I kept a few core methods (e.g., exact_core and apply_core). Reason:
if we use default parameters
meta constant exact (e : expr) (md := semireducible) : tactic unit
then, we will not be able to write
to_expr p >>= exact
The workaround is
do t <- to_expr p, exact t
or
to_expr p >>= (fun x, exact x)
One alternative is to change how we handle default parameters, and
eta-expand applications that involve default parameters.
We may also have an attribute [eta_expand]. Then
attribute [eta_expand] foo
instructs the elaborator to automatically eta-expand foo-applications.
The attribute would give users more control, and avoid potential
performance problems. Without the attribute, then for every function
application the elaborator has to check the type and decide whether it
must be eta-expanded or not.
@gebner @kha What do you think?
Summary:
We minimize the number of "'sorry' used warning messages". We also
re-target the error to the main declaration. Example: foo._main ==> foo
We do not report for auxiliary declarations such as "_example" and
"foo.equations._eqn_1"
Get rid of the redundant error message "error : failed" for tactics.
We added "silent failures" in the tactic framework.
We do not store line/col information for tactics nested in notation
declarations. Before this commit, we would have tactics such
as (tactic.save_info line col) nested inside of notation declarations.
Before this commit, an user could define their own prelude and change
the types of quot, quot.mk, quot.lift or quot.ind.
By doing that, they could prove false.
This commit prevents this kind of abuse.
It also modifies the definition of `quot` and avoids the `setoid`
dependency.
The previous `quot` type is now called `quotient`, and it is defined
using the new `quot` type provided by the kernel.
See discussion at #1330
After this commit, new interactice tactic classes can be added without
writing C++ code (see example: tests/lean/run/my_tac_class.lean).
The tactic_evaluator was simplified, and all the complexity has been
moved to tactic_notation, and lean code.
We can now inspect the intermediate states produced by the rewrite
tactic.
The function (@scope_trace _ line col thunk) can be used to position trace
messages produced by thunk. If line/col are not provided (i.e., we
just write (scope_trace thunk)), then line/col are filled with the
position of this term by the elaborator.
We can visualize the intermediate tactic states inside nested blocks
such as (try