Domain-driven Design (DDD) in R. Put domain experts and developers on a level playing field, which produces software that makes perfect sense to the business, not just the coders.
ddd 
Domain-driven Design (ddd)
ddd?Sorry, most likely ddd is not for you.
ddd for?Here are a few assumptions about you:
You are an R user
You are an experienced Data Scientist
You have seen and created clunkers, that is, analytic applications
with an abundance of cruft
You have a passion for producing analytic applications of
high-quality every day
If you belong to the intersection of all the sets above, then ddd may
benefit you.
ddd is (but first, what it is not)Because ddd is an R package that caters for Data Scientists, it is
natural to think it may contain a new learning algorithm, statistical
method, a visualisation tool, or data processing toolkit. ddd offers
none of these things. Instead, ddd automates repetitive tasks that
arise during software development in R. In that sense, ddd is akin to
the usethis package.
ddd is an acronym of a software development approach called
domain-driven
design, or DDD
for short. ddd facilitates the implementation of key DDD patterns
including Entity, Value Object and Domain Services. Together these
building blocks form a domain model.
Before getting into specifics about domain-driven design, lets first
discuss the importance of software design and how it is inevitable.
When you develop software to solve a new business problem, what guides
your software design? Obviously, you strive to eliminate unintended
behaviours (bugs) from the software. Perhaps you are using tests and
even a Test-Driven Development (TDD) approach. However, guiding your
design with tests is analogous to driving a car by using the guardrails.
It keeps you safe from the bar ditches while you drive, though it has no
navigation utility.
All too often, Data Scientists produce software without thoughtful
design. On the one hand, Data Scientists are typically not Software
Engineers, nor are they taught good practices of software engineering.
On the other hand, Data Scientists often take the role of Software
Engineers when they develop analytic applications. Perhaps, the lack of
software engineering knowledge, leads Data Scientists to overlook the
significance of good design.
No design is a fallacy. This is because design still flows from the
brains of the Data Scientists through their fingertips as they wrangle
with the code. Thus, questions about whether a design is necessary or
affordable are quite beside the point: design is inevitable. The
alternative to good design is bad design, not no design at all.
You can install ddd by using:
install.packages("remotes")remotes::install_github("tidylab/ddd")
Optionally, you can make a project independent of the ddd package by
calling use_ddd(). ddd is needed for software development and has no
added value in production. use_ddd() will copy the necessary R classes
and functions into your project.
Evans, Eric. 2004. Domain-driven design: tackling complexity in the
heart of software. Addison-Wesley Professional.
Percival, Harry, and Bob Gregory. 2020. Architecture Patterns with
Python Enabling Test-Driven Development, Domain-Driven Design, and
Event-Driven Microservice. Edited by 1 Edition. O’Reilly Media, Inc.
Vernon, Vaughn. 2013. Implementing Domain-Driven Design. 1st Editio.
Addison-Wesley Professional.
Wlaschin, Scott. 2018. Domain Modeling Made Functional. Pragmatic
Bookshelf.