question: how to correctly decode CBOR-encoded data that was encoded with cbor.StringToByteString

[shanebishop]

I see support for decoding and encoding CBOR data with byte strings was added in issue #446 in PR #465. I would like to use this feature in order to automatically transform input data that can include invalid UTF-8 into CBOR that I can successfully decode (if this is a misuse of the feature, please let me know).

I encoded some data in CBOR with the following encoding options:

options := cbor.CoreDetEncOptions()
options.Time = cbor.TimeRFC3339Nano
options.String = cbor.StringToByteString

I am now trying to decode the CBOR, but everything I have tried so far gives me an error.

This code gives me the error cbor: cannot unmarshal byte string into Go struct field wmd.WmdEvent.created of type time.Time:

decMode, err := cbor.DecOptions{ByteStringToString: cbor.ByteStringToStringAllowed}.DecMode()
if err != nil {
	return nil, err
}
err = decMode.Unmarshal(b.Bytes(), v)

When I use

cbor.DecOptions{DefaultByteStringType: reflect.TypeOf(cbor.ByteString(""))}.DecMode()

I get the error cbor: cannot unmarshal byte string into Go struct field wmd.WmdEvent.id of type string.

How can I correctly decode my CBOR data?

[benluddy]

Hi @shanebishop, have you tried:

cbor.DecOptions{
  ByteStringToString: cbor.ByteStringToStringAllowed,
  DefaultByteStringType: reflect.TypeOf(""),
}

ByteStringToStringAllowed should let you decode a CBOR byte string into the string field.

When decoding into time.Time, the current implementation effectively decodes the data item into interface{}, then tries to convert the resulting Go value to time.Time:

cbor/decode.go

Lines 1270 to 1300 in cfbd0ff

	var content interface{}
	content, err = d.parse(false)
	if err != nil {
	return
	}
	switch c := content.(type) {
	case nil:
	return
	case uint64:
	return time.Unix(int64(c), 0), nil
	case int64:
	return time.Unix(c, 0), nil
	case float64:
	if math.IsNaN(c) || math.IsInf(c, 0) {
	return
	}
	f1, f2 := math.Modf(c)
	return time.Unix(int64(f1), int64(f2*1e9)), nil
	case string:
	tm, err = time.Parse(time.RFC3339, c)
	if err != nil {
	tm = time.Time{}
	err = errors.New("cbor: cannot set " + c + " for time.Time: " + err.Error())
	return
	}
	return
	default:
	err = &UnmarshalTypeError{CBORType: t.String(), GoType: typeTime.String()}
	return
	}

In any case, the above configuration for DefaultByteStringType should allow you to decode an untagged byte string into time.Time. I'm not sure it's desirable that an option controlling decodes into interface{} has a side effect on decodes into time.Time. I was surprised to discover it for myself yesterday (#497 (comment)).

[shanebishop]

Thanks @benluddy, your suggestion worked! 🎉

Now I just need to figure out how to get com.fasterxml.jackson.dataformat.cbor.databind.CBORMapper in the consuming Java application to play nice with the byte string-encoded CBOR, but that is outside the scope of this question.

[fxamacker]

Hey @shanebishop, thanks for opening this issue!

Based on more discussion with @benluddy, the workaround currently relies on unintended side effect 🐞 of implementation details.

New feature/bugfix for this will be implemented this month (probably by adding a new decoding option).

I'll open new issue and will also comment here when the new feature becomes available for you to try.