Dev builds deterministic RdSAP export pipeline by isolating all side effects

A developer built a mobile export pipeline this week for UK energy assessment data, converting survey responses into RdSAP XML files with checksums and audit logs. The core design decision was making the serializer a pure function — no file system, no clock, no random IDs — so identical inputs always produce identical output. Side effects like file writes, hashing, and database logging are handled exclusively by a separate orchestrator layer through swappable dependency ports. In testing environments where native modules like Expo crypto or file system APIs are unreliable, in-memory substitutes run the full real pipeline without mocks. The developer notes this strict boundary separation also makes AI-assisted coding more reliable, since large language models perform better when schema mapping, IO, and logging are kept in clearly separated layers.