Feature Request: Automatic Expansion of uninitialized record fields

[ProofOfKeags]

When a warning for "Fields of ‘SomeRecord’ not initialised: a, b, c, d" fires, I'd like a CodeAction that allows it to initialize all of those rows with typed holes.

Flow would look like this:

1. let thing = SomeRecord{}
2. Ctrl+'.' (or other code action trigger)
3. let thing = SomeRecord{} --> let thing = SomeRecord { a = _, b = _, c = _, d = _ }

The reason that this is somewhat of a big deal is that in the presence of long record names with the conventional prefix based row names (someRecordA, someRecordB, ...), the typing for this is nontrivial and also entirely mechanical and so the editor is perfectly capable of doing this for me.

[ProofOfKeags]

More sophisticated versions of this could infer what fields are already initialized, and only supply the ones that are missing in this way.

[ProofOfKeags]

Is this something that belongs in ghcide or hls?

[googleson78]

Even cooler would be, agda supports straight up introducing the entire record.
So
0.

let
  thing :: SomeRecordType 
  thing = _

1. <introduce-record

let
  thing :: SomeRecordType 
  thing =
    SomeRecordType
      { field1 = _
      , ...
      }

[jneira]

I would say that i could be done in a plugin in hls, but not sure.

[ProofOfKeags]

@googleson78 yeah that's also great. That said, when the typedef expands (like many records do over time), I'd still like to be able to introduce the missing fields.

[antoine-fl]

Somewhat related, it would be nice to have a shortcut to convert from using the "normal" constructor to using record syntax.

For example if I have the following:

data Person = Person { firstName :: Text, lastName :: Text }

let p = Person "Robert" "Dupont"

Some shortcut could convert the second line to:

let p = Person { firstName = "Robert", lastName = "Dupont" }

[jneira]

Now tactics has a code action: Wingman: Use constructor Person which converts

person :: Person
person = _

in

person :: Person
person = Person {firstName = _wf, lastName = _wg}

So i think the suggestion of @googleson78 is already covered.

It still misses the other two requests:

• Person {} to Person {firstName = _wf, lastName = _wg}
• Person wf wg to Person {firstName = _wf, lastName = _wg}

For the first one there is a dedicated ghc warning usable to add a quick fix:

• Fields of ‘Person’ not initialised: firstName, lastName
• In the expression: Person {}
  In an equation for ‘person’: person = Person {}typecheck(-Wmissing-fields)

So i would say it should not be very much dificult to get

[naufik]

Had a look at the codebase and I think this might be a good start for me to start contributing. Can I attempt this issue?

My outline would be to add a code action for the Fields of '<X>' uninitialized warning. Right now, there is a code action for "suppress errors" so I suppose I can add an alternative quick fix as @jneira mentioned.

For suggestion 2 I'd probably have to have a look on how LHS detects if a token is a record constructor invoked using function syntax.

Hoping to spend the week on this 🤞

[jneira]

@naufik nice! thanks for willing to contribute, feel free to ask anything here or in the chat channels.