---
title: "Getting started with stbl"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Getting started with stbl}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)
```

```{r setup}
library(stbl)
```

The goal of {stbl} is to help you stabilize function arguments *before* you use them. This is especially important when a function performs a slow or expensive operation, like writing to a database or calling a web API. You want to fail fast and with a clear error message if the inputs aren't right.

This vignette demonstrates this principle by incrementally building a single function, `register_user()`, which validates several arguments before hypothetically sending them to an external service.

## The `register_user()` Function

Here is the base function we'll be improving. Without any checks, it's vulnerable to bad inputs that could cause cryptic errors later on or send corrupt data to our external service.

```{r}
register_user <- function(username,
                          email_address,
                          age,
                          is_premium_member,
                          interests) {
  # Imagine this is a slow API call
  list(
    username = username,
    email_address = email_address,
    age = age,
    is_premium_member = is_premium_member,
    interests = interests
  )
}
```

## Step 1: Handling a Vector with `to_*()`

Let's start adding checks. The first check will be for the `interests` argument. We expect this to be a character vector, but we're not picky about the content. `to_chr()` will convert inputs that are character-like (like factors or a simple list of strings) into a proper character vector.

```{r}
register_user <- function(username,
                          email_address,
                          age,
                          is_premium_member,
                          interests) {
  interests <- to_chr(interests)

  list(
    username = username,
    email_address = email_address,
    age = age,
    is_premium_member = is_premium_member,
    interests = interests
  )
}

# It works with a character vector or a list of characters. We use `str()` to 
# make the output easier to see.
register_user(
  username = "test_user", 
  email_address = "test@example.com", 
  age = 42, 
  is_premium_member = TRUE, 
  interests = c("R", "hiking")
) |> str()
register_user(
  username = "test_user", 
  email_address = "test@example.com", 
  age = 42, 
  is_premium_member = TRUE, 
  interests = list("R", "hiking")
) |> str()
```

If the input is something that cannot be reasonably flattened to a character vector, it fails with a helpful message.

```{r, error = TRUE}
# Fails because the list contains a function, which is not character-like.
register_user(
  username = "test_user", 
  email_address = "test@example.com", 
  age = 42, 
  is_premium_member = TRUE, 
  interests = list("R", mean)
)
```

## Step 2: Simple Scalar Coercion with `to_*_scalar()`

Next, we'll add checks for `age` and `is_premium_member`. These arguments must each contain a single value. We'll use the `_scalar` variants: `to_int_scalar()` and `to_lgl_scalar()`. These functions are liberal in what they accept. For example, `to_lgl_scalar()` understands that `1`, `"T"`, and `"True"` all mean `TRUE`.

```{r}
register_user <- function(username,
                          email_address,
                          age,
                          is_premium_member,
                          interests) {
  interests <- to_chr(interests)
  age <- to_int_scalar(age)
  is_premium_member <- to_lgl_scalar(is_premium_member)

  list(
    username = username,
    email_address = email_address,
    age = age,
    is_premium_member = is_premium_member,
    interests = interests
  )
}

# This works.
register_user(
  username = "test_user", 
  email_address = "test@example.com", 
  age = "42", 
  is_premium_member = TRUE, 
  interests = c("R", "hiking")
) |> str()
```

They fail clearly if the input isn't a single value or can't be coerced.

```{r, error = TRUE}
# Fails because age is not a single value.
register_user(
  username = "test_user", 
  email_address = "test@example.com", 
  age = c(30, 31), 
  is_premium_member = TRUE, 
  interests = c("R", "hiking")
)

# Fails because "forty-two" cannot be converted to an integer.
register_user(
  username = "test_user", 
  email_address = "test@example.com", 
  age = "forty-two", 
  is_premium_member = TRUE, 
  interests = c("R", "hiking")
)
```

## Step 3: Complex Validation with `stabilize_*()`

Finally, let's add our most complex validation for `username` and `email_address`. For these, simple type coercion isn't enough; we need to check their content and structure using `stabilize_chr_scalar()`. This function first coerces the input, then applies a list of validation `rules`. You should prefer the faster `to_*()` functions and only "upgrade" to `stabilize_*()` when you need these additional checks.

This is our final, fully stabilized function:

```{r}
register_user <- function(username,
                          email_address,
                          age,
                          is_premium_member,
                          interests) {
  interests <- to_chr(interests)
  age <- to_int_scalar(age)
  is_premium_member <- to_lgl_scalar(is_premium_member)

  space_regex <- c("must not contain spaces" = "\\s")
  attr(space_regex, "negate") <- TRUE
  username <- stabilize_chr_scalar(
    username,
    regex = space_regex
  )

  email_regex <- "^[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\\.[a-zA-Z]{2,}$"
  email_address <- stabilize_chr_scalar(
    email_address,
    regex = c("must be a valid email address" = email_regex)
  )

  list(
    username = username,
    email_address = email_address,
    age = age,
    is_premium_member = is_premium_member,
    interests = interests
  )
}

# A successful call.
register_user(
  username = "test_user", 
  email_address = "test@example.com", 
  age = 42, 
  is_premium_member = TRUE, 
  interests = c("R", "hiking")
) |> str()
```

And here are some examples of it failing, with informative error messages for each of our new rules.

```{r, error = TRUE}
# Fails because the username has a space.
register_user(
  username = "test user", 
  email_address = "test@example.com", 
  age = 42, 
  is_premium_member = TRUE, 
  interests = c("R", "hiking")
)

# Fails because the email address is invalid.
register_user(
  username = "test_user", 
  email_address = "not-a-valid-email", 
  age = 42, 
  is_premium_member = TRUE, 
  interests = c("R", "hiking")
)
```