08/07/2021

Erik Kusch

Erik Kusch (erik.kusch@bio.au.dk), PhD

Student

Department of Biology

Section for Ecoinformatics & Biodiversity

Center for Biodiversity Dynamics in a

Changing World (BIOCHANGE)

Aarhus University

Erik Kusch

CODING PRACTICES

Life with R

08/07/2021 SalGo-Team - Coding Practices

08/07/2021

Erik Kusch

Code Structure

SalGo-Team - Coding Practices

Code Folding

 Braced Regions (“{…}”)

 Code Sections

Code Sections

 Title followed by 4:

 “#”

 “=”

 “-”

 Section Hierarchy determined

by preceding “#”

 Requires Rstudio Version ≥ 1.4

Header

Title

Contents

Author

Preamble

rm(list=ls()

Directories

Packages

Sourcing

Data

Download

Manipulation

Analysis

Statistical Tests/Models

Export

Results

Plots

Manipulated Data

08/07/2021

Erik Kusch

Header

SalGo-Team - Coding Practices

 Information of:

 Project Membership

 Contents

 Dependencies

 Authorship

 Can be used to track edit dates, but file versioning is better for this

Header

Title

Contents

Author

Preamble

rm(list=ls()

Directories

Packages

Sourcing

Data

Download

Manipulation

Analysis

Statistical Tests/Models

Export

Results

Plots

Manipulated Data

08/07/2021

Erik Kusch

rm(list=ls())

 Clears working directory

Assures your code is self-contained

Directories

 Avoid hard-coding directories

Use getwd() and file.path() for soft-coded directories

Packages

 Do not use library() to load packages

Automatically install packages that are needed, but not installed

Sourcing/Functionality

 Source other code documents

 Create small helper functions

Preamble

SalGo-Team - Coding Practices

Header

Title

Contents

Author

Preamble

rm(list=ls()

Directories

Packages

Sourcing

Data

Download

Manipulation

Analysis

Statistical Tests/Models

Export

Results

Plots

Manipulated Data

08/07/2021

Erik Kusch

Data

SalGo-Team - Coding Practices

Header

Title

Contents

Author

Preamble

rm(list=ls()

Directories

Packages

Sourcing

Data

Download

Manipulation

Analysis

Statistical Tests/Models

Export

Results

Plots

Manipulated Data

File Checks

 Use file.exists(…) to check if data files are present

Download

 httr::GET(…) for data download given a link

 utils::unzip(…) for extraction of .zip archives

 Do not load data via dialogue boxes!

Use code to load data

Manipulation

 Do not manipulate data outside of R

 Make all data manipulation traceable through R-Code

08/07/2021

Erik Kusch

Analysis

SalGo-Team - Coding Practices

Header

Title

Contents

Author

Preamble

rm(list=ls()

Directories

Packages

Sourcing

Data

Download

Manipulation

Analysis

Statistical Tests/Models

Export

Results

Plots

Manipulated Data

Data

 No more manipulation of data

 Subsetting is permissible

Analyses

 Variable Selection

 Analyses

 Model Comparison

 Model Evaluation

Naming

 Do not override model objects

Use unique names for model objects/outputs

08/07/2021

Erik Kusch

Export

SalGo-Team - Coding Practices

Header

Title

Contents

Author

Preamble

rm(list=ls()

Directories

Packages

Sourcing

Data

Download

Manipulation

Analysis

Statistical Tests/Models

Export

Results

Plots

Manipulated Data

Results

 Export model objects as .Rdata with save(…)

 Export model summaries as .txt with sink(…) … sink()

Plotting

 Export ggplot2 figures with ggsave(…)

 Export base plot figures with png(…) … dev.off()

Manipulated Data

 Do not override original data files!

Append suffix to data name (e.g. “_clean”)

 Do not save as excel-readable files unless strictly necessary

Use saveRDS(…) to store .rds files

08/07/2021

Erik Kusch

Naming Schemes

SalGo-Team - Coding Practices

Specificity

 What does the object contain?

Spaces

 “_”for objects and “.” for functions

Capitalisation

 Choose your style, but be consistent

Classes

 Suffix/Prefix to indicate object class

Iterators

 Avoid single-letter iterators

Logical Operators

 Do not use “T” or “F”

Object Function

Bad Name

mydata fun

Specificity

NDVI1981 Additionfunction

Spaces

NDVI_1981 Addition.function

Capitalisation

NDVI_1981 Addition.Function

Classes

NDVI1981_ras FUN.Addition

08/07/2021

Erik Kusch

Comments

SalGo-Team - Coding Practices

What?

 Comments

Start with “#”

Are not executed as code

Why?

 Convey goal of the code lines

 Revisit code

How?

