Reproducible Data Workflows With Drake

class: left bottom hide-count
background-image: url(assets/images/bench-accounting-xxeAftHHq6E-unsplash.jpg)
background-size: 100%

---

.f1.pt4[
<img class="icon-huge pr3" src="assets/images/cloud-upload.svg" width="100px"/>
`rstudio.cloud/project/405721`
]

.f1[
<img class="icon-huge pr3" src="assets/images/desktop-download.svg" width="100px"/>
.code.moffitt-blue[usethis::use_zip(".moffitt-orange[github &#x1F447;]")]
]

.f1[
<img class="icon-huge pr3" src="assets/images/mark-github.svg" width="100px"/>
[github.com/.moffitt-orange[gadenbuie/drake-intro]](https://github.com/gadenbuie/drake-intro)
]

.f2.silver.center.mt5[Find someone to sit next to and share laptops]

---

## What is drake?

![](assets/images/drake-meme.jpg)

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-02.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-03.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-04.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-05.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-06.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-07.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-08.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-09.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-10.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-11.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-12.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-13.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-14.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-15.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-16.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-17.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-18.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-19.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-20.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-21.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-22.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-23.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-24.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-25.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-26.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-27.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-28.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-29.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-30.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-31.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/messy-workflow-32.jpg')
background-size: contain

---

class: middle
background-image: url('assets/images/drawn/messy-workflow-32.jpg')
background-size: contain

.f2.moffitt-bg-light-blue-o90.white.pa5.shadow-3[
* Will this work when I come back to it later?

* What happens if I re-run _everything_?

* Am I certain that the results are still valid?
]

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/project-spectrum-33.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/project-spectrum-34.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/project-spectrum-35.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/project-spectrum-36.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/drake-intro-37.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/drake-intro-38.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/drake-intro-39.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/drake-intro-40.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/drake-intro-41.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/drake-intro-42.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/drake-intro-43.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/drake-intro-44.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/drake-intro-45.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/drake-intro-46.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/drake-intro-47.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/drake-intro-48.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/drake-intro-49.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/drake-intro-50.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/drake-intro-51.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/drake-intro-52.jpg')
background-size: contain

---

class: hide-count animated fadeIn
background-image: url('assets/images/drawn/drake-intro-53.jpg')
background-size: contain

---

layout: true
class: animated fadeIn
background-image: url('assets/images/drake-infographic.svg')
background-size: 100%
background-position: 40% left
---

.bg-white.h-100.w-80.fixed.o-90.cover-2-3[

]

---

.bg-white.h-100.w-80.fixed.o-90.cover-3[

]

---

---
layout: false
class: center middle moffitt-bg-blue inverse hide-count

# Detour — Functions

---
layout: true
class: animated fadeIn
---

.f1.mt4.pa4.code[
.o-0[verb <- .dark-gray[function(].moffitt-bg-blue.moffitt-blue.pa3.o-0[x, y = 0, ...]) {]  
.moffitt-green.mh3.pv1.ph1[
x_range <- max(x) - min(x)

(x - min(x)) / x_range + y
]

.o-0.dark-gray[}]
]
 
---

.f1.mt4.pa4.code.dark-gray[
.o-0[verb <- function(.moffitt-bg-blue.moffitt-blue.pa3.o-0[x, y = 0, ...]) {]  
.moffitt-bg-green.moffitt-green.mh3.pv1.ph1[
x_range <- max(x) - min(x)

(x - min(x)) / x_range + y
]

---

.f1.mt4.pa4.code.dark-grey[
verb <- function() {  
.moffitt-bg-green.moffitt-green.mh3.pv1.ph1.o-60[
x_range <- max(x) - min(x)

(x - min(x)) / x_range + y
]

}
]

---

.f1.mt4.pa4.code.dark-gray[
verb <- function(.moffitt-bg-blue.moffitt-blue.pa3[x, y = 0, ...]) {  
.moffitt-bg-green.moffitt-green.mh3.pv1.ph1.o-40[
x_range <- max(x) - min(x)

