This modification is relevant for fixing regressions on recent changes
to the auto implicit behavior for inductive families.
The following declarations are now accepted:
```lean
inductive HasType : Fin n → Vector Ty n → Ty → Type where
| stop : HasType 0 (ty :: ctx) ty
| pop : HasType k ctx ty → HasType k.succ (u :: ctx) ty
inductive Sublist : List α → List α → Prop
| slnil : Sublist [] []
| cons l₁ l₂ a : Sublist l₁ l₂ → Sublist l₁ (a :: l₂)
| cons2 l₁ l₂ a : Sublist l₁ l₂ → Sublist (a :: l₁) (a :: l₂)
inductive Lst : Type u → Type u
| nil : Lst α
| cons : α → Lst α → Lst α
```
TODO: universe inference for `inductive` should be improved. The
current approach is not good enough when we have auto implicits.
TODO: allow implicit fixed indices that do not depend on indices that
cannot be moved to become parameters.
Since we are going to make `mkMatcher` reversible, it's quite useful to have the input be a `structure`. This way we make sure, that the inverse function always returns the same type as `mkMatcher` needs as input.
@Kha This one required a bunch of manual fixes. The main issue is that
before we added the string interpolation feature, we created
`MessageData`s using `++` and coercions. For example, given
`(e : Expr)`, we would write
```
let msg : MessageData := "type: " ++ e
```
and rely on the coercions `String -> MessageData` and
`Expr -> MessageData`, and the instance `Append MessageData`.
However, heterogeneous operators "block" the expected type propagation downwards.
This kind of code is obsolete now since we can write a more compact
version using string interpolation
```
let msg := m!"type: {e}"
```
@Kha This is a first attempt to improve the error message for examples
like the one Andrew Kent posted on Zulip.
I created a simpler example using "vectors".
We go back to the original approach where we pattern matching
alternative variables as `FVar`s.
We fix the original problem we had by implementing a simple
unification procedure for alternative `FVar`s.
