Stream: beginners

Topic: Alias for destructured record

Ilya Shmygol (Apr 23 2025 at 16:43):

Isn't it possible to do {a, b} as full_record in the function arguments as is it is in when-is?

Ilya Shmygol (Apr 23 2025 at 16:44):

This definition fails:

add_user : Database, { name : Str, owes ?? Dict Str Dec, owed_by ?? Dict Str Dec, balance ?? Dec} -> Result Database _
add_user = |database, { name, owes ?? Dict.empty({}), owed_by ?? Dict.empty({}), balance ?? 0.0} as user|

with error:

── MALFORMED ARGS LIST in ./Database.roc ───────────────────────────────────────

I was trying to parse the arguments list for a function, but I got
stuck here:

19│  add_user = |database, { name, owes ?? Dict.empty({}), owed_by ?? Dict.empty({}), balance ?? 0.0} as user|
                                                                                                      ^

I was expecting to find a | next.%

Anton (Apr 23 2025 at 17:14):

Not yet supported, but seems reasonable to add. Do you expect any problems from this @Anthony Bullard, @Joshua Warner?

Sam Mohr (Apr 23 2025 at 21:12):

That should be fine so long as the left side of the x as y is an exhaustive pattern. I believe this has been already implemented in the way that @Ilya Shmygol wrote it in the new compiler's parsing code, but I might be wrong

Brendan Hansknecht (Apr 23 2025 at 22:31):

Yeah, I think it just doesn't work with ??

Brendan Hansknecht (Apr 23 2025 at 22:31):

But not fully sure

Anthony Bullard (Apr 24 2025 at 00:55):

I honestly can’t remember if this will in fact “just work” in the new parser but I think so. I wish I had time to look at it right now

Sam Mohr (Apr 24 2025 at 02:44):

Brendan is probably right, but for those who don't know, ?? is going away in the rewrite

Anthony Bullard (Apr 24 2025 at 02:46):

That’s news to me and the parser :joy:

Richard Feldman (Apr 24 2025 at 03:13):

wait, me too! :laughing:

Richard Feldman (Apr 24 2025 at 03:13):

unless I'm misremembering?

Brendan Hansknecht (Apr 24 2025 at 05:24):

Optional/default value record fields are going away

Brendan Hansknecht (Apr 24 2025 at 05:24):

I think ?? is switching to be Result.with_default(...)

Brendan Hansknecht (Apr 24 2025 at 05:25):

At least that is what I remember being decided

Brendan Hansknecht (Apr 24 2025 at 05:25):

But not fully sure on with default.

Anthony Bullard (Apr 24 2025 at 10:35):

That’s somewhat sad, since it was implemented not even six months ago (by me) and has just as much value in a static dispatch world as before. And with_default has been there and will remain for when you want to chain it (we almost had ?? Desugar to it, but chose to use a when instead)

Ilya Shmygol (Apr 24 2025 at 10:41):

I liked ??...

Richard Feldman (Apr 24 2025 at 12:55):

yeah I don't remember that change haha

Richard Feldman (Apr 24 2025 at 12:55):

in my mind, the plan was still to have ?? work the same as today

Richard Feldman (Apr 24 2025 at 12:56):

can anyone find a thread where we discussed changing that?

Anthony Bullard (Apr 24 2025 at 13:44):

It may have stemmed from this discussion?

Brendan Hansknecht said:

Now that we have static dispatch do we expect to see issues with ? and ?? ?
They no longer play nice with pipelines

out =
    (((load_config(in_file) ?? default_config)
        .stage1(abc) ? Stage1Err)
        .stage2() ? Stage2Err)
        .stage3() ? Stage3Err

Brendan Hansknecht (Apr 24 2025 at 15:52):

To clarify my statement. In the new compiler, I would expect this to still just work:

out =
    (((load_config(in_file) ?? default_config)
        .stage1(abc) ? Stage1Err)
        .stage2() ? Stage2Err)
        .stage3() ? Stage3Err

Brendan Hansknecht (Apr 24 2025 at 15:55):

I just recall a discussion on default value record fields like:

fn = |record : {x ?? i32, y ?? i32}|
    {x ?? 0, y ?? 0} = record
    x + y

In that discussion, richard suggested we should remove them cause overall they had been confusing and error prone for users.

Brendan Hansknecht (Apr 24 2025 at 16:13):

I think this is the most official message of removal, but significant discussion definitely happened in idea threads much earlier:
Richard Feldman said:

of note, the "demanded" enum for records isn't necessary anymore due to optional record fields going away

Richard Feldman (Apr 24 2025 at 16:52):

ohh yeah, I misunderstood

Richard Feldman (Apr 24 2025 at 16:52):

so my understanding is that:

• ?? is still an infix operator, which works the same way as today
• we no longer support default record fields, which used to also use ?? as their syntax for specifying the default fields

Brendan Hansknecht (Apr 24 2025 at 16:59):

yeah, exactly that.

Sam Mohr (Apr 25 2025 at 00:23):

yep!

Anthony Bullard (Apr 25 2025 at 13:00):

Sweet, that makes sense

Last updated: Jul 06 2025 at 12:14 UTC