proposal: net/url: implement TextMarshaler/TextUnmarshaler interfaces for url.URL

[YaroslavDev]

In order to be able to pass url.URL variable to flag.TextVar(&url, ...), url.URL type must implement encoding.TextMarshaler and encoding.TextUnmarshaler interfaces.
This approach was suggested here instead of making a dedicated flag.URLVar function in flag package.

[earthboundkid]

I think this would be good, but is it a breaking change? Right now, the output of this code

	u, _ := url.Parse("http://example.com")
	b, _ := json.Marshal(u)
	fmt.Println(string(b))

is

{
 "ForceQuery": false,
 "Fragment": "",
 "Host": "example.com",
 "Opaque": "",
 "Path": "",
 "RawFragment": "",
 "RawPath": "",
 "RawQuery": "",
 "Scheme": "http",
 "User": null
}

I think it would be better if the output were "http://example.com" but maybe someone somewhere would miss the old output?

[YaroslavDev]

@carlmjohnson nice catch!
I guess there are 2 options:

1. Make a breaking change since url.URL typically is not used as JSON(maybe only for debugging/logging purposes)
2. Implement explicitly json.Marshaler and json.Unmarshaler interfaces on url.URL as well to maintain backward compatibility

[rsc]

This proposal has been added to the active column of the proposals project
and will now be reviewed at the weekly proposal review meetings.
— rsc for the proposal review group

[rsc]

It seems like JSON decoding the { } form might still work since that's not a JSON string, so it won't try the TextUnmarshaler. So an old encoding might be readable by a new Go where URL implements these methods. The reverse would not be true.

Have we decided before whether this is OK? I seem to remember JSON encoding of IP addresses used to be base64 byte encodings, but I think we decided that was just so broken we could fix it.

Not changing the JSON seems like a big inconsistency too since I assume the goal here is to get better JSON support for URLs (in addition to flag support).

[seankhliao]

It seems like JSON decoding the { } form might still work since that's not a JSON string, so it won't try the TextUnmarshaler.

encoding/json currently doesn't have that kind of fallback: https://go.dev/play/p/z8QXBU6Ov2z

json: cannot unmarshal object into Go value of type *main.New

---

Have we decided before whether this is OK?

#5086 for net.IP: done in go1.2
#10244 for net.HardwareAddr: declined for no standard format
#10275 for stdlib types in general: undecided
#12803 for net.IPNet: Not ok
#16039 for time.Duration: declined for no standard format
#24174 for time.Duration: folded into #10275
#25705 for url.URL: folded into #10275
#29678 for net.HardwareAddr: undecided
#40436 for net.IPMask: no comments

[rsc]

Thanks very much for that list, @seankhliao! It looks like we should merge this one into #10275 too, and then bring 10275 over to the non-Go2 proposal process.

[rsc]

This proposal is a duplicate of a previously discussed proposal, as noted above,
and there is no significant new information to justify reopening the discussion.
The issue has therefore been declined as a duplicate.
— rsc for the proposal review group