(x - min(x)) / x_range + y
]

}
]

---

.f1.mt4.pa4.code.dark-gray[
verb <- function(.moffitt-bg-blue.moffitt-blue.pa3[x, y = 0, ...]) {  
.moffitt-bg-green-o10.moffitt-green.mh3.pv1.ph1[
x_range <- max(x) - min(x)

(x - min(x)) / x_range + y
]

}
]

---

.f1.mt4.pa4.code.dark-gray[
verb <- function(.moffitt-bg-blue-o20.moffitt-blue.pa3[x, y = 0, ...]) {  
.moffitt-bg-green-o10.moffitt-green.mh3.pv1.ph1[
x_range <- max(.moffitt-blue[x]) - min(.moffitt-blue[x])

(.moffitt-blue[x] - min(.moffitt-blue[x])) / x_range + .moffitt-blue[y]
]

}
]

---

}
]

---

.moffitt-bg-orange-o20[.moffitt-orange[return(](.moffitt-blue[x] - min(.moffitt-blue[x])) / x_range + .moffitt-blue[y].moffitt-orange[)]]
]

}
]

---

.f1.mt4.pa4.code.dark-gray.moffitt-bg-grey-o20.ba.b--dotted.bw2.b--dark-gray.border-box[
verb <- function(.moffitt-bg-blue.moffitt-blue.pa3.o-30[x, y = 0, ...]) {  
.moffitt-bg-green.moffitt-green.mh3.pv1.ph1.o-40[
x_range <- max(x) - min(x)

(x - min(x)) / x_range + y
]

}
]

---

.f1.mt4.pa4.code.dark-gray.moffitt-bg-grey-o20[
verb <- function(.moffitt-bg-blue-o20.moffitt-blue.pa3[x, y = 0, ...]) {  
.moffitt-bg-green-o10.moffitt-green.mh3.pv1.ph1[
x_range <- max(.moffitt-blue[x]) - min(.moffitt-blue[x])

(.moffitt-blue[x] - min(.moffitt-blue[x])) / x_range + .moffitt-blue[y]
]

}
]

---

.f1.mt4.pa4.code.dark-gray.moffitt-bg-grey-o20[
scale <- function(.moffitt-bg-blue-o20.moffitt-blue.pa3[x, y = 0, ...]) {  
.moffitt-bg-green-o10.moffitt-green.mh3.pv1.ph1[
x_range <- max(.moffitt-blue[x]) - min(.moffitt-blue[x])

(.moffitt-blue[x] - min(.moffitt-blue[x])) / x_range + .moffitt-blue[y]
]

}
]

---

.f1.mt4.pa4.code.dark-gray.moffitt-bg-grey-o20[
scale <- function(.moffitt-bg-blue-o20.moffitt-blue.pa3[x, y = 0, .moffitt-red[...]]) {  
.moffitt-bg-green-o10.moffitt-green.mh3.pv1.ph1[
x_r <- max(.moffitt-blue[x], .moffitt-red[...]) - min(.moffitt-blue[x], .moffitt-red[...])

(.moffitt-blue[x] - min(.moffitt-blue[x], .moffitt-red[...])) / x_r + .moffitt-blue[y]
]

}
]

---
layout: false

## Function Review

* Functions take **inputs** .muted[(also called .code[formals()])]

* ... use the inputs in the function **.code[body()]**

* ... or find variables within their **scope** .muted[(or .code[environment()])]

* ... and **.code[return()]** a value

---

## Your Turn!

The formula for converting Celsius to Fahrenheit is

.center[
<span style="font-size: 100%; display: inline-block; position: relative;" id="MathJax-Element-1-Frame" tabindex="0" role="presentation" class=""><span class="MJX_Assistive_MathML MJX_Assistive_MathML_Block" role="presentation"><math xmlns="http://www.w3.org/1998/Math/MathML" display="block"><msub><mi>T</mi><mi>F</mi></msub><mo>=</mo><mfrac><mn>9</mn><mn>5</mn></mfrac><msub><mi>T</mi><mi>C</mi></msub><mo>+</mo><mn>32</mn></math></span></span>
]

* Write a function that converts &deg;C to &deg;F.  
  What is 20&deg;C?

