You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Follow-up from PR #296 review (jsteinich). The generator currently hard-fails with an actionable error when provider-function names collide after sanitization: two functions to one method name, a function landing on a reserved wrapper member (constructor, providerLocalName), two parameters to one identifier, or a provider config attribute colliding with the generated functions getter. The reviewer accepted this "to start with" but flagged it as a long-term support risk: a provider release introducing a collision turns into a deterministic cdktn get / prebuilt-regeneration failure.
Evidence the risk is currently theoretical: the RFC-04 collision sweep (cdktn-planning branch rfc04/collision-sweep) ran the exact compiled sanitizers against all 29 catalog providers + Grafana + MongoDB Atlas — 31/31 clean.
Scope
Design a deterministic dedupe (e.g. suffix rules with emitted diagnostics, mirroring uniqueClassName's approach but loudly) to replace the hard-fail so a colliding upstream release degrades to a discoverable rename instead of a generation failure.
Evaluate adopting jsii's per-language reserved-word lists proactively in the generator-side table (jsii currently handles these at compile time behind --silence-warnings reserved-word).
Context
Follow-up from PR #296 review (jsteinich). The generator currently hard-fails with an actionable error when provider-function names collide after sanitization: two functions to one method name, a function landing on a reserved wrapper member (
constructor,providerLocalName), two parameters to one identifier, or a provider config attribute colliding with the generatedfunctionsgetter. The reviewer accepted this "to start with" but flagged it as a long-term support risk: a provider release introducing a collision turns into a deterministiccdktn get/ prebuilt-regeneration failure.Evidence the risk is currently theoretical: the RFC-04 collision sweep (cdktn-planning branch
rfc04/collision-sweep) ran the exact compiled sanitizers against all 29 catalog providers + Grafana + MongoDB Atlas — 31/31 clean.Scope
uniqueClassName's approach but loudly) to replace the hard-fail so a colliding upstream release degrades to a discoverable rename instead of a generation failure.--silence-warnings reserved-word).Relations: #296, #309, cdktn-planning
rfc04/collision-sweep.