 Good comments:

 Specify (what the code is doing)

 Justify (what is being done)

BAD!

GOOD!

08/07/2021

Erik Kusch

Object Assignment

 Use “<-” instead of “=“

Readability

 Spaces in function specification

 Line breaks between individual chunks of code

 Soft-wrap in RStudio to prevent scrolling side-ways

Consistency

 Develop a style and be consistent in using it

Quality of Life with R

SalGo-Team - Coding Practices

08/07/2021

Erik Kusch

Reproducibility

SalGo-Team - Coding Practices

Why Care?

 Reproducible research as gold-standard of science

 Journal Requirements

 Code sharing

 Data sharing

How?

 STOP. HARDCODING.

 Test your scripts on

different machines.

08/07/2021

Erik Kusch

Reproducibility – Cardinal Sins

SalGo-Team - Coding Practices

08/07/2021

Erik Kusch

Packages

SalGo-Team - Coding Practices

What People Do

 Use library(…) to load packages

Not everyone has the same packages installed

 Force installation via install.packages(…)

May disrupt local package versions

How To Do Better

 Identify packages that need installing and only

install those prior to loading

08/07/2021

Erik Kusch

Directories

SalGo-Team - Coding Practices

What People Do

 Hardcode working directories with setwd(…)

 Others do not have access to your hard drive

How To Do Better

 Organise projects in separate folders

 Softcode working directories:

 getwd(…) to index project folder

 file.path(…) to index sub-directories

Project Folder

Data

Raw

Manipulated

Exports

Plots

Models

08/07/2021

Erik Kusch

File Checks

SalGo-Team - Coding Practices

What People Do

 Overwrite existing data files

Can possibly break reproducibility and cement errors

 Carry out analyses whose output is already present

 Waste of processing time and power

How To Do Better

 Check whether files are present with file.exists(…)

Load/Manipulate

Load/Write

Load/Analyse

08/07/2021

Erik Kusch

Random Processes

SalGo-Team - Coding Practices

What People Do

 Sample data sets

 Partition data into test and training data sets

May induce severe bias

How To Do Better

 Set a seed to make random processes reproducible

set.seed(…)

 Sample multiple times and average out to remove bias

Bootstrapping

08/07/2021

Erik Kusch

Reproducibility - Reporting

SalGo-Team - Coding Practices

Code is as essential as your written

report.

08/07/2021

Erik Kusch

Using Rmarkdown for your research comes with a multitude of advantages:

1. Entire workflow in one program (RStudio)

2. Research and reports reproducible at the click of one button

3. Combines R functionality and LATEX formatting (if desired)

4. Consistent formatting

5. Clear presentation of code

6. Dynamic documents (you can generate various output document types)

7. Applicable for almost all document types you may desire as an output

Reporting Code

SalGo-Team - Coding Practices

08/07/2021

Erik Kusch

More Quality of Life Improvements

SalGo-Team - Coding Practices

Everything hereafter is nice-to-have or nice-

to-do, but not essential.

08/07/2021

Erik Kusch

Functions & Sourcing

SalGo-Team - Coding Practices

Big projects lend themselves well to a multi-document workflow:

Functions

Fun <- function(…){} to create a custom

function

Useful for:

 Soft-coding analyses

 Repeating code for different

input parameters

Should be:

 internally consistent

 well-documented

 easy to understand

Sourcing

Source(…) command to load/execute

R scripts

Useful for:

 Keeping code structured and

concise

 Storing extra functionality

Should:

 Use sensible file names

 Sourced functions need to be

called

08/07/2021

Erik Kusch

Progress Bar

 Updates on how code is progressing

 Especially useful when your code involves loops

Estimators

 When to expect code to finish processing

 Most useful in loop-based approaches as they need a

baseline for the estimation

Progress Bar & Estimators

SalGo-Team - Coding Practices

08/07/2021

Erik Kusch

Errors

 Stop code execution

 Report message to console

Warnings

 Do not stop code execution

 Report message to console

User-Input

 Pause code execution

 Wait for user-input in console

Errors, Warnings & User-Inputs

SalGo-Team - Coding Practices

08/07/2021

Erik Kusch

 By default, R only uses one core

Inefficient use of computational power

 Use parallel processing to make use of remaining cores

 Particularly powerful when paired with eval(parse(text=“…”))

Parallel Processing

SalGo-Team - Coding Practices

08/07/2021

Erik Kusch

Versioning

SalGo-Team - Coding Practices

Use GitHub to:

 Track code and development

You may revert to previous versions of code if you broke

something

 Share code with others

You can make available R functionality and packages on GitHub

 Collaborate with others

Pull requests and the issue feature allow for collaborative code

development