Remove function$ entirely.
#7208
Conversation
cristianoc
force-pushed
the
remove-function$-entirely
branch
3 times, most recently
from
cc48899 to
c250201
Compare
| @@ -1195,7 +1195,9 @@ | |||
| addValueReference OptArg.res:1:4 --> OptArg.res:1:29 | |||
| addValueReference OptArg.res:3:4 --> OptArg.res:3:17 | |||
| addValueReference OptArg.res:3:4 --> OptArg.res:3:27 | |||
| DeadOptionalArgs.addReferences foo called with optional argNames:x argNamesMaybe: OptArg.res:5:7 | |||
There was a problem hiding this comment.
Need to look into these changes: the loc of parameters has changed.
There was a problem hiding this comment.
These changes appear to be correct.
| @@ -9,11 +9,12 @@ ContextPath Value[someFunc] | |||
| Path someFunc | |||
| argAtCursor: unlabelled<0> | |||
| extracted params: | |||
| [int, ~two: string=?, ~three: unit => unit, ~four: someVariant, unit] | |||
| [( | |||
There was a problem hiding this comment.
Need to look into these changes: the loc of parameters has changed.
There was a problem hiding this comment.
Given this function type:
type t = ( x: int
^ ^
f t
The parser updates the location of the top-level node to start at location f.
So in the case of function$< Tarrow> the top-level node starts at f and the Tarrow node starts at t.
Now that Tarrow is the top-level node, it starts at f.
Neither is optimal, perhaps at some point one could consider a better representation of function types, where the locations of the overall type, and individual labels is straightforward. (Labels themselves are a little funny with optional parameters).
cristianoc
force-pushed
the
remove-function$-entirely
branch
from
c250201 to
c2f8cca
Compare
cristianoc
force-pushed
the
remove-function$-entirely
branch
2 times, most recently
from
49acad1 to
674c4bf
Compare
cristianoc
force-pushed
the
remove-function$-entirely
branch
from
674c4bf to
b71324f
Compare
| @@ -132,7 +132,7 @@ type t = @attr (foo, @attr2 ~f: bar, @attr3 ~f: baz) => unit | |||
| type t = @attr (string => @attr (int => unit)) | |||
| type t = @attr (string, int) => @attr (int, float) => unit | |||
| type t = @attr (int => @attr (int, float) => @attr (unit => unit => unit)) | |||
| type t = @attr (@attr2 ~f: int, @attr3 ~g: float) => unit | |||
| type t = (@attr @attr2 ~f: int, @attr3 ~g: float) => unit | |||
There was a problem hiding this comment.
This change doesn't look correct?
There was a problem hiding this comment.
There's a single node.
No way to distinguish inside and outside the parens.
There was a problem hiding this comment.
I thought @attr was on the whole function and @attr2 was on the first arg.
So what exactly are @attr and @attr2 on now?
There was a problem hiding this comment.
Later on, a good representation should probably take a list of params, each one with its own locations and attributes.
Then the single function node could have its own location and attributes.
We're bound to having to convert back and forth with the original parse tree for ast compatibility, so either one would have to preserve all information,or at least all the info required for the later phases of the compiler (other than pretty printing).
There was a problem hiding this comment.
I thought
@attrwas on the whole function and@attr2was on the first arg.So what exactly are
@attrand@attr2on now?
The only attributes are at the level of ast nodes.
The type (a, b) => c is just a pretty printed form for 2 ast nodes. (The outer one has the arity info).
There was a problem hiding this comment.
So (a,b) => c is really a =2=> (b => c)
Attributes are on the arrow nodes so the there's no distinction between an attribute on a, and on the whole type: it's the same node, the outer one.
There was a problem hiding this comment.
Cleaning up these things is exactly what we need custom asts for.
This aspect can be cleaned up in a further refinement.
|
Tested against a real world project without any issues! |
No description provided.