Emit static Effect schema declarations

[patroza]

Summary

Adds declaration-emitter support for Effect schema model facades so .d.ts emit can materialize the same static shape we currently generate in scanner source:

• materializes namespace X { interface Encoded ... } from S.StructNestedEncoded<typeof X>
• synthesizes interface X, namespace X.Make, X.DecodingServices, and X.EncodingServices
• rewrites emitted schema bases to S.OpaqueFacade<X, X.Encoded, X.Make, X.DecodingServices, X.EncodingServices, {}>
• preserves class static members by intersecting the facade with fields, mapFields, copy, to, and from
• emits an intermediate __X_base / __X wrapper for model classes with own members, matching native source codegen shape

This intentionally lives in declaration emit and checker/node-builder plumbing only, so it can be carried as a patch onto Effect's TypeScript compiler fork later.

Validation

• npm run build:compiler
• scanner comparison with the forked compiler:
  • no-static main + compatibility patch: Aggregate Instantiations: 13,917,040
  • static-native branch + forked compiler: Aggregate Instantiations: 13,840,894
  • no-static main + stock compiler: Aggregate Instantiations: 17,871,742

The d.ts-emitter path recovered roughly 98% of the static-native source-codegen instantiation benefit in the scanner comparison.

Notes / Gotchas

• Existing native __X_base helper classes are not reinterpreted. Rewriting those to namespace service aliases can make old generated type Make = X.Make aliases self-reference and collapse constructor input to never.
• Textual d.ts parity is close but not byte-for-byte: ordering, import qualification, and boolean vs false | true can still differ.
• This targets the effect-app/TypeScript fork only; do not retarget to the official Microsoft repository.

[patroza]

Update: the emitter now also handles direct bare schema models like class X extends S.Opaque<X>()(...) {} when no namespace X existed before.

For those classes it now synthesizes:

• top-level interface X from the schema Type
• namespace X
• X.Encoded
• X.Make
• X.DecodingServices
• X.EncodingServices
• S.OpaqueFacade<X, X.Encoded, X.Make, X.DecodingServices, X.EncodingServices, {}> base rewrite

It also tightens recognition to schema aliases S.Opaque / Schema.Opaque, so Context.Opaque service tags are not swept into this path.

Validation:

• npm run build:compiler ✅
• tiny declaration fixture for bare S.Opaque<DirectBare>()(...) emitted the full OpaqueFacade + interface + namespace shape ✅

[patroza]

Updated again after measuring direct bare S.Opaque<X>()(...) support.

Latest scanner comparison with this compiler:

• main/no-static + stock TypeScript: 17,871,742
• main/no-static + forked emitter before bare Opaque: 13,917,040
• main/no-static + forked emitter with bare Opaque: 12,128,418
• static-native branch + forked emitter with bare Opaque: 11,724,286

Bare Opaque support adds another 1,788,622 instantiation reduction on no-static main vs the previous emitter pass. Current no-static compiler emit is 404,132 above static-native, so it recovers about 93% of the current stock-to-static-native benefit.

Also fixed synthesized interface/namespace export-ness so non-exported local schema classes merge correctly inside their source module.

[patroza]

Summary

Adds declaration-emitter support for Effect schema model facades so .d.ts emit can materialize the same static shape we currently generate in scanner source:

• materializes namespace X { interface Encoded ... } from S.StructNestedEncoded<typeof X>
• synthesizes interface X, namespace X.Make, X.DecodingServices, and X.EncodingServices
• rewrites emitted schema bases to S.OpaqueFacade<X, X.Encoded, X.Make, X.DecodingServices, X.EncodingServices, {}>
• preserves class static members by intersecting the facade with fields, mapFields, copy, to, and from
• emits an intermediate __X_base / __X wrapper for model classes with own members, matching native source codegen shape

This intentionally lives in declaration emit and checker/node-builder plumbing only, so it can be carried as a patch onto Effect's TypeScript compiler fork later.

