Stream: contributing

Topic: deprecating WSA (whitespace application for function calls)

jan kili (Feb 12 2025 at 21:18):

As discussed in #ideas > reflecting on static dispatch... and discussing removing WSA , we're going to start officially deprecating WSA (whitespace application syntax for function calls). For example, Num.add 1 2 should now be written as Num.add(1, 2) instead, using the new PNC (parentheses and commas) function call syntax.

There are several aspects for us to discuss further:

• WSA is currently also used in our type system, for applying type variables (List Str). I infer that tag payloads (Custom Str) are simply a subset of type variables, but maybe they count as a distinct use case. Should we deprecate WSA in the type system, too?
• When is the first day that we want anyone reading this thread to stop writing new WSA code?
• When do we want to announce that everyone should migrate all their active code from WSA to PNC?
• When is the first day that we want Roc to start warning about WSA code?
• When is the first day that we want anyone not reading this thread to receive PRs migrating their WSA code to PNC?
• When is the first day that we want Roc to stop compiling WSA code?
• When is the first day that we want Roc to stop recognizing WSA code?
• Anything else we need to discuss for a smooth deprecation?

Luke Boswell (Feb 12 2025 at 21:20):

I think the plan is that types are WSA and values are PNC, its subtle but I appreciate having them distinct visually.

Sam Mohr (Feb 12 2025 at 21:21):

#ideas > Using parens for types

Luke Boswell (Feb 12 2025 at 21:23):

e.g. [Number I64, String Str] is the type of a tag union, with Number(42), or String("a string") being values of this type.

Brendan Hansknecht (Feb 12 2025 at 22:49):

I think we have a very easy removal point. Just never port WSA to the new zig compiler

Brendan Hansknecht (Feb 12 2025 at 22:50):

Whenever that is the default compiler, it will be officially removed

Brendan Hansknecht (Feb 12 2025 at 22:50):

It is already deprecated in the old rust compiler and will auto format to PNC

Brendan Hansknecht (Feb 12 2025 at 22:51):

So I would label it as deprecated already

Brendan Hansknecht (Feb 12 2025 at 22:51):

I don't think we need to do anything more aggressive in the old rust compiler.

Anthony Bullard (Feb 18 2025 at 17:12):

I agree with Brendan

jan kili (Feb 18 2025 at 18:42):

These simple replies seem like consensus! CC @Richard Feldman

• WSA is currently also used in our type system, for applying type variables (List Str). I infer that tag payloads (Custom Str) are simply a subset of type variables, but maybe they count as a distinct use case. Should we deprecate WSA in the type system, too?
  • #ideas > Using parens for types is resolving this.
• When is the first day that we want anyone reading this thread to stop writing new WSA code?
  • Today.
• When do we want to announce that everyone should migrate all their active code from WSA to PNC?
  • Tomorrow?
• When is the first day that we want Roc to start warning about WSA code?
  • Never? Hopefully the new compiler can have an errorfix hint like "It looks like you forgot parentheses around function arguments.", but maybe that's an unreasonable insight.
• When is the first day that we want anyone not reading this thread to receive PRs migrating their WSA code to PNC?
  • Tomorrow?
• When is the first day that we want Roc to stop compiling WSA code?
  • The day we release v0.1.0.
• When is the first day that we want Roc to stop recognizing WSA code?
  • The day we release v0.1.0.
• Anything else we need to discuss for a smooth deprecation?
  • I just want to advocate for us encouraging ecosystem tooling to not drop support for any v0.0.0 syntax until v0.1.0 fully releases.

:thumbs_up:s?

Last updated: Jul 06 2025 at 12:14 UTC