Commit graph

20 commits

Author SHA1 Message Date
Leonardo de Moura
c7ae8354fd feat: improve type mismatch error messages
Use heuristic to automatically annotate terms with `pp.explicit`.
2020-12-17 07:11:52 -08:00
Leonardo de Moura
7a4b544b1c feat: improve application type mismatch error message
If the type error is at an implicit argument, we annotate
application with `pp.explicit := true`

Given the type incorrect definition
```
def f {a b c : α} : a = c :=
  Eq.trans (a := a) (b := b = c)
```
We now generate the error
```
error: application type mismatch
  @Eq.trans α a (b = c)
argument
  b = c
has type
  Prop
but is expected to have type
  α
```
@Kha Note that we only enable `pp.explicit := true` for the relevant
application. That is, we set `pp.explicit := false` for each children.

Unfortunately, there is a corner case.
```
set_option pp.explicit true
def f {a b c : α} : a = c :=
  Eq.trans (a := a) (b := b = c)
```
produces the error
```
error: application type mismatch
  @Eq.trans α a (b = c)
argument
  @Eq α b c
has type
  Prop
but is expected to have type
  α
```
The reset `pp.explicit := false` overwrote the user option.
I think the simplest solution is the following
1- The delaborator saves the initial set of Options `Init`
2- When it finds a node annotated with a `pp` options, it only
consider the option if it is not set by `Init`.

What do you think?
2020-12-09 13:58:08 -08:00
Leonardo de Moura
ae5aa51712 chore: add explicit discard 2020-12-08 06:18:18 -08:00
Leonardo de Moura
36eb03c601 chore: cleanup Check.lean 2020-11-24 16:13:52 -08:00
Leonardo de Moura
396e767f3d refactor: move Ref to Prelude and rename it to MonadRef
`MacroM` will implement `MonadRef` because
1- It will be easier to throw errors from macros
2- We will be able to `getRef` to retrieve the syntax node at macro
rules.

I renamed `Ref` to `MonadRef` to make it consistent with other classes
providing monadic methods (e.g. `MonadEnv`, `MonadState`, etc).

cc @Kha
2020-11-13 16:00:31 -08:00
Leonardo de Moura
5481999560 chore: cleanup 2020-10-26 14:25:38 -07:00
Leonardo de Moura
13c2a8ff51 chore: remove #lang lean4 header 2020-10-25 09:54:07 -07:00
Leonardo de Moura
1495f403a1 chore: use builtin_initialize instead of initialize at src/Lean 2020-10-19 15:17:02 -07:00
Leonardo de Moura
9abe45ae4f chore: move to new frontend 2020-10-19 06:03:09 -07:00
Leonardo de Moura
bbc1f4d461 fix: throwAppTypeMismatch should be polymorphic
We use it from `TermElabM` and `MetaM`, and they have different
`TermElabM` implementations.
2020-09-22 09:35:59 -07:00
Leonardo de Moura
e181c1adee refactor: add DefEqM
The idea is to make clear that the field `posponed` is transient
state. It is only used during `isDefEq`.
The refactoring was motivated by a bug I found where the `posponed`
constraints were not being handled correctly. For example,
the `check (e : Expr)` method was returning `true`, but leaving pending
universe constraints at `postponed`.

cc @Kha
2020-09-12 16:42:16 -07:00
Leonardo de Moura
02fc447b04 fix: error message 2020-09-09 16:42:41 -07:00
Leonardo de Moura
12f69a78b7 chore: change meaning of lambdaTelescope
`lambdaTelescope` now only consumes `lambda`s.
`lambdaLetTelescope` implements the old behavior.
2020-09-04 13:35:17 -07:00
Leonardo de Moura
238c38fed9 feat: preserve nonDep flag at LocalDecl.ldecl 2020-09-03 09:08:59 -07:00
Leonardo de Moura
4bc1be17f4 chore: cleanup 2020-08-28 09:18:22 -07:00
Leonardo de Moura
ac565de96c refactor: add MonadMetaM class 2020-08-24 12:17:47 -07:00
Leonardo de Moura
ae5b4defb1 chore: simplify isExprDefEqStuck and isLevelDefEqStuck 2020-08-21 12:03:51 -07:00
Leonardo de Moura
05a0e7f6d0 refactor: build all main monads on top of ECoreM 2020-08-20 18:36:04 -07:00
Leonardo de Moura
249bda16c0 chore: remove prelude commands from Lean package 2020-06-25 11:21:17 -07:00
Leonardo de Moura
4ccc3fef52 chore: move Init.Lean files to Lean package 2020-05-26 15:04:35 -07:00
Renamed from src/Init/Lean/Meta/Check.lean (Browse further)