Validation

• npm run build:compiler
• scanner comparison with the forked compiler:
  • no-static main + compatibility patch: Aggregate Instantiations: 13,917,040
  • static-native branch + forked compiler: Aggregate Instantiations: 13,840,894
  • no-static main + stock compiler: Aggregate Instantiations: 17,871,742

The d.ts-emitter path recovered roughly 98% of the static-native source-codegen instantiation benefit in the scanner comparison.

Notes / Gotchas

• Existing native __X_base helper classes are not reinterpreted. Rewriting those to namespace service aliases can make old generated type Make = X.Make aliases self-reference and collapse constructor input to never.
• Textual d.ts parity is close but not byte-for-byte: ordering, import qualification, and boolean vs false | true can still differ.
• This targets the effect-app/TypeScript fork only; do not retarget to the official Microsoft repository.

[patroza]

Update: the emitter now also handles direct bare schema models like class X extends S.Opaque<X>()(...) {} when no namespace X existed before.

For those classes it now synthesizes:

• top-level interface X from the schema Type
• namespace X
• X.Encoded
• X.Make
• X.DecodingServices
• X.EncodingServices
• S.OpaqueFacade<X, X.Encoded, X.Make, X.DecodingServices, X.EncodingServices, {}> base rewrite

It also tightens recognition to schema aliases S.Opaque / Schema.Opaque, so Context.Opaque service tags are not swept into this path.

Validation:

• npm run build:compiler ✅
• tiny declaration fixture for bare S.Opaque<DirectBare>()(...) emitted the full OpaqueFacade + interface + namespace shape ✅

[patroza]

Updated again after measuring direct bare S.Opaque<X>()(...) support.

Latest scanner comparison with this compiler:

• main/no-static + stock TypeScript: 17,871,742
• main/no-static + forked emitter before bare Opaque: 13,917,040
• main/no-static + forked emitter with bare Opaque: 12,128,418
• static-native branch + forked emitter with bare Opaque: 11,724,286

Bare Opaque support adds another 1,788,622 instantiation reduction on no-static main vs the previous emitter pass. Current no-static compiler emit is 404,132 above static-native, so it recovers about 93% of the current stock-to-static-native benefit.

Also fixed synthesized interface/namespace export-ness so non-exported local schema classes merge correctly inside their source module.

[patroza]

Summary

Adds declaration-emitter support for Effect schema model facades so .d.ts emit can materialize the same static shape we currently generate in scanner source:

• materializes namespace X { interface Encoded ... } from S.StructNestedEncoded<typeof X>
• synthesizes interface X, namespace X.Make, X.DecodingServices, and X.EncodingServices
• rewrites emitted schema bases to S.OpaqueFacade<X, X.Encoded, X.Make, X.DecodingServices, X.EncodingServices, {}>
• preserves class static members by intersecting the facade with fields, mapFields, copy, to, and from
• emits an intermediate __X_base / __X wrapper for model classes with own members, matching native source codegen shape

This intentionally lives in declaration emit and checker/node-builder plumbing only, so it can be carried as a patch onto Effect's TypeScript compiler fork later.

Validation

• npm run build:compiler
• scanner comparison with the forked compiler:
  • no-static main + compatibility patch: Aggregate Instantiations: 13,917,040
  • static-native branch + forked compiler: Aggregate Instantiations: 13,840,894
  • no-static main + stock compiler: Aggregate Instantiations: 17,871,742

The d.ts-emitter path recovered roughly 98% of the static-native source-codegen instantiation benefit in the scanner comparison.

Notes / Gotchas

• Existing native __X_base helper classes are not reinterpreted. Rewriting those to namespace service aliases can make old generated type Make = X.Make aliases self-reference and collapse constructor input to never.
• Textual d.ts parity is close but not byte-for-byte: ordering, import qualification, and boolean vs false | true can still differ.
• This targets the effect-app/TypeScript fork only; do not retarget to the official Microsoft repository.

[patroza]

Update: the emitter now also handles direct bare schema models like class X extends S.Opaque<X>()(...) {} when no namespace X existed before.

For those classes it now synthesizes:

• top-level interface X from the schema Type
• namespace X
• X.Encoded
• X.Make
• X.DecodingServices
• X.EncodingServices
• S.OpaqueFacade<X, X.Encoded, X.Make, X.DecodingServices, X.EncodingServices, {}> base rewrite

It also tightens recognition to schema aliases S.Opaque / Schema.Opaque, so Context.Opaque service tags are not swept into this path.

