Suppose we have a discriminated union
export const duParser = z.discriminatedUnion('type', [
z.object({
type: z.literal('a'),
id: aParser,
}),
z.object({
type: z.literal('b'),
id: bParser,
}),
]);
The inferred result using z.infer would look like
{type: 'a', id: A} | {type: 'b', id: B}Now, the goal is to obtain a parser and create a specific type from it, for example:
Rename the union field (type) to another name (e.g. type -> type2):
{type2: 'a', id: A} | {type2: 'b', id: B}Map other fields (although I acknowledge this can be more complex):
{type: 'a', id2: A} | {type: 'b', id2: B}
The scenario involves having a generic parser that outputs {type: ..., id: ...}, which needs to be combined with a flat structure parser, resulting in prefixed fields for the discriminated union. The requirement for the flat parser stems from the fact that the parsed data represents an HTTP query string:
rootField1=value1&rootField2=value2&subfieldId=subValue1&subfieldType=asubfieldId and subfieldType correspond to the "id" and "type" attributes of the original discriminatedUnion. However, retaining them as "id" and "type" is not ideal as these attributes are part of a different abstraction level (the root entity may have its own id and type).
Typically, a nested structure would be utilized in this case, but there's curiosity about whether Zod allows for flattening and mapping field names simultaneously.