Reproduce the Problem

class: center, middle, inverse, title-slide

# Reproduce the Problem
## ⚔ in order to solve it
### Julia Piaskowski
### 2022-03-08

---

class: center, middle

## How to ask for help 
 
 
*the coding error* 
*unsolvable mystery* 
*make it simple, friend*

---
# Getting Help (again)

* Eventually, you will probably need outside help to solve a programming problem.
* there are many skilled R programmers out there: Stack Overflow, R4DS, RStudio Community, that one grad student who knows everything, ...
* You are asking them to kindly help you, but in order for them to do that, they need to understand your problem.

---
# Example 1

>I want to ask: are there significant differences in the yield responses of corn to 3 different levels of manure application with and without biofertilizer? What R code should I use? Thank you so much! Manure application levels = none, low, normal Biofertilizer = Y or N.

---
# Example 2

>I am new to R so I am struggling to figure out how to do this. I have a data frame with thousands of observations that looks like this  
x    total  subgroup1  subgroup2   
.03  1000     500       500   
.04  1050     550       500   
.07  1100     525       575   
.08  9900     4900      5000   
I am trying to come up with a weighted value for x based on the following equation.   
weighted x for subgroup1 = (Σ x*subgroup1) / (Σ subgroup1)   
The idea is that for each observation we multiply the x value by the subgroup 1 value and add them all together and divide it by the sum of all the subgroup values. I also want to do this for subgroup 2. I imagine I will need some type of loop function. However, I am not sure where to start in R. Any leads would be great.

---
class: center, middle

## Anyone feel ready to troubleshoot those problems in your R session?

---
# Reprex = reproducible example

= minimum version of what you are doing ~~in R~~ that is generating an error

![lion and cub](images/lion_and_cub.jpg)

---
# A good reprex includes...

1. A short description of what you are trying to do programatically (e.g. 'filter a data set based on condition')
1. Code in the language you are using that reproduces the problem for someone trying to solve this *on their system* (people often cannot just look at code and see what is wrong, they need to work with it).
1. The error message being produced and/or the incorrect or undesirable output produced by your code.

---
# Some good examples of a reprex

* [Example 1](https://stackoverflow.com/questions/71403633/conditional-formatting-of-cells-using-if-else-or-hiding-columns-in-dt-package-i) 
* [Example 2](https://stackoverflow.com/questions/71402703/how-to-obtain-the-specific-grouping-of-cases-and-controls-in-r)  
* [Example 3](https://stackoverflow.com/questions/71403688/save-kable-in-r-writes-empty-file)

---
# A minimal data set for a reprex

1. Use smallest, simplest data set. Preferably a built-in data set (e.g. iris or cars, yawn).  
1. If you have to create a data set, keep it small, keep it simple. 
  * minimum size (e.g. number of rows), minimum variables
  * enough information to replicate your situation - at least, the situation that is creating a problem

<center> **Remember, the goal is to recreate your problem!** </center>
<center> **Tailor your (minimal) data set to meet that goal.** </center>

---
# Tools to create a minimal data set

* `data.frame()`, `matrix()`, `list()` and other base R functions
* if you simulate new data through a random process, use `set.seed()`.

```r
set.seed(208)
my_df <- data.frame(x1 = c(1, 2, 3, 4),
 x2 = c("a", "a", "b", "b"),
 x3 = rnorm(4))
```

------------------

**mini-exercise:** make a `data.frame` with at least 2 variables and 5 rows.

---
# Tools to create a minimal data set

* Find those data sets! Use `data()`, `data(package = "dplyr")` or search the documentation (their data sets are listed in the help files).

*(no point in running these in an Rmarkdown document - do this in your R session)*

------------------

**mini-exercise:** find a data set from a tidyverse package you've never used before and examine it. How big is it? What type of variables are present?

---
# datapasta, a convenience package

[<img src="images/datapasta_hex.png" style="max-width:22%;min-width:60px;"
alt="datapasta" />](https://milesmcbain.github.io/datapasta/index.html)

* Format a table so it is represented as R code, and put it on your system clipboard:

```r
library(datapasta); library(dplyr)
iris %>% head() %>% df_format() 
```

* Next, paste the R-formatted data ([CTRL + V] or [CMD + V] on many systems) into an R source file.