Validation:

• npm run build:compiler ✅
• tiny declaration fixture for bare S.Opaque<DirectBare>()(...) emitted the full OpaqueFacade + interface + namespace shape ✅

[patroza]

Updated again after measuring direct bare S.Opaque<X>()(...) support.

Latest scanner comparison with this compiler:

• main/no-static + stock TypeScript: 17,871,742
• main/no-static + forked emitter before bare Opaque: 13,917,040
• main/no-static + forked emitter with bare Opaque: 12,128,418
• static-native branch + forked emitter with bare Opaque: 11,724,286

Bare Opaque support adds another 1,788,622 instantiation reduction on no-static main vs the previous emitter pass. Current no-static compiler emit is 404,132 above static-native, so it recovers about 93% of the current stock-to-static-native benefit.

Also fixed synthesized interface/namespace export-ness so non-exported local schema classes merge correctly inside their source module.

[patroza]

Summary

Adds declaration-emitter support for Effect schema model facades so .d.ts emit can materialize the same static shape we currently generate in scanner source:

• materializes namespace X { interface Encoded ... } from S.StructNestedEncoded<typeof X>
• synthesizes interface X, namespace X.Make, X.DecodingServices, and X.EncodingServices
• rewrites emitted schema bases to S.OpaqueFacade<X, X.Encoded, X.Make, X.DecodingServices, X.EncodingServices, {}>
• preserves class static members by intersecting the facade with fields, mapFields, copy, to, and from
• emits an intermediate __X_base / __X wrapper for model classes with own members, matching native source codegen shape

This intentionally lives in declaration emit and checker/node-builder plumbing only, so it can be carried as a patch onto Effect's TypeScript compiler fork later.

Validation

• npm run build:compiler
• scanner comparison with the forked compiler:
  • no-static main + compatibility patch: Aggregate Instantiations: 13,917,040
  • static-native branch + forked compiler: Aggregate Instantiations: 13,840,894
  • no-static main + stock compiler: Aggregate Instantiations: 17,871,742

The d.ts-emitter path recovered roughly 98% of the static-native source-codegen instantiation benefit in the scanner comparison.

Notes / Gotchas

• Existing native __X_base helper classes are not reinterpreted. Rewriting those to namespace service aliases can make old generated type Make = X.Make aliases self-reference and collapse constructor input to never.
• Textual d.ts parity is close but not byte-for-byte: ordering, import qualification, and boolean vs false | true can still differ.
• This targets the effect-app/TypeScript fork only; do not retarget to the official Microsoft repository.

[patroza]

Update: the emitter now also handles direct bare schema models like class X extends S.Opaque<X>()(...) {} when no namespace X existed before.

For those classes it now synthesizes:

• top-level interface X from the schema Type
• namespace X
• X.Encoded
• X.Make
• X.DecodingServices
• X.EncodingServices
• S.OpaqueFacade<X, X.Encoded, X.Make, X.DecodingServices, X.EncodingServices, {}> base rewrite

It also tightens recognition to schema aliases S.Opaque / Schema.Opaque, so Context.Opaque service tags are not swept into this path.

Validation:

• npm run build:compiler ✅
• tiny declaration fixture for bare S.Opaque<DirectBare>()(...) emitted the full OpaqueFacade + interface + namespace shape ✅

[patroza]

Updated again after measuring direct bare S.Opaque<X>()(...) support.

Latest scanner comparison with this compiler:

• main/no-static + stock TypeScript: 17,871,742
• main/no-static + forked emitter before bare Opaque: 13,917,040
• main/no-static + forked emitter with bare Opaque: 12,128,418
• static-native branch + forked emitter with bare Opaque: 11,724,286

Bare Opaque support adds another 1,788,622 instantiation reduction on no-static main vs the previous emitter pass. Current no-static compiler emit is 404,132 above static-native, so it recovers about 93% of the current stock-to-static-native benefit.

Also fixed synthesized interface/namespace export-ness so non-exported local schema classes merge correctly inside their source module.

