Talk: 15:05-15:50 (English)
Delta encodings help separate business logic from database operations
In this talk, I describe delta encodings and how they helped me refactor Haskell code in which business logic was mixed with database operations.
In my case, the business logic is about managing funds in a blockchain wallet. Each block of the chain is mapped to a delta of the wallet state, which is then written to a database file. Before using delta encodings, the business logic was entangled with IO, making it hard to test and reason about.
In particular, I will describe
-
delta encodings, which we define as data types that describe changes to other data types. Delta encodings allow us to separate the description of a change (business logic) from its execution (database operation). Multiple delta encodings may apply to the same data type.
-
DBVar, which we define to be mutable variable that holds a Haskell value (business logic), but also ensures that the value is stored on disk (database operation). Delta encodings ensure that writing to disk remains performant.
The combination of these techniques is particularly useful in situations where the database is predominantly written to, and where its contents still fits into memory.
Heinrich Apfelmus
Heinrich Apfelmus has been programming in Haskell for more than 15 years, and develops several open source libraries. He has studied mathematics and completed a PhD in theoretical physics. As a student, he has participated in international programming competitions.
- Slides
- apfelmus.pdf
