2019-12-03_pure-functions.Rmd

---
output: github_document
---

# [Avoid hidden arguments](https://principles.tidyverse.org/args-hidden.html)

<https://twitter.com/mauro_lepore>

License: [CCO](https://creativecommons.org/choose/zero/?lang=es)

## Hidden arguments make code harder to reason about, because to correctly predict the output you also need to know some other state

```{r}
y <- 1
add <- function(x) {
  x + y
}
add(1)

y <- 10 ## It is hard to keep track of this

add(1)  
```

## Functions are easier to understand if the results depend only on the values of the inputs

##  How can I remediate the problem?

If you have an existing function with a hidden input:

1. Make sure the input is an explicit option.
2. Make sure it’s printed.

## For example, take `prepare_data()`

The output depends on `data`, but it is hidden.

```{r}
prepare_data <- function() {
  data <- read.csv(path)
  data[1:2, 1:2]
}

path <- tempfile()
readr::write_csv(mtcars, path)

prepare_data()
```

## 1. `prepare_data()` gains the explicit argument `data`

```{r}
prepare_data <- function(data = read.csv(path)) {
  data[1:2, 1:2]
}

prepare_data()
```

## 2. `prepare_data()` now prints `data`

```{r}
prepare_data <- function(data = read.csv(path)) {
  if (missing(data)) {
    message(
      "Using `data` with names: ", paste(names(data), collapse = ", ")
    )
  }
  data[1:2, 1:2]
}

prepare_data()

prepare_data(read.csv(path))
```

## But `data` should be supplied

> Data arguments provide the core data. They are required, and are usually vectors and often determine the type and size of the output. Data arguments are often called `data`, `x`, or `y` -- [tidyverse design guide](https://principles.tidyverse.org/args-data-details.html).

```{r}
prepare_data <- function(data) {
  data[1:2, 1:2]
}

try(prepare_data())

data <- read.csv(path)
prepare_data(data)
```

# Some functions do need to depend on external state ...
 
## A function has hidden arguments when it returns different results with the same inputs in a surprising way

## Surprising

```{r}
getOption("stringsAsFactors")
data.frame(x = "a")$x

old_options <- options(stringsAsFactors = FALSE)
on.exit(old_options)

getOption("stringsAsFactors")
data.frame(x = "a")$x
```

Global options should not affect computation.

## Not surprising

`read_csv(path)` depends not only on `path` but also on the contents of the file, but that is not surprising.

```{r, message=FALSE}
library(readr)
path <- tempfile()

write_csv(mtcars, path)

names(read_csv(path))

write_csv(iris, path)

names(read_csv(path))
```