[patroza]

Summary

Adds declaration-emitter support for Effect schema model facades so .d.ts emit can materialize the same static shape we currently generate in scanner source:

• materializes namespace X { interface Encoded ... } from S.StructNestedEncoded<typeof X>
• synthesizes interface X, namespace X.Make, X.DecodingServices, and X.EncodingServices
• rewrites emitted schema bases to S.OpaqueFacade<X, X.Encoded, X.Make, X.DecodingServices, X.EncodingServices, {}>
• preserves class static members by intersecting the facade with fields, mapFields, copy, to, and from
• emits an intermediate __X_base / __X wrapper for model classes with own members, matching native source codegen shape

This intentionally lives in declaration emit and checker/node-builder plumbing only, so it can be carried as a patch onto Effect's TypeScript compiler fork later.

Validation

• npm run build:compiler
• scanner comparison with the forked compiler:
  • no-static main + compatibility patch: Aggregate Instantiations: 13,917,040
  • static-native branch + forked compiler: Aggregate Instantiations: 13,840,894
  • no-static main + stock compiler: Aggregate Instantiations: 17,871,742

The d.ts-emitter path recovered roughly 98% of the static-native source-codegen instantiation benefit in the scanner comparison.

Notes / Gotchas

• Existing native __X_base helper classes are not reinterpreted. Rewriting those to namespace service aliases can make old generated type Make = X.Make aliases self-reference and collapse constructor input to never.
• Textual d.ts parity is close but not byte-for-byte: ordering, import qualification, and boolean vs false | true can still differ.
• This targets the effect-app/TypeScript fork only; do not retarget to the official Microsoft repository.

[patroza]

Update: the emitter now also handles direct bare schema models like class X extends S.Opaque<X>()(...) {} when no namespace X existed before.

For those classes it now synthesizes:

• top-level interface X from the schema Type
• namespace X
• X.Encoded
• X.Make
• X.DecodingServices
• X.EncodingServices
• S.OpaqueFacade<X, X.Encoded, X.Make, X.DecodingServices, X.EncodingServices, {}> base rewrite

It also tightens recognition to schema aliases S.Opaque / Schema.Opaque, so Context.Opaque service tags are not swept into this path.

Validation:

• npm run build:compiler ✅
• tiny declaration fixture for bare S.Opaque<DirectBare>()(...) emitted the full OpaqueFacade + interface + namespace shape ✅

[patroza]

Updated again after measuring direct bare S.Opaque<X>()(...) support.

Latest scanner comparison with this compiler:

• main/no-static + stock TypeScript: 17,871,742
• main/no-static + forked emitter before bare Opaque: 13,917,040
• main/no-static + forked emitter with bare Opaque: 12,128,418
• static-native branch + forked emitter with bare Opaque: 11,724,286

Bare Opaque support adds another 1,788,622 instantiation reduction on no-static main vs the previous emitter pass. Current no-static compiler emit is 404,132 above static-native, so it recovers about 93% of the current stock-to-static-native benefit.

Also fixed synthesized interface/namespace export-ness so non-exported local schema classes merge correctly inside their source module.

[patroza]

Summary

Adds declaration-emitter support for Effect schema model facades so .d.ts emit can materialize the same static shape we currently generate in scanner source:

• materializes namespace X { interface Encoded ... } from S.StructNestedEncoded<typeof X>
• synthesizes interface X, namespace X.Make, X.DecodingServices, and X.EncodingServices
• rewrites emitted schema bases to S.OpaqueFacade<X, X.Encoded, X.Make, X.DecodingServices, X.EncodingServices, {}>
• preserves class static members by intersecting the facade with fields, mapFields, copy, to, and from
• emits an intermediate __X_base / __X wrapper for model classes with own members, matching native source codegen shape

This intentionally lives in declaration emit and checker/node-builder plumbing only, so it can be carried as a patch onto Effect's TypeScript compiler fork later.

Validation

• npm run build:compiler
• scanner comparison with the forked compiler:
  • no-static main + compatibility patch: Aggregate Instantiations: 13,917,040
  • static-native branch + forked compiler: Aggregate Instantiations: 13,840,894
  • no-static main + stock compiler: Aggregate Instantiations: 17,871,742

The d.ts-emitter path recovered roughly 98% of the static-native source-codegen instantiation benefit in the scanner comparison.

