3 replies

Validating Objects Without `_tag` in `Schema.TaggedError`

When using

Schema.TaggedError

Schema.TaggedError

, is it possible to validate/decode objects that may not have the

_tag

_tag

property? I've tried a few different solutions, but none seem... right. Here's some examples of modeling an XRPC spec-compliant error response from atproto

export class XrpcError extends Schema.TaggedError<XrpcError>()("XrpcError", {
  _tag: Schema.optional(Schema.Literal("XrpcError")),
  error: Schema.String,
  message: Schema.optional(Schema.String),
}) {}

export class XrpcError extends Schema.TaggedError<XrpcError>()("XrpcError", {
  _tag: Schema.optional(Schema.Literal("XrpcError")),
  error: Schema.String,
  message: Schema.optional(Schema.String),
}) {}

Any usage of the field

_tag

_tag

results in a

DuplicatePropertySignatureError

DuplicatePropertySignatureError

export class XrpcError extends Schema.TaggedError<XrpcError>()("XrpcError", {
  error: Schema.String,
  message: Schema.optional(Schema.String),
}) {
  static readonly Response = XrpcError.pipe(Schema.omit("_tag"));
}

export class XrpcError extends Schema.TaggedError<XrpcError>()("XrpcError", {
  error: Schema.String,
  message: Schema.optional(Schema.String),
}) {
  static readonly Response = XrpcError.pipe(Schema.omit("_tag"));
}

This should otherwise be fine, but I have to replicate this everywhere I define a response-driven tagged error. Not to mention, for some reason, the type looks like this:

(property) XrpcError.Response: Schema.SchemaClass<{
    readonly error: string;
    [Symbol.iterator]: () => Effect.EffectGenerator<Effect.Effect<never, XrpcError, never>>;
    pipe: {
        <A>(this: A): A;
...

(property) XrpcError.Response: Schema.SchemaClass<{
    readonly error: string;
    [Symbol.iterator]: () => Effect.EffectGenerator<Effect.Effect<never, XrpcError, never>>;
    pipe: {
        <A>(this: A): A;
...

export class XrpcError extends Schema.TaggedError<XrpcError>()("XrpcError", {
  error: Schema.String,
  message: Schema.optional(Schema.String),
}) {
  static readonly Response = Schema.encodedBoundSchema(XrpcError).pipe(Schema.omit("_tag"));
}

export class XrpcError extends Schema.TaggedError<XrpcError>()("XrpcError", {
  error: Schema.String,
  message: Schema.optional(Schema.String),
}) {
  static readonly Response = Schema.encodedBoundSchema(XrpcError).pipe(Schema.omit("_tag"));
}

This works, but I mean come on, no way this is the way that it has to be.

I've also tried cleverly using the constructor as a validator since it uses

ParseResult.validateSync

ParseResult.validateSync

under the hood for validation and it helpfully doesn't require the

_tag

_tag

and adds it after parsing. the only downside: I have to opt out of type checking and the constructor throws

(u: unknown) => new XrpcError(u as any)

(u: unknown) => new XrpcError(u as any)

I'm curious what the preferred approach is in this case

message.txt4.92KB

Validating Objects Without `_tag` in `Schema.TaggedError`

When using

Schema.TaggedError

Schema.TaggedError

, is it possible to validate/decode objects that may not have the

_tag

_tag

property? I've tried a few different solutions, but none seem... right. Here's some examples of modeling an XRPC spec-compliant error response from atproto

export class XrpcError extends Schema.TaggedError<XrpcError>()("XrpcError", {
  _tag: Schema.optional(Schema.Literal("XrpcError")),
  error: Schema.String,
  message: Schema.optional(Schema.String),
}) {}

export class XrpcError extends Schema.TaggedError<XrpcError>()("XrpcError", {
  _tag: Schema.optional(Schema.Literal("XrpcError")),
  error: Schema.String,
  message: Schema.optional(Schema.String),
}) {}

Any usage of the field

_tag

_tag

results in a

DuplicatePropertySignatureError

DuplicatePropertySignatureError

export class XrpcError extends Schema.TaggedError<XrpcError>()("XrpcError", {
  error: Schema.String,
  message: Schema.optional(Schema.String),
}) {
  static readonly Response = XrpcError.pipe(Schema.omit("_tag"));
}

export class XrpcError extends Schema.TaggedError<XrpcError>()("XrpcError", {
  error: Schema.String,
  message: Schema.optional(Schema.String),
}) {
  static readonly Response = XrpcError.pipe(Schema.omit("_tag"));
}

This should otherwise be fine, but I have to replicate this everywhere I define a response-driven tagged error. Not to mention, for some reason, the type looks like this:

(property) XrpcError.Response: Schema.SchemaClass<{
    readonly error: string;
    [Symbol.iterator]: () => Effect.EffectGenerator<Effect.Effect<never, XrpcError, never>>;
    pipe: {
        <A>(this: A): A;
...

(property) XrpcError.Response: Schema.SchemaClass<{
    readonly error: string;
    [Symbol.iterator]: () => Effect.EffectGenerator<Effect.Effect<never, XrpcError, never>>;
    pipe: {
        <A>(this: A): A;
...

export class XrpcError extends Schema.TaggedError<XrpcError>()("XrpcError", {
  error: Schema.String,
  message: Schema.optional(Schema.String),
}) {
  static readonly Response = Schema.encodedBoundSchema(XrpcError).pipe(Schema.omit("_tag"));
}

export class XrpcError extends Schema.TaggedError<XrpcError>()("XrpcError", {
  error: Schema.String,
  message: Schema.optional(Schema.String),
}) {
  static readonly Response = Schema.encodedBoundSchema(XrpcError).pipe(Schema.omit("_tag"));
}

This works, but I mean come on, no way this is the way that it has to be.

I've also tried cleverly using the constructor as a validator since it uses

ParseResult.validateSync

ParseResult.validateSync

under the hood for validation and it helpfully doesn't require the

_tag

_tag

and adds it after parsing. the only downside: I have to opt out of type checking and the constructor throws

(u: unknown) => new XrpcError(u as any)

(u: unknown) => new XrpcError(u as any)

I'm curious what the preferred approach is in this case

message.txt4.92KB

Validating Objects Without `_tag` in `Schema.TaggedError`

Validating Objects Without `_tag` in `Schema.TaggedError`

Similar Threads

Similar Threads

Similar Threads