R Packages

background-image: url("img/logo_padded.001.jpeg")
background-position: left
background-size: 60%
class: middle, center,

.pull-right[

<br>

## .base_color[R Packages]
 
<br>

#### .navy[Kelly McConville]

#### .navy[ Stat 108 | Week 10 | Spring 2023]

]

---

## Announcements

* Fill out [this form](https://forms.gle/sY39jB8WLSdFYwfMA) to help us create the Project 2 groups.
* Project 1 is due at 10pm on Wednesday.
    + Come by office hours with questions.

************************

## Week's Goals

.pull-left[

**Mon Lecture**

* Why `R` packages?
* `R` package creation demo

]

.pull-right[

**Wed Lecture**

* Documentation
* Package metadata

]

---

### Project 1

#### Suggestions

* Make sure to consider the peer feedback and to go through the Rubric/Requirements.
* Going from MVP to final product should take **several** iterations.
* Don't just focus on functionality.  Also, consider lay-out and aesthetics.
* Graphs should be high-quality, include appropriate context, and should follow Best Practices.
* Don't forget about the Data Scientist's Statement.

#### Feedback after looking over the dashboards:

* Use `theme()` or `update_theme()` to increase the font sizes.
* Don't worry if your app takes a while to load.  (If we were using the non-free version of Shinyapps, it would be faster.)
* Include more text to orient the user but be concise.
    + Use **bolding/color** and headers to help the user pull out the most relevant text.

---

### Refactoring Recap

* First write code that works.
* Then refactor it to be more
    + Readable
    + Efficient
    + Defensive
    + General
* But make sure the code still works after the refactoring!

---

### One More Refactoring Thought

* Jenny Bryan inspired much of the discussion from last Wed's lecture.  Reading her materials and watching her presentations has made me a better coder.

* Close with a thought from her:

<img src="img/cakes.png" width="65%" style="display: block; margin: auto;" />
 
---

class: middle, center

## Sharing Code

#### How do you share code?

---

### Sharing Code

#### Options

* `R` scripts and `R` Markdown documents
* RStudio Projects
* GitHub Repository
* R Package

#### How to pick between the options?

---

### When To Create an `R` Package

* If you have certain recurring operations that occur across multiple projects.
* Want to share data and documentation and not worry about file paths, file types, the documentation getting lost...

* **BUT**, you will still use an RStudio Project (with a corresponding GitHub repo) for the code and data that are specific to a given project.

---

### R Packages

* What is an R package?

> "R packages are the fundamental unit of R-ness". -- Jenny Bryan

* Contains functions, (possibly) datasets, documentation, and tests.

* "base R": 14 base packages that are preloaded
    + There are 15 other packages that also come preloaded.

* CRAN has > 19,000 more packages
    + `install.packages("dplyr")`
    + `library(dplyr)`

* And then there are all the packages on `GitHub`:
    + `devtools::install_github("hadley/dplyr")`
    + `library(dplyr)`

---

### R Data Packages

* Great way to share data!
    
    
* Example 1:
    + `library(mosaicData)`
    + `data(package = "mosaicData")`
    + `?Births2015`

* Example 2:    
    + `library(pdxTrees)`
    + `get_pdxTrees_parks()`
    + `get_pdxTrees_streets()`

* Example 3: 
    + `library(gbfs)`

---

### R Packages

#### Why create an R package?

* Very portable.
* Includes documentation.
* Provides a useful structure for organizing your work:
    + `R` folder: Code
    + `tests` folder: Testing functions
    + `data` folder: Storing data
* Lots of helper functions in other packages to automate parts of the process.

---

### Creating an R Package

Key packages:

* [`devtools`](https://cran.r-project.org/web/packages/devtools/index.html): supports the development and dissemination of the package

* [`usethis`](https://usethis.r-lib.org/): automates steps of package creation, such as constructing the data file

* [`roxygen2`](https://cran.r-project.org/web/packages/roxygen2/vignettes/roxygen2.html): simplifies writing documentation

---

class: middle, center

## Let's Build an `R` Package!

---

#### Our Package: `camDogs`

**Goals**:

* Share the [Dogs of Cambridge dataset](https://data.cambridgema.gov/General-Government/Dogs-of-Cambridge/sckh-3xyx), along with documentation.
* Create a function that outputs a dataset of the dogs with the 10 most common names.

* Useful MVP code:

```r
library(tidyverse)
camDogs <- read_csv("https://data.cambridgema.gov/api/views/sckh-3xyx/rows.csv?accessType=DOWNLOAD")

camDogs <- mutate(camDogs,
               Breed = if_else(Dog_Breed == "Mixed Breed",
                               "Mixed", "Single"))

top10 <- function(data, x){

# Find the 10 top based on x
  top10x <- count(data, {{x}}) %>%
    slice_max(n = 10, order_by = n) %>%
    select({{x}}) %>%
    pull()

# Filter dataset to only the 10 top based on x
  return(filter(data, {{x}} %in% top10x))
}
```

---

### Steps

* Let's go through the "Creating an R Package" hand-out.

* I will demo the process with [this Dogs of Cambridge dataset](https://data.cambridgema.gov/General-Government/Dogs-of-Cambridge/sckh-3xyx).

* If following along, clone this repo: [https://github.com/harvard-stat108s23/pkgDemo](https://github.com/harvard-stat108s23/pkgDemo)

---

### Wrap-Up Thought

#### Most Common Confusion When Moving from Scripts/R Markdowns to Package Writing

* Requires new ways of working with functions in other packages.
    * `DESCRIPTION` file to declare dependencies.
    * Use `package_name::function_name()`.
    * Can't use `library(package_name)`.
    * Want to avoid unnecessary dependencies.
    
    
#### Other questions?