| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This fixes a mistake in macro impl ref typechecking that used to have
an heuristic to figure out whether it looks at a bundle method ref or at
a vanilla object method ref. Under some circumstances the heuristic could
fail, and then the macro engine would reject perfectly good macro impls.
Now every macro impl ref is typechecked twice - once as a bundle method ref
and once as a vanilla object method ref. Results are then analyzed,
checked against ambiguities (which are now correctly reported instead
of incorrectly prioritizing towards bundles) and delivered to the macro
engine.
The only heuristic left in place is the one that's used to report errors.
If both bundle and vanilla typechecks fail, then if a bundle candidate
looks sufficiently similar to a bundle, a bundle typecheck error is reported
providing some common bundle definition hints.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
With the new focus on quasiquotes in macro implementations, we now have
to change the way how inference of macro def return types works.
Previously, if the return type of a macro def wasn’t specified, we looked into
the signature of its macro impl, took its return type (which could only
be c.Expr[T]) and then assigned T to be the return type of the macro def.
We also had a convenient special case which inferred Any in case when
the body of the macro impl wasn’t an expr. That avoided reporting spurious
errors if the macro impl had its body typed incorrectly (because in that
case we would report a def/impl signature mismatch anyway) and also provided
a convenience by letting macro impls end with `???`.
However now we also allow macro impls to return c.Tree, which means that
we are no longer able to do any meaningful type inference, because c.Tree
could correspond to tree of any type.
Unfortunately, when coupled with the type inference special case described
above, this means that the users who migrate from exprs to quasiquotes
are going to face an unpleasant surprise. If they haven’t provided
explicit return types for their macro defs, those types are going to be
silently inferred as `Any`!
This commit plugs this loophole by prohibiting type inference from
non-expr return types of macro impls (not counting Nothing). Moreover,
it also deprecates c.Expr[T] => T inference in order to avoid confusion
when switching between exprs and quasiquotes.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
With the advent of quasiquotes, we allowed both arguments and return types
of macro impls to be c.Tree's (as opposed to traditional c.Expr[T]'s).
This warrants an update of macro def <-> macro impl signature mismatch
errors that include a printout of suggested macro impl signatures. Now
along with a signature that contains exprs, we suggest another signature
that has all exprs replaced by trees
|
|
|
|
|
|
|
|
|
|
|
| |
There are three kinds of assign-like trees:
1. Assign(lhs, rhs) // $lhs = $rhs
3. AssignOrNamedArg(lhs, rhs) // $lhs = $rhs
2. Apply(Select(f, nme.update), args :+ rhs) // $f(..$args) = $rhs
New syntactic combinator unifies all of them and lets users not to think
about these implementations details.
|
|
10229316db allowed macro impls to take and return values of type c.Tree
in addition to the usual c.Expr. However it didn't take into account that
it is often useful to return subtypes of trees (e.g. with quasiquotes
that expand into expressions typed as precisely as possible). This patch
fixes that oversight.
|