol) %>% 
  mutate(R = cumsum(return)) %>% 
  
  ggplot + 
    geom_line(aes(x = date, y = R, color = symbol)) +
    scale_y_continuous(labels = scales::percent_format(accuracy = 1)) +

    facet_grid(jumper ~ ., scales = "free") +
    
    labs(
      title = "Currencies Exchange Rates", subtitle = "Return of Investment for last 10 years",
      x = "", y = "Return of Investment", 
      caption = currencies_ds$description) +
    theme_minimal() +
  
    theme(
      legend.position = "top", legend.title = element_blank(),
      plot.caption = element_text(size = 8)
    )