Input string to output number while preserve the schema

[sumerokr]

Hi,
How can I receive a string input, coerce it to number, only then validate it as a number and also make Input and Output type correct?

Or, in other words, how can I create a schema that

• returns string for Input<typeof MySchema>
• coerces incoming string value to number type before validation
• validates coerced value as a number
• returns number value
• returns number for Output<typeof MySchema>

---

What I have found in docs

const NumberSchema = coerce(number(), Number);
const numberOutput = parse(NumberSchema, '1234'); // 1234

While string can be passed here, its Input<typeof NumberSchema> returns number.

---

transform(string(), Number);

Sets correct Input (string) and Output (number) types, but validation happens against string, before the transformation takes place.

[fabian-hiller]

I am aware of this problem, but have not had time to investigate it in detail. Currently, the behavior you expect is not possible. Do you know how it is with other schema libraries?

[sumerokr]

Hey, thanks for the answer! Unfortunatelly, I am not aware of others implementation.

[fabian-hiller]

Ok. I hope to find time to look into this in 1 to 2 weeks.

[sillvva]

@sumerokr Here's a custom transform function, which takes an input type, an output schema, performs the transform, and validates only the output.

export function ioTransform<TInput, TOutput extends BaseSchema>(
	outputSchema: TOutput,
	action: (value: TInput) => Output<TOutput>
): BaseSchema<TInput, Output<TOutput>> {
	return {
		...outputSchema,
		_parse(input: TInput, info?: ParseInfo) {
			return outputSchema._parse(action(input), info);
		}
	};
}

It does require a custom validator to handle NaN values, which do pass the base number() check which is typeof input === "number".

[fabian-hiller]

This looks almost the same as our coerce method. The only difference is that the input type can be defined. However, we could improve this. Feel free to create an issue or PR.

[huyhoang160593]

@sumerokr Here's a custom transform function, which takes an input type, an output schema, performs the transform, and validates only the output.

export function ioTransform<TInput, TOutput extends BaseSchema>(
	outputSchema: TOutput,
	action: (value: TInput) => Output<TOutput>
): BaseSchema<TInput, Output<TOutput>> {
	return {
		...outputSchema,
		_parse(input: TInput) {
			const result = outputSchema._parse(action(input));
			return result.issues ? result : { output: result.output };
		}
	};
}

It does require a custom validator to handle NaN values, which do pass the base number() check which is typeof input === "number".

@sumerokr Here's a custom transform function, which takes an input type, an output schema, performs the transform, and validates only the output.

export function ioTransform<TInput, TOutput extends BaseSchema>(
	outputSchema: TOutput,
	action: (value: TInput) => Output<TOutput>
): BaseSchema<TInput, Output<TOutput>> {
	return {
		...outputSchema,
		_parse(input: TInput) {
			const result = outputSchema._parse(action(input));
			return result.issues ? result : { output: result.output };
		}
	};
}

It does require a custom validator to handle NaN values, which do pass the base number() check which is typeof input === "number".

Can we add this to the core library, I'm in very need of this to perform my validation to transform string to array string and validate that array

[fabian-hiller]

This looks almost the same as our coerce method. The only difference is that the input type can be defined. However, we could improve this. Feel free to create an issue or PR.