Notes / Gotchas

• Existing native __X_base helper classes are not reinterpreted. Rewriting those to namespace service aliases can make old generated type Make = X.Make aliases self-reference and collapse constructor input to never.
• Textual d.ts parity is close but not byte-for-byte: ordering, import qualification, and boolean vs false | true can still differ.
• This targets the effect-app/TypeScript fork only; do not retarget to the official Microsoft repository.

[patroza]

Update: the emitter now also handles direct bare schema models like class X extends S.Opaque<X>()(...) {} when no namespace X existed before.

For those classes it now synthesizes:

• top-level interface X from the schema Type
• namespace X
• X.Encoded
• X.Make
• X.DecodingServices
• X.EncodingServices
• S.OpaqueFacade<X, X.Encoded, X.Make, X.DecodingServices, X.EncodingServices, {}> base rewrite

It also tightens recognition to schema aliases S.Opaque / Schema.Opaque, so Context.Opaque service tags are not swept into this path.

Validation:

• npm run build:compiler ✅
• tiny declaration fixture for bare S.Opaque<DirectBare>()(...) emitted the full OpaqueFacade + interface + namespace shape ✅

[patroza]

Updated again after measuring direct bare S.Opaque<X>()(...) support.

Latest scanner comparison with this compiler:

• main/no-static + stock TypeScript: 17,871,742
• main/no-static + forked emitter before bare Opaque: 13,917,040
• main/no-static + forked emitter with bare Opaque: 12,128,418
• static-native branch + forked emitter with bare Opaque: 11,724,286

Bare Opaque support adds another 1,788,622 instantiation reduction on no-static main vs the previous emitter pass. Current no-static compiler emit is 404,132 above static-native, so it recovers about 93% of the current stock-to-static-native benefit.

Also fixed synthesized interface/namespace export-ness so non-exported local schema classes merge correctly inside their source module.

[patroza]

Summary

Adds declaration-emitter support for Effect schema model facades so .d.ts emit can materialize the same static shape we currently generate in scanner source:

• materializes namespace X { interface Encoded ... } from S.StructNestedEncoded<typeof X>
• synthesizes interface X, namespace X.Make, X.DecodingServices, and X.EncodingServices
• rewrites emitted schema bases to S.OpaqueFacade<X, X.Encoded, X.Make, X.DecodingServices, X.EncodingServices, {}>
• preserves class static members by intersecting the facade with fields, mapFields, copy, to, and from
• emits an intermediate __X_base / __X wrapper for model classes with own members, matching native source codegen shape

This intentionally lives in declaration emit and checker/node-builder plumbing only, so it can be carried as a patch onto Effect's TypeScript compiler fork later.

Validation

• npm run build:compiler
• scanner comparison with the forked compiler:
  • no-static main + compatibility patch: Aggregate Instantiations: 13,917,040
  • static-native branch + forked compiler: Aggregate Instantiations: 13,840,894
  • no-static main + stock compiler: Aggregate Instantiations: 17,871,742

The d.ts-emitter path recovered roughly 98% of the static-native source-codegen instantiation benefit in the scanner comparison.

Notes / Gotchas

• Existing native __X_base helper classes are not reinterpreted. Rewriting those to namespace service aliases can make old generated type Make = X.Make aliases self-reference and collapse constructor input to never.
• Textual d.ts parity is close but not byte-for-byte: ordering, import qualification, and boolean vs false | true can still differ.
• This targets the effect-app/TypeScript fork only; do not retarget to the official Microsoft repository.

[patroza]

Update: the emitter now also handles direct bare schema models like class X extends S.Opaque<X>()(...) {} when no namespace X existed before.

For those classes it now synthesizes:

• top-level interface X from the schema Type
• namespace X
• X.Encoded
• X.Make
• X.DecodingServices
• X.EncodingServices
• S.OpaqueFacade<X, X.Encoded, X.Make, X.DecodingServices, X.EncodingServices, {}> base rewrite

It also tightens recognition to schema aliases S.Opaque / Schema.Opaque, so Context.Opaque service tags are not swept into this path.

