Literal field becomes optional in typescript #36

Activity

1. alythobani commented on Mar 20, 2025

   

   alythobani

   on Mar 20, 2025

   Last edited by alythobani

   More actions

   I could be misunderstanding your use case here, but I don't actually think this should be treated differently from #28.

   If MyModel is being sent from TypeScript to Python (e.g. as a request), the presence of the default on the Pydantic model means that even if TS doesn't include my_bool_property, it will still be populated as False on the Python side when processing the request data (converting the JSON into a MyModel instance). So in that case, it should be safe to mark my_bool_property as optional on the TS side.

   And if it's being used as a discriminator on the TS side and should be required, that sounds like something pydantic2ts wouldn't have direct knowledge of (but a flag to avoid defaulted-fields becoming optional would solve the issue).

   But yeah maybe I'm missing something in what you were looking for here.
2. sebbekarlsson commented on Jun 7, 2025

   

   sebbekarlsson

   on Jun 7, 2025

   More actions

   I'm having this issue too
3. lubosmato commented on Oct 3, 2025

   

   lubosmato

   on Oct 3, 2025

   More actions

   I might have similar "issue" / use-case. I would like to use generated types for calling JSON-RPC 2.0 API in TS environment. Now I use Literal types to "pass" method name into TS as a string literal type.

   from typing import Any, Literal, Optional
   from pydantic import BaseModel
   
   
   JSONRPC_VERSION = "2.0"
   
   
   class JsonRpcPayload(BaseModel):
       method: str | Any
       params: None | Any
       jsonrpc: Literal["2.0"] = JSONRPC_VERSION
       id: Optional[str] = None
   
   
   class SomeRpcMethodPayload(JsonRpcPayload):
       method: Literal["some.rpc.method"] = "some.rpc.method"
       params: None = None
   
   
   payload = SomeRpcMethodPayload()
   print(
       payload.model_dump_json()
   )  # {"method":"some.rpc.method","params":null,"jsonrpc":"2.0","id":null}

   export interface JsonRpcPayload {
     method: unknown;
     params: unknown;
     jsonrpc?: "2.0"; // <-- 
     id?: string | null;
   }
   
   export interface SomeRpcMethodPayload {
     method?: "some.rpc.method"; // <-- 
     params?: null;
     jsonrpc?: "2.0"; // <-- 
     id?: string | null;
   }

   Usage:

   const payload: SomeRpcMethodPayload = {}
   // ^ this approach would work if the property method were not optional
   // and caller would be required to set property method, e.g.:
   // const payload: SomeRpcMethodPayload = {method: "some.rpc.method"}
   
   const result = await fetch("...", {
     method: "POST",
     body: JSON.stringify(payload) // -> "{}" 
   }) // body is empty json

   At the moment this is a bit limiting when Literal python type hint is represented by optional property in Typescript - allowing passing no method property. I understand that this is a feature (not a bug).

   Indeed a flag that would avoid defaulted-fields becoming optional would make possible this use-case.
4. lubosmato commented on Oct 3, 2025

   

   lubosmato

   on Oct 3, 2025

   More actions

   On second thought, I realize my approach might be somewhat hacky. The underlying pipeline is:
   Pydantic objects → JSON Schema → json2ts → TypeScript types

   Implementing a flag to make defaulted fields required would essentially break Pydantic's serialization conventions. JSON Schema semantics dictate that fields with default values should be marked as optional, which makes perfect sense from a schema validation perspective.