Adv-R-Programming · Epsian · Apr 10, 2023 · Apr 10, 2023 · Apr 10, 2023 · Apr 10, 2023
diff --git a/content/class_worksheets/29_tests/29_tests.qmd b/content/class_worksheets/29_tests/29_tests.qmd
@@ -14,7 +14,7 @@ editor_options:
 ---
 
 ```{r, include=FALSE}
-answers = FALSE
+answers = TRUE
 ```
 
 ## Overview

diff --git a/content/class_worksheets/29_tests/_index.en.md b/content/class_worksheets/29_tests/_index.en.md
@@ -21,7 +21,7 @@ editor_options:
 
 ## Overview
 
-Tests are your early warning system is something starts to go awry in
+Tests are your early warning system if something starts to go awry in
 your project. The more complex your project, the more important it
 becomes to have a robust set of tests. This is especially true if you
 have collaborators.
@@ -144,6 +144,21 @@ your test, make sure it works by using `devtools::test()`.
 
 </div>
 
+<div class="answer">
+
+test_that(“Are there any commas in the results?”, {
+
+test_vec = c(“Mon, Tues, Wed”) possible = c(“Mon”, “Tues”, “Wed”,
+“Thurs”, “Fri”)
+
+out = comma_split(test_vec, possible)
+
+expect_false(all(grepl(“,”, out)))
+
+})
+
+</div>
+
 ### Check your Progress
 