Validation:

• npm run build:compiler ✅
• tiny declaration fixture for bare S.Opaque<DirectBare>()(...) emitted the full OpaqueFacade + interface + namespace shape ✅

[patroza]

Updated again after measuring direct bare S.Opaque<X>()(...) support.

Latest scanner comparison with this compiler:

• main/no-static + stock TypeScript: 17,871,742
• main/no-static + forked emitter before bare Opaque: 13,917,040
• main/no-static + forked emitter with bare Opaque: 12,128,418
• static-native branch + forked emitter with bare Opaque: 11,724,286

Bare Opaque support adds another 1,788,622 instantiation reduction on no-static main vs the previous emitter pass. Current no-static compiler emit is 404,132 above static-native, so it recovers about 93% of the current stock-to-static-native benefit.

Also fixed synthesized interface/namespace export-ness so non-exported local schema classes merge correctly inside their source module.

[patroza]

Summary

Adds declaration-emitter support for Effect schema model facades so .d.ts emit can materialize the same static shape we currently generate in scanner source:

• materializes namespace X { interface Encoded ... } from S.StructNestedEncoded<typeof X>
• synthesizes interface X, namespace X.Make, X.DecodingServices, and X.EncodingServices
• rewrites emitted schema bases to S.OpaqueFacade<X, X.Encoded, X.Make, X.DecodingServices, X.EncodingServices, {}>
• preserves class static members by intersecting the facade with fields, mapFields, copy, to, and from
• emits an intermediate __X_base / __X wrapper for model classes with own members, matching native source codegen shape

This intentionally lives in declaration emit and checker/node-builder plumbing only, so it can be carried as a patch onto Effect's TypeScript compiler fork later.

Validation

• npm run build:compiler
• scanner comparison with the forked compiler:
  • no-static main + compatibility patch: Aggregate Instantiations: 13,917,040
  • static-native branch + forked compiler: Aggregate Instantiations: 13,840,894
  • no-static main + stock compiler: Aggregate Instantiations: 17,871,742

The d.ts-emitter path recovered roughly 98% of the static-native source-codegen instantiation benefit in the scanner comparison.

Notes / Gotchas

• Existing native __X_base helper classes are not reinterpreted. Rewriting those to namespace service aliases can make old generated type Make = X.Make aliases self-reference and collapse constructor input to never.
• Textual d.ts parity is close but not byte-for-byte: ordering, import qualification, and boolean vs false | true can still differ.
• This targets the effect-app/TypeScript fork only; do not retarget to the official Microsoft repository.

[patroza]

Update: the emitter now also handles direct bare schema models like class X extends S.Opaque<X>()(...) {} when no namespace X existed before.

For those classes it now synthesizes:

• top-level interface X from the schema Type
• namespace X
• X.Encoded
• X.Make
• X.DecodingServices
• X.EncodingServices
• S.OpaqueFacade<X, X.Encoded, X.Make, X.DecodingServices, X.EncodingServices, {}> base rewrite

It also tightens recognition to schema aliases S.Opaque / Schema.Opaque, so Context.Opaque service tags are not swept into this path.

Validation:

• npm run build:compiler ✅
• tiny declaration fixture for bare S.Opaque<DirectBare>()(...) emitted the full OpaqueFacade + interface + namespace shape ✅

[patroza]

Updated again after measuring direct bare S.Opaque<X>()(...) support.

Latest scanner comparison with this compiler:

• main/no-static + stock TypeScript: 17,871,742
• main/no-static + forked emitter before bare Opaque: 13,917,040
• main/no-static + forked emitter with bare Opaque: 12,128,418
• static-native branch + forked emitter with bare Opaque: 11,724,286

Bare Opaque support adds another 1,788,622 instantiation reduction on no-static main vs the previous emitter pass. Current no-static compiler emit is 404,132 above static-native, so it recovers about 93% of the current stock-to-static-native benefit.

Also fixed synthesized interface/namespace export-ness so non-exported local schema classes merge correctly inside their source module.