omahar_nse_presentation.Rmd

---
title: "Standard Non-Standard Evaluation"
output: html_notebook
---


```{r}
library(tidyverse)
```

```{r}
mtcars
```
```{r}
mtcars %>%
  summarise(avg = mean(mpg))
```

```{r}
mtcars %>%
  summarise(avg = mean(hp))
```

```{r}
mtcars %>%
  summarise(avg = mean(wt))
```

## Hadley's Rule

Repeat yourself three times? Time for a function

```{r}
meaner <- function(data, column) {
  data %>%
    summarise(avg = mean(column))
}
```

```{r}
meaner(mtcars, hp)
```

```{r}
hp
```

```{r}
meaner(mtcars, "hp")
```

```{r}
mtcars %>%
  summarise(avg = mean("hp"))
```

```{r}
meaner2 <- function(data, column) {
  column <- enquo(column)
  data %>%
    summarise(avg = mean(!!column))
}
```

```{r}
meaner2(mtcars, hp)
```

## Enquo Magic!!!

`!!` and `enquo` magic worked!

Why?




## Names and Variables

```{r}
foo <- 4
```


The name of the variable is `foo`, but the value is 4


But what about `hp`?

```{r}
mtcars %>%
  summarise(avg = mean(hp))
```

`hp` is not a variable. It has no value.

```{r}
hp
```

But the `hp` has meaning in the context of the `mtcars`:

```{r}
with(mtcars, mean(hp))
```

```{r}
mean(mtcars$hp)
```


So in the tidyverse, names have special meaning.



So why doesn't this work?

```{r}
meaner <- function(data, column) {
  data %>%
    summarise(avg = mean(column))
}
```
```{r}
meaner(mtcars, hp)
```

Because the tidyverse `summarise` is looking for the column named `column`.




Tidyverse only cares about *names* not *values*.

```{r}
meaner2 <- function(data, column) {
  column <- enquo(column)
  data %>%
    summarise(avg = mean(!!column))
}
```
```{r}
meaner2(mtcars, hp)
```
`enquo` looks for the *name* typed in by the user: `hp`


`!!` tells tidyverse functions to use the *value* (which is a name)




## Why is it so complicated???

* most other programming languages don't use it:

> Zen of Python: Explicit is better than implicit

* progammers are taught to use variable *values* not *names*





## What are the benefits?

* Less typing during interactive programming

```{r}
mtcars %>%
  filter(cyl == 6, mpg > 20, am == 1)
```
vs.
```{r}
mtcars[mtcars$cyl == 6 & mtcars$mpg > 20 & mtcars$am == 1, ]
```


* Can use complicated expressions in functions

```{r}
mtcars %>%
  transmute(standardized = (hp - mean(hp)/sd(hp)))
```

```{r}
standardizer <- function(data, col) {
  col <- enquo(col)
  data %>%
    transmute(standardized = (!!col - mean(!!col)/sd(!!col)))
}
```
```{r}
standardizer(mtcars, hp)
```

* Can use `!!` + `enquo` like normal variables:
  - multiply, divide
  - built-in functions







## What about strings?




I want this:

```{r}
mtcars %>%
  group_by(cyl) %>%
  summarise(count = n())
```

By using a string:

```{r}
column <- "cyl"
mtcars %>%
  group_by(column) %>%
  summarise(count = n())
```

Same problem as before:
- tidyverse uses the *name* not the *value*

There is no column with *name* `column`







Does `enquo` + `!!` work?

```{r}
enquoed_column <- enquo(column)
mtcars %>%
  group_by(!!column) %>%
  summarise(count = n())
```

No.



`enquo` only works for *names*. 






We want the *value* `"cyl"` to be converted to a *name*.




Introducing: `sym`

```{r}
symed_column <- sym(column)
mtcars %>%
  group_by(!!symed_column) %>%
  summarise(count = n())
```

`!!` stays the same

We use `sym` instead of `enquo` if we want to turn 
a *value* into a *name*




We can also put this into a function

```{r}
grouper <- function(data, col) {
  col <- sym(col)
  data %>%
    group_by(!!col) %>%
    summarise(count = n())
}
```
```{r}
grouper(mtcars, "cyl")
```



## Any questions?

### Any functions you want to see?