---
# datapasta, a convenience package

* Pull part of table into R-formatted syntax:

First, copy part of a table from a spreadsheet (e.g. Googlesheets, Excel, Numbers, Calc).

Next, run this line of code to generate R code that will format your copied spreadsheet cells into an R data.frame:

```r
df_paste()
```

You can also run this line of code to generate R code that will format your copied spreadsheet cells into an R tibble:

```r
tribble_paste() 
```

---
class: center, middle, inverse

# Why do something so tedious??

### Because we need this code for a reprex. This is not for us - it's for the anonymous programmers out there we are asking to help us.

---
class: center, middle, inverse

# [exercise](exercises/exercise-8.html)

---
# Narrow down the problem:

* DON'T provide a long description of why you want to do something, "We plan to save the world by comparing storm drainage rates..."
* DON'T dump a long R script with the single descriptor "it didn't work" 
* DO Be very judicious in your use of screenshots. In most cases, text and code describes the problem. 
* DO find the line of code where it all breaks down (this does not mean digging into the source code - find the error in *your* code) 
* DO include all lines of code needed to recreate the problem - including libraries that must be loaded 
* DON'T load a meta-library like `library(tidyverse)` - only load the packages needed to recreate the problem
* DO ruthlessly strip out every bit of unneeded code, any extra lines, extraneous arguments that can be ignored and still the error persists

---
# Be considerate of your anonymous helpers

.pull-left[
<img src="images/kind_rewind.jpg" width="840" />
]

.pull-right[
* no `rm = list(ls())` since it is running on someone else's system
* no `setwd("path/to/my/computer")`
* if your code changes systems systems (e.g. `par(...)`, make sure you reset it at the end of the reprex code
* don't mask built-in functions (e.g. `mean()`)  with your own custom functions
]

---
# Last step: using `reprex()`

* **reprex** is a package supporting how to generate reproducible examples. 
* like **datapasta**, it can make your life easier

Let's pretend the code below is generating an error. Copy both lines of code.

```r
x<- 1:10
plot(x)
```

Next, run this code in an R session:

```r
reprex::reprex()
```
**You won't see much, but, valuable information has been put on your system clipboard!**

Paste your clipboard contents on the resource  where you are seeking help (e.g. [GitHub issue](https://github.com/IdahoAgStats/what-they-forgot-202203/issues), Stack Overflow)

---
# Example: Let's Write a Reprex!

![](images/need_reprex.png)

---
class: middle, center, inverse

# [exercise](exercises/exercise-9.html)

---
# Use good style

### (it's easier to read)

* There is no cap on white space usage:

```r
# No need to pack in code this tightly! 
# (We aren't trying to send this data stream to the moon)
myfunc<-onebigfunc(mydf,arg=1,arg2=TRUE,arg4="Go")
more<-evenmore(myfunc,argA=c("Idaho","Montana","Utah"))
if(condition){dothis}
else{dothat}
```

* Pay attention to function examples in documentation for code styling 
* There is actually a [code guide!](https://style.tidyverse.org/syntax.html) and a package, [styler](https://styler.r-lib.org/) to support this.

Tip: CTRL + Shift + a will reformat things in RStudio.

---
# Show me the money!

* Creating a good reprex is **a lot** of work.
* If you want to elicit feedback to solve your problem, this is how to do that. 
* Very frequently, the process of creating a reprex will reveal an error, a mistaken assumption, or something that will help you resolve the problem.

---
# Other guides for creating a good reprex

### [Stack Overflow](https://stackoverflow.com/help/minimal-reproducible-example)

### [Rstudio Community](https://community.rstudio.com/t/faq-how-to-do-a-minimal-reproducible-example-reprex-for-beginners/23061)

### [Tidyverse Reprex Do's and Don'ts](https://reprex.tidyverse.org/articles/reprex-dos-and-donts.html)