Record update syntax sugar with pattern matching

[ricardo-valero]

These are examples of how to get and update records (and their "desugared" versions)

readA = \record -> record |> .a
# readA = \record -> record.a

writeA = \value -> \record -> record |> &a value
# writeA = \value -> \record -> { record & a: value }

This one works

readT = \type -> \record ->
        record
        |>
        when type is
            A -> .a
            B -> .b
            C -> .c

# readT = \type -> \record ->
#        when type is
#            A -> record.a
#            B -> record.b
#            C -> record.c

expect (readT A) { a: "Roc", b: "is", c: "great" } == "Roc"

But the next one isn't valid

writeT = \type, value -> \record ->
        record
        |>
        when type is
            A -> &a value
            B -> &b value
            C -> &c value

# writeT = \type, value -> \record ->
#        when type is
#            A -> { record & a: value }
#            B -> { record & b: value }
#            C -> { record & c: value }

expect (writeT C "the best") { a: "Roc", b: "is", c: "great" } == { a: "Roc", b: "is", c: "the best" }

It's a silly example but I think the point is understood.

[ricardo-valero]

Related to #2659

[lukewilliamboswell]

@ricardo-valero this is a good first issue if you are interested exploring further.

The record update is parsed into roc_parse::ast::Expr::RecordUpdater and then de-sugared in crates/compiler/can/src/desugar.rs. It's pretty much all contained to one pattern match in desugar_expr, and a fun place to see how the front end of the compiler AST is handled.