Partial decoding using Regex/Seqexp

[arunvickram]

Right now the implementation of :catn has it so that the following example doesn't decode "1.0" if TestEnum fails:

(ns sample.core 
  (:require [malli.core :as m]
            [malli.transform :as mt]))
​
(def GreaterThanZero
	[:fn {:decode/string mt/-string->double}
	 '#(> % 0)])
​
(def TestEnum
	[:enum "A" "B"])
​
(m/decode
 [:catn 
  [:amount GreaterThanZero]
  [:layout TestEnum]]
 ["1.0" "U"]
 mt/string-transformer) ; => ["1.0" "U"]

I was wondering if there were any plans to support partial decoding?

[nilern]

This might be hard (or not). IMHO we should not make any guarantees about how (partially) invalid input is decoded. (Ceterum censeo, it should produce an error but that is not on the roadmap.)

[arunvickram]

Sorry for the long delay, I've been flooded with work. I'm not sure why this would be difficult, but perhaps I'm missing some context on how decoders work?

[nilern]

We would have to keep the longest partial decoding around and then return that if the full sequence cannot be decoded. Maybe not super difficult but definitely additional complexity and reduced performance and usually the input has (or should have) been validated beforehand anyway. I would rather have #404.

[arunvickram]

Either way works, in fact a coercer may even be better because at the end of the day, I'm getting stringified data and I'd like to be able to get the correct error reported for fn schemas and schemas with custom errors.

[nilern]

Ah I might have it the wrong way around, in that case you would first decode and then validate... still I think the coercer would be more optimal. Either way quite a bit of work to improve the errors.