The speaker shares his background working in small, autonomous teams and how that experience shaped his approach to software configuration.
The talk focuses on solving configuration for internal systems by leveraging Python's internals rather than focusing on performance or complex code.
Configuration is defined not as a variable or a true constant, but as a value derived from the environment that remains fixed after startup.
Classic configuration files are often unnecessary for internal Python projects, as demonstrated by tools like Sphinx, Plotly, and Nox.
Code at the module level executes during import, and the execution path differs when running a file as a script versus importing it as a module.
Most "constants" in Python, including system attributes, can be modified at runtime, highlighting the need for patterns that enforce immutability.
A singleton class using the @property decorator provides a sweet spot for creating dynamic, read-only configuration values that are easily testable.
The speaker answers audience questions about version handling in a monorepo, why configuration shouldn't change at runtime, and potential pitfalls of these patterns.
ROSEN Technology and Research Center GmbH
Lingen (Ems), Germany
25:29Diana Gastrin
57:46Adrian Schmitt
24:59Daniel Hepper
58:28Meredydd Luff
27:11Eli Holderness
37:37Eli Holderness
33:44Fabian Schindler
39:04Bobur Umurzokov
Jobs that call for the skills explored in this talk.
Cover Genius
Amsterdam, Netherlands
Cover Genius
Amsterdam, Netherlands
Schwarz Corporate Solutions
Barcelona, Spain
