| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Implicit search detects likely cycles by looking at the stack of
open implicits and checking the same implicit appears twice, and
if the second occurrence is trying satisfy an implicit search for
a "dominant" type.
Originally, this condition immediately failed the entire implicit
search. However, since Scala 2.10, this mechanism has been refined to
continue searching after the first divergent implicit is detected.
If a second divergence is found, we fail immediately. If the followup
search fails, we report the first divergence. Otherwise, we
take the successful result.
This mechanism was originally built around exceptions. This proved
to be fragile, and was refactored in SI-7291 / accaa314 to instead
use the `Context.errors` to control the process.
But, since that change, the pattern of implicits in scalanlp/breeze
and Shapeless have been prone to reporting the divergent implicit
errors where they used to recover.
So long as we left the `DivergentImplictTypeError` that originates
from a nested implicit search in `context.errors`, we are unable to
successfully typecheck other candidates. This commit instead
stashes the first such error away in `DivergentImplicitRecovery`,
to clear the way for the alternative path to succeed.
We must retain any other divergent implicit errors, as witnessed by
test/files/neg/t2031.scala, which loops unless we retain divergent
implicit errors that we don't stash in `DivergentImplicitRecovery`.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This simple patch makes it possible for implicit views to benefit from
fundep-guided type inference, eliminating a nasty special case in
implicit inference.
Here are the changes that I had to apply to the tests (they exposed
quite a number of interesting questions that I was happy to answer):
1) neg/t5845.scala now works, so I moved it to pos. That easily makes sense,
because the test was just a canary to track previous state of things.
2) neg/divergent_implicit.scala, neg/t5578.scala and neg/t7519.scala
changed their messages to less informative ones, because inapplicable
implicit views now get disqualified early and therefore don't display
their error messages to the user. This is an unfortunate but necessary
byproduct of this change, and one can argue that the behavior is now
completely consistent with implicit vals (that also don't explain why
this or that implicit val got disqualified, but rather display a generic
implicit value not found message).
3) scaladoc/implicits-chaining.scala and scaladoc/implicits-base.scala.
Immediate culling of apriori inapplicable implicit views messes things up
for Scaladoc, because it's interested in potentially applicable views,
having infrastructure to track various constraints (type bounds, requirements
for implicit parameters, etc) that guide applicability of views and then
present it to the user. Therefore, when scaladoc is detected, implicit search
reverts to the old behavior.
4) We still cannot have Jason's workaround to type constructor inference
mentioned in comments to SI-3346, because it's not really about implicit
parameters of implicit views, but about type inference flowing from the
implicit parameter list to a preceding parameter list in order to affect
inference of an implicit view. This is something that's still too ambitious.
|
|
|
|
|
| |
This reverts commit 884e1ce762d98b29594146d37b85384581d9ba96, reversing
changes made to f6fcc4431f272c707d49de68add532c452dd4b0f.
|
|
|
|
|
|
|
|
|
|
|
|
| |
The parser hole I found while working on the generated positions
serves as the umbrella for a host of improvements. Upgraded
positions assigned during some specific challenging situations mostly
involving the creation of synthetic trees, e.g. for comprehensions
and closures. While doing so improved some error messages.
Eliminated some of the most glaring duplication in the parser.
It's written like there is some payoff associated with being
spectacularly imperative. Not so far.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Proliferating the number of debugging modes because it's still way
too hard to see what's going on in there. Until we get hubert's type
debugger with its whiz-bang whizbanginess, we'll have to struggle along
with somewhat prettier ascii. This introduces:
-Yinfer-debug
which tries to print in readable fashion what is happening in the
worlds of inference and implicit search. It should be made a bit more
complementary and less overlappy with -Ytyper-debug. No review.
|
|
|
|
|
| |
Updated neg check files to the new output of r17773
|
|
|
|
|
| |
Better inference for implicits; some preparations for new collections.
|
|
(1) more changes for dynamic cycle detection of implicits 2) change
(to wildcard scoping in types 3) suppress $tag generation for purely
(abstract traits 4) increase stacksize for sabbus to 32m (otherwise
(problems on Windows)
|