.muted.w-60[
Extra challenge:

* Your friend is Canadian and keeps sharing their local weather in &deg;C.
  Write a function that decides if it's hot or not based on the infinitely more reasonable threshold of 78&deg;F.
  
* 78&deg;F is a debatable, make it a parameter
]

???

Check with someone else. What is 20&deg;C? It should be 68&deg;F

---

## Function Style Guide

.f2.pt4[
* Clear **names** like `verb()` or `verb_thing()`
]

.f2.pt4[
* **Short** functions that do **one**.muted[-ish] thing
]

.f2.pt4[
* Limit **side-effects** or be explicit .silver[e.g. .code[output = ""]]
]

---
layout: true
## Fail or Return Early

---

Check your inputs at the top of your function. For type checking you can use

```r
verb <- function(x) {
  stopifnot(is.numeric(x))
  
  # ... more code
}
```

or for more control over error messages

```r
verb <- function(x) {
  if (x < 0) {
    stop("`x` must be non-negative")
  }
}
```

---

Return early from functions to minimize `if ... else` indenting

.absolute.pa3.red.bg-washed-red.bottom-2.right-0.w-40.f2[&#x2639;]

```r
verb <- function(df) {
  
  # check that data is valid
  
  if (nrow(df) > 0) {
    # process the valid data
    answer <- df %>% 
      # ... many
      # ... lines
      # ... of code
  } else {
    # or give back a default value
    answer <- FALSE
  }
  
  answer
}
```

---

Return early from functions to minimize `if ... else` indenting

```r
verb <- function(df) {
  
  # check that data is valid
  
  if (nrow(df) == 0) {
    return(FALSE)
  }
  
  # ... process the valid data
}
```

---
layout: false

## Pass the Dots

Use the power of the .moffitt-red.moffitt-bg-red-o10.code[...] with the .pkg[tidyverse]

```r
grouped_mean <- function(df, ...) {
  df %>% 
    group_by(...) %>% 
    summarize(mean_expr = mean(expr))
}
```

.footnote[
Jenny Bryan: [Lazy Evaluation](https://resources.rstudio.com/rstudio-conf-2019/lazy-evaluation), rstudio::conf(2019)  
Sharla Gelfand: [tidyeval](https://sharla.party/posts/tidyeval/), sharla.party
]

---
class: middle center moffitt-bg-light-blue white hide-count

## Your Turn Again

Update your temperature conversion function to reject  
non-numeric, unreasonable inputs

---
exclude: true

.mt1.pa4.code.mw7[
verb <- .dark-gray[function(].moffitt-bg-blue.moffitt-blue.pa3[x, y = 0, ...]) {  
.moffitt-bg-green.moffitt-green.mh3.pv1.ph4[
x_range <- max(x) - min(x)

(x - min(x)) / x_range + y
]

---

## An Example drake Project

---
layout: true

#### .code.gray[usethis::use_course("gadenbuie/drake-intro")]
]

<hr />

---

* Open .moffitt-bg-grey-o20.moffitt-blue.code.pa2[drake.R]

* What packages and functions are loaded?

* Walk through .moffitt-bg-grey-o20.moffitt-blue.code.pa2[R/plan.R]

* Preview the work plan

* Run the plan

---

* Use .moffitt-bg-grey-o20.moffitt-red.code.pa2[readd()] and .moffitt-bg-grey-o20.moffitt-red.code.pa2[loadd()] to view results

* View the .moffitt-bg-grey-o20.moffitt-blue.code.pa2[report.html]

* How are _targets_ used in .moffitt-bg-grey-o20.moffitt-blue.code.pa2[report.Rmd]?

