Debugging

class: center, middle, inverse, title-slide

# Debugging
## ⚔<br/>in R
### Julia Piaskowski
### 2022-03-08

---

# When you encounter an R error:

1. Check package documentation
1. Ask around: RStudio community, R4DS slack, Stack Overflow, your local oracle
1. Check source code (maybe)

#### You will probably need to work with debugging functions eventually

---
class: center, middle, inverse

## debugging = removing code errors  
<br> 
# How do you debug?

---
# Retrace your steps

1. Go through a process step-by-step and see where things went wrong
1. Start with the action that occurred that first and proceed to the second action, the third, etc...

---
# Retrace Example

[Agricultural Data from Tidy Tuesday](https://github.com/rfordatascience/tidytuesday/tree/master/data/2020/2020-09-01)

```r
library(dplyr); library(tidyr); library(purrr)
agdata <- readRDS(here::here("static", "slides", "data", "agdata.rds"))
yields <- agdata[[4]]
colnames(yields) <- gsub("\$tonnes per hectare\$", "", names(yields))

wheat <- yields %>% tidyr::pivot_wider(id_cols = Year, 
                                  names_from = Entity,
                                  values_from = Wheat) %>% 
  purrr::discard(~all(is.na(.))) %>% 
  mutate(ethiopia = case_when(
    is.na(Ethiopia) ~ `Ethiopia PDR`,
    TRUE ~ Ethiopia)) %>% select(-Ethiopia, -`Ethiopia PDR`)
```

```
## Error in `stop_subscript()`:
## ! Can't subset columns that don't exist.
## x Column `Wheat` doesn't exist.
```

<mark>What went wrong? </mark>

---
# Fixed:

```r
yields <- agdata[[4]]
colnames(yields) <- gsub(" \$tonnes per hectare\$", "", names(yields))

---
# Another retrace example

**Working with nested functions makes debugging hard**

#### Hard:

```r
result <- function4(function3(function2(function1(my_object, extra_args = 1),more_args = TRUE), even_more_args = "show_me_the_money"), final_arg = FALSE)
```
--
#### Easier

```r
pre_result1 <- function1(my_object, extra_args = 1)
pre_result2 <- function2(pre_result1, more_args = TRUE)
pre_result3 <- function3(pre_result2, even_more_args = "show_me_the_money")
result <- function4(pre_result3, final_arg = FALSE)
```
*(Lesson: avoid writing nested functions)*

---
# Consider writing a function if....

* You need to do something repeatedly
* That thing that needs to be repeated has several steps
* If you have to *cut-paste-replace* more than twice, consider a function to automate that process

---
# Why write functions?

* To save yourself work  
* To avoid mistakes
<br><br>
  
👍  DRY = **D**on't **R**epeat **Y**ourself  
<br>
  
👎  WET = **W**rite **E**verything **T**wice / **We** **E**njoy **T**yping

---
# Crash course in functions

R functions follow a general structure:

```r
my_function_name <- function(arg1, arg2) {
  final_output <- action(arg1, arg2)
  return(final_output)
}
```

Their usage:

```r
my_result <- my_function_name(arg1 = ..., arg2 = ...)
```

---
# Crash course in functions

* Let's define a function (plotting)
* And describe the arguments it takes

```r
my_function_name <- function(arg1, arg2) {
  # arg1 & arg2 = numeric vectors of identical lengths
  plot(arg1, arg2)
}
```

---
# Crash course in functions

* Call the function just written:

```r
my_function_name(1:10, 1:10 + rnorm(10))
```

*(functions are usually more complicated than this example)*

.footnote[[longer function writing tutorial](https://agstats.io/post/writing-r-functions/) on writing functions]

---
class: center, middle,  inverse

# [exercise](exercises/exercise-4.html)

---
## What to do when function errors happen

* `traceback`

```r
f <- function(x) x + 1
g <- function(x) f(x)^3
g("a")
```

```
## Error in x + 1: non-numeric argument to binary operator
```

```r
#traceback()
```

---
class: center, middle, inverse

# [mini-exercise](exercises/exercise-5.html)

---
# Trace

* sometimes you only want R to alert you when something specific happens - this is where [trace](https://stat.ethz.ch/R-manual/R-devel/library/base/htmlgi/trace.html) can be helpful
* this examines `print()` when it receives numeric input that is greater than 3.

```r
trace(print, quote(if (is.numeric(x) && x > 3) 
  cat("We have a problem: \n")), print = FALSE)
```

```
## Tracing function "print" in package "base"
```

```
## [1] "print"
```

```r
print(4)
```

```
## We have a problem: 
## [1] 4
```

Remember to turn `trace` off:

```r
untrace(print)
```

```
## Untracing function "print" in package "base"
```

---
## What to do when function errors happen

* sprinkle `print()` and/or `str()` statements in your function
(not a formal debug method, just something that works)

```r
log_hist <- function(var, color = "darkcyan", do_log=TRUE) {
  if(do_log == TRUE) {
    print("doing the log var")
    logvar = log(var)
    str(logvar)
    hist(logvar, col = color); abline(v = mean(logvar), lwd = 2, col = "red3")
  }
  else hist(var, col = color); abline(v = mean(var), lwd = 2, col = "red3")
}
```

```r
log_hist(rnorm(100), do_log=TRUE)
```

```
## [1] "doing the log var"
```

```
## Warning in log(var): NaNs produced
```

```
##  num [1:100] NaN -2.206 NaN -2.823 -0.953 ...
```

![](debugging_files/figure-html/unnamed-chunk-16-1.png)

---
## `browse()`

A more powerful way to examine objects in a function:

```r
fertil <- agdata[[2]]
colnames(fertil) <- c("country", "code", "year", 
                      "cereal_yield", "N_appl")
```

```r
my_lm_fun <- function(df, y, cty) {
  m1 = lm(y ~ year, data = df,
          subset = df[,df$country == cty])
  a1 = anova(m1)
  r2 = summary(m1)$r2
  return(list(coefficients = coef(m1),
              anova = a1, rsq = r2))
}
```

```r
my_lm_fun(df = fertil, country = "Canada")
```

```
## Error in my_lm_fun(df = fertil, country = "Canada"): unused argument (country = "Canada")
```

---
class: center, middle, inverse

# [exercise](exercises/exercise-6.html)

---
## How do you examine an error for a function you didn't write?

* `debug()`

```r
debug(stats::lm)
```

```r
test <- lm(cereal_yield ~ N_appl, data = fertil)
```

Turn if off:

```r
undebug(stats::lm)
```

---
# You can put this in your .Rprofile!

e.g.

```r
options(error = traceback)
```
or

```r
options(error = browser)
```

....and more

---
class: center, middle, inverse

## Keep Calm and Debug On

---
# Miscellaneous R tip: running R scripts in a terminal

```r
Rscript -e "Sys.Date()"
```

```r
Rscript path/to/my/script.R 
```

**Why do this?**

You might be running a long process that does not require your supervision or an interactive RStudio session. R can use *a lot* of memory just running processes, but RStudio, the graphical user interface, is also a memory hog.