Problem Decomposition

Approaches to solving problems

• Real computational tasks are complicated.

• To accomplish them effectively you need to think before you code.

• Decompose a coding problem:

1. Understand the problem.
   • Identify inputs and outputs.
     • Input = Portal surveys csv
     • Output = plot of yearly abundance for large and small rodents

    # Overall goal: visualize how many large (>50g) and 
    # small (<50g) mammals are collected each year
   

2. Break the problem down into a few pieces.
   • Bullet list on paper or comments in script.

    # Subtasks
        # Import data
        # Get column of small and large categories
        # Count number of each per year
        # Plot this
   

3. Break those pieces into codeable chunks.

    # TODO: function that returns size class from weight
   

4. Code one chunk at a time.

Coding one chunk

• Write function to return size classes

get_size_class()

• Start with if statement
• Pseudocode
  • Write code in English
  • “If weight value is bigger than 50, return large; otherwise return small”

if(weight > 50){
  size_class <- "large"
} else {
  size_class <- "small"
}

• Test code using values with known output

weight <- 10

• Turn into function

get_size_class <- function(weight){
  if(weight > 50){
    size_class <- "large"
  } else {
    size_class <- "small"
  }
}
get_size_class(100)

• Why doesn’t this work?

get_size_class <- function(weight){
  if(weight > 50){
    size_class <- "large"
  } else {
    size_class <- "small"
  }
  return(size_class)
}

• Generalize function
• Easier to change values later
• Always test after changing

get_size_class <- function(weight, threshold){
  if(weight > threshold){
    size_class <- "large"
  } else {
    size_class <- "small"
  }
  return(size_class)
}
get_size_class(100, 50)
get_size_class(100, 150)

• Make one change at a time

• Decompose a code chunk:

  1. Think about it.
  2. Write it.
  3. Test it ( on it’s own ).
  4. Fix any problems.

Rest of code in decomposition-example.R.