* How does .pkg[drake] track dependencies in R Markdown?

* Do we need to run .moffitt-bg-grey-o20.moffitt-blue.code.pa2[make.R] to be able to update .moffitt-bg-grey-o20.moffitt-blue.code.pa2[report.Rmd]?
 
---

The life expectancy plot currently only shows results for Tampa.

* Modify .moffitt-bg-grey-o20.code.pa2[plot_life_exp_gender_income()] in .moffitt-bg-grey-o20.moffitt-blue.code.pa2[R/functions.R] to have a `czname` argument.

* Update the .moffitt-bg-grey-o20.moffitt-blue.code.pa2[R/plan.R] to plot .b[Denver].

* Add .b[Denver] to the .moffitt-bg-grey-o20.moffitt-blue.code.pa2[report.Rmd] to compare with .b[Tampa].

---

* Convert the code in .moffitt-bg-grey-o20.moffitt-blue.code.pa2[R/scratch.R] into a function .moffitt-bg-grey-o20.code.pa2[plot_life_exp_income()] and add it to the .moffitt-bg-grey-o20.moffitt-blue.code.pa2[R/plan.R].

---
layout: false

## But wait, there's more!

There is a lot more that .pkg[drake] can do, including:

1. [Predict total runtime](https://ropenscilabs.github.io/drake-manual/time.html#predict-total-runtime) of your plan

```r
  predict_runtime(config)
  ```

1. Parameterized targets for hyperparameter selection in [large plans](https://ropenscilabs.github.io/drake-manual/plans.html#large-plans)

1. [Parallel computation](https://ropenscilabs.github.io/drake-manual/hpc.html#parallel-backends) of targets

1. Distribute and run targets on [HPC clusters](https://ropenscilabs.github.io/drake-manual/hpc.html)

---

### Learn More with These Resources

* [User Manual](https://ropenscilabs.github.io/drake-manual/)
* [Package Docs](https://ropensci.github.io/drake/)
* [ropensci/drake](https://github.com/ropensci/drake)
]
.w-60.ph4[
Functions

* [Advanced R: Functions](https://adv-r.hadley.nz/functions.html)
* [Functions - Nice R Code](https://nicercode.github.io/guides/functions/)
* [Programming with R: Creating Functions](https://swcarpentry.github.io/r-novice-inflammation/02-func-R/)
]
]

.footnote[
**Icons** by [Github octicons](https://octicons.github.com/) and the following [Noun Project](https://thenounproject.com/search/?q=Report&i=180805) icon creators:
[Kirby Wu, TW](https://thenounproject.com/search/?q=json&i=966215),
[Iga](https://thenounproject.com/search/?q=document&i=2711779),
[Lil Squid](https://thenounproject.com/search/?q=report&i=149914),
[Wichai Wi](https://thenounproject.com/search/?q=results&i=2294590),
[Nick Kinling](https://thenounproject.com/search/?q=detour&i=788361).

**Built with** [rmarkdown](https://rmarkdown.rstudio.com), [xaringan](https://slides.yihui.name/xaringan), [xaringanthemer](https://pkg.garrickadenbuie.com/xaringanthemer), [remark.js](http://remarkjs.com/), [tachyons.css](http://tachyons.io/), [animate.css](https://daneden.github.io/animate.css/)
]

.drake-logo {
  background-image: url('assets/images/drake-logo.svg');
  background-size: cover;
  background-repeat: no-repeat;
}

.talk-meta {
  font-family: Overpass;
  position: absolute;
  text-align: left;
  bottom: 10px;
  left: 25px;
}

.talk-author {
  color: #444;
  font-weight: bold;
  font-size: 1.5em;
  line-height: 1em;
}

.talk-date {
  color: #666;
  font-size: 1.25em;
  line-height: 0;
}

.icon-huge {
  position: relative; 
  top: 35px;
}

.cover-2-3 {
  left: 33%;
}

.cover-3 {
  left: 66%;
}
</style>