Question: For CBOR are Maps with keys of Major type 2 possible?

[eikeon]

I've not been able to work out if it's possible to codec maps with key of major type 2 re: CBOR. I was converting my keys from []byte into string so that I can use them as keys in my map. But they are then getting tagged as major type 3.

[ugorji]

Thanks.

There's a quick hack fix, but I don't like it.

I instead want to take a second stab at how we handle golang string type during encoding, and also how we decode in corresponding context.

This may cause the fix to take a week or so. I will be in touch once fixed.

[eikeon]

Thank you!

I will look forward to the enhancement and can test it once it’s ready for testing.

[ugorji]

@eikeon I couldn't wait to try out my fix.

Please test out and let me know that it solves your concerns.

[eikeon]

Yes, with the EncodeOptions.StringToRaw = true the map keys are now turning into major type 2.

Question: If my map values are also string should the StringToRaw option also apply to them? The resulting CBOR I'm now getting is a map where the keys are encoded with type 2 and the values with type 3. Which in my current use case is fine, I think, as they are 'proper' strings. But want to understand how the determination is made for the map values?

Seems like Go's type system is letting us down a bit here :(

[ugorji]

It's should be working as expected. But I can only be sure with a reproducer.

Send me a reproducer (code I can run with go test or go run), and I can explain what is happening, or fix if something isn't working right.

[ugorji]

@eikeon ^^

[ugorji]

@eikeon seems there is a path within the code which doesn't take the StringToRaw flag into consideration. Fixing now.

[eikeon]

@ugorji, The clarification in the doc is very helpful. I am now seeing StringToRaw being consistently applied.

I am not able to determine how I can use encoding.TextMarshaler so that my type gets encoded as a UTF-8 string. Here's my explainer:

package main

import (
	"bytes"
	"encoding/hex"
	"log"

	"github.com/ugorji/go/codec"
)

type Path string

type PathT string

func (p PathT) MarshalText() (text []byte, err error) {
	return []byte(p), nil
}

type Hash string

type Manifest map[Hash]interface{}

func (m Manifest) Bytes() []byte {
	var ch codec.CborHandle
	var b []byte
	ch.BasicHandle.EncodeOptions.Canonical = true
	ch.BasicHandle.EncodeOptions.StringToRaw = true
	enc := codec.NewEncoderBytes(&b, &ch)
	err := enc.Encode(m)
	if err != nil {
		log.Fatal(err)
	}
	return b
}

func main() {
	bufA, _ := hex.DecodeString("da8a33")
	bufB, _ := hex.DecodeString("5ea3f0")

	m := Manifest{Hash(bufB): Path("b"), Hash(bufA): Path("a")}
	mm := Manifest{Hash(bufB): PathT("b"), Hash(bufA): PathT("a")}

	if bytes.Compare(m.Bytes(), mm.Bytes()) == 0 {
		log.Println("not different")
	}
}

[ugorji]

see https://godoc.org/github.com/ugorji/go/codec#hdr-Custom_Encoding_and_Decoding

Type must implement both sides of the symmetry i.e. MarshalText and UnmarshalText

[ugorji]

@eikeon

BTW, for performance and ergonomic reasons, I would write the code snippet as below (see https://godoc.org/github.com/ugorji/go/codec#hdr-Usage ):

type Manifest map[Hash]interface{}

var ch codec.CborHandle // handles maintain cache of things on first use - make shared, see docs above
func init() {
	ch.Canonical = true // leveraging ergonomic embedding rules of go
	ch.StringToRaw = true // leveraging ergonomic embedding rules of go
}
func (m Manifest) Bytes() (b []byte) {
	enc := codec.NewEncoderBytes(&b, &ch)
	err := enc.Encode(m)
	if err != nil {
		log.Fatal(err)
	}
	return
}