 Once you have made your test and it works as intended, try running a

diff --git a/content/class_worksheets/30_benchmark/30_benchmark.qmd b/content/class_worksheets/30_benchmark/30_benchmark.qmd
@@ -0,0 +1,177 @@
+---
+pre: <b>4/12. </b>
+title: "Benchmarks"
+weight: 30
+summary: "Gotta go fast."
+format:
+    gfm:
+      toc: true
+      output-file: "_index.en.md"
+      reference-links: true
+      code-link: true
+editor_options: 
+  chunk_output_type: console
+---
+
+```{r, include=FALSE}
+answers = FALSE
+```
+
+## Overview
+
+For our worksheet today, we are going to be tinkering with `comma_split()` again, this time doing some benchmarking. Our goal by the end is to try and optimize the function to be a tad faster. 
+
+
+
+### Our Test Case
+
+For the worksheet today, you will need to load in the class survey data, you can do so by running the following:
+
+```{r, eval=FALSE, echo=TRUE}
+survey = readRDS(url("https://github.com/Adv-R-Programming/Adv-R-Reader/raw/main/class_survey.rds"))
+```
+
+You will also need to have a copy of the `comma_split()` function. No need to work inside a package today.
+
+```{r, eval=FALSE, echo=TRUE}
+comma_split = function(vector_to_split, possible_columns){
+
+  # make a base dataframe with rows for each of our cases.
+  output = data.frame(
+    "id" = 1:length(vector_to_split)
+    )
+
+  # iterate through all options and create a column with NAs for it
+  for(option in possible_columns){
+
+    # make a new column with a character version of each possible option.
+    output[, as.character(option)] = NA
+
+  }
+
+  # fill output df
+  for(option in possible_columns){
+
+    # fill dataframe iterativly.
+    output[ , option] = grepl(option, vector_to_split, ignore.case = TRUE)
+
+  }
+
+  # clear all know options
+  for(option in possible_columns){
+
+    # remove all known options
+    vector_to_split = gsub(pattern = option, vector_to_split, replacement = "", ignore.case = TRUE)
+
+  }
+
+  # clear commas and whitespace
+  vector_to_split = gsub(pattern = ",", vector_to_split, replacement = "", ignore.case = TRUE)
+  vector_to_split = trimws(vector_to_split)
+
+  # Fill in "other"
+  output$other = vector_to_split
+  # Turn blanks into NAs
+  output[output$other == "" & !is.na(output$other), "other"] = NA
+
+  # return output
+  return(output)
+}
+```
+
+### Making a Plan & Baseline
+
+The `comma_split()` as it is now is alright; room for improvement, but not bad. For example, you can see that we pre-allocate space for our output rows in the dataframe by making an ID column of the appropriate length, so we aren't appending anything. However, there are a few ways to improve things.
+
+Before we go about making improvements however, we need to set a baseline. To do so, we are going to use `bench::mark()` to get a sense of how long our code takes to run now.
+
+::: {.question}
+Use `bench::mark()` to run `comma_split()` on a column from the class survey. Run at least 500 iterations using the `iterations` argument.
+:::
+
+```{r, include=answers, results='asis', echo=FALSE}
+cat('
+::: {.answer}
+
+bench::mark({
+
+  comma_split(survey$pets, possible_columns = c("dog", "cat", "fish", "bird", "reptile", "rock", "none"))
+
+}, iterations = 500)
+
+:::
+')
+```
+
+### Make Some Improvements
+
+No that we have an idea of how long the code takes now, we are going to start tinkering. Refer to the list of hints below (and in lecture slides) for ideas of things to focus on.
+
+Rules of Thumb for efficient code:
+
+* Don't repeat yourself!
+* Subset data to only what you will use before running analyses (no need to analyse data that will be tossed out)
+* Pre-specify anything you can (e.g. manually set levels in factors, column names when reading in large data, what method you want for generic functions)
+* Pre-allocate for outputs (never amend to lists/dataframes)
+* Always work in a vector if possible (don't work on individual values at a time)
+
+::: {.question}
+Create a new version of `comma_split()` that attempts to speed it up, called `comma_split_2()`. There are a few ways you can attempt this. Once you have created your improved function, benchmark it to see if your changes had the intended effect.
+:::
+
+```{r, include=answers, results='asis', echo=FALSE}
+cat('
+::: {.answer}
+
+# here I left a for loop but still cut the time in half!
+comma_split_2 = function(vector_to_split, possible_columns){
+
+  # make list of results
+  splits = lapply(possible_columns, FUN = function(option, vector_to_split){
+    return(grepl(option, vector_to_split, ignore.case = TRUE))
+  }, vector_to_split)
+
+  # add names
+  names(splits) = possible_columns
+
+  # clear all know options
+  for(option in possible_columns){
+
+    # remove all known options
+    vector_to_split = gsub(pattern = option, vector_to_split, replacement = "", ignore.case = TRUE)
+
+  }
+
+  # clear commas and whitespace
+  vector_to_split = gsub(pattern = ",", vector_to_split, replacement = "", ignore.case = TRUE)
+  vector_to_split = trimws(vector_to_split)
+  vector_to_split[vector_to_split == ""] = NA
+
+  # Fill in "other"
+  splits[["other"]] = vector_to_split
+
+  # turn to df
+  class(splits) = "data.frame"
+  attr(splits, "row.names") <- .set_row_names(length(test[[1]]))
+
+  # return output
+  return(splits)
+}
+
+# do benchmarks
+
+bench::mark({
+  comma_split(survey$pets, possible_columns = c("dog", "cat", "fish", "bird", "reptile", "rock", "none"))
+}, iterations = 500, time_unit = "ms")
+
+
+bench::mark({
+  comma_split_2(survey$pets, possible_columns = c("dog", "cat", "fish", "bird", "reptile", "rock", "none"))
+}, iterations = 500, time_unit = "ms")
+
+:::
+')
+```
+
+
+
diff --git a/content/class_worksheets/30_benchmark/_index.en.md b/content/class_worksheets/30_benchmark/_index.en.md
@@ -0,0 +1,134 @@
+---
+pre: <b>4/12. </b>
+title: "Benchmarks"
+weight: 30
+summary: "Gotta go fast."
+format:
+    gfm:
+      toc: true
+      output-file: "_index.en.md"
+      reference-links: true
+      code-link: true
+editor_options: 
+  chunk_output_type: console
+---
+
+- [Overview][]
+  - [Our Test Case][]
+  - [Making a Plan & Baseline][]
+  - [Make Some Improvements][]
+
+## Overview
+
+For our worksheet today, we are going to be tinkering with
+`comma_split()` again, this time doing some benchmarking. Our goal by
+the end is to try and optimize the function to be a tad faster.
+
+### Our Test Case
+
+For the worksheet today, you will need to load in the class survey data,
+you can do so by running the following:
+
+``` r
+survey = readRDS(url("https://github.com/Adv-R-Programming/Adv-R-Reader/raw/main/class_survey.rds"))
+```
+
+You will also need to have a copy of the `comma_split()` function. No
+need to work inside a package today.
+
+``` r
+comma_split = function(vector_to_split, possible_columns){
+
+  # make a base dataframe with rows for each of our cases.
+  output = data.frame(
+    "id" = 1:length(vector_to_split)
+    )
+
+  # iterate through all options and create a column with NAs for it
+  for(option in possible_columns){
+
+    # make a new column with a character version of each possible option.
+    output[, as.character(option)] = NA
+
+  }
+
+  # fill output df
+  for(option in possible_columns){
+
+    # fill dataframe iterativly.
+    output[ , option] = grepl(option, vector_to_split, ignore.case = TRUE)
+
+  }
+
+  # clear all know options
+  for(option in possible_columns){
+
+    # remove all known options
+    vector_to_split = gsub(pattern = option, vector_to_split, replacement = "", ignore.case = TRUE)
+
+  }
+
+  # clear commas and whitespace
+  vector_to_split = gsub(pattern = ",", vector_to_split, replacement = "", ignore.case = TRUE)
+  vector_to_split = trimws(vector_to_split)
+
+  # Fill in "other"
+  output$other = vector_to_split
+  # Turn blanks into NAs
+  output[output$other == "" & !is.na(output$other), "other"] = NA
+
+  # return output
+  return(output)
+}
+```
+
+### Making a Plan & Baseline
+
+The `comma_split()` as it is now is alright; room for improvement, but
+not bad. For example, you can see that we pre-allocate space for our
+output rows in the dataframe by making an ID column of the appropriate
+length, so we aren’t appending anything. However, there are a few ways
+to improve things.
+
+Before we go about making improvements however, we need to set a
+baseline. To do so, we are going to use `bench::mark()` to get a sense
+of how long our code takes to run now.
+
+<div class="question">
+
+Use `bench::mark()` to run `comma_split()` on a column from the class
+survey. Run at least 500 iterations using the `iterations` argument.
+
+</div>
+
+### Make Some Improvements
+
+No that we have an idea of how long the code takes now, we are going to
+start tinkering. Refer to the list of hints below (and in lecture
+slides) for ideas of things to focus on.
+
+Rules of Thumb for efficient code:
+
+- Don’t repeat yourself!
+- Subset data to only what you will use before running analyses (no need
+  to analyse data that will be tossed out)
+- Pre-specify anything you can (e.g. manually set levels in factors,
+  column names when reading in large data, what method you want for
+  generic functions)
+- Pre-allocate for outputs (never amend to lists/dataframes)
+- Always work in a vector if possible (don’t work on individual values
+  at a time)
+
+<div class="question">
+
+Create a new version of `comma_split()` that attempts to speed it up,
+called `comma_split_2()`. There are a few ways you can attempt this.
+Once you have created your improved function, benchmark it to see if
+your changes had the intended effect.
+
+</div>
+
+  [Overview]: #overview
+  [Our Test Case]: #our-test-case
+  [Making a Plan & Baseline]: #making-a-plan-baseline
+  [Make Some Improvements]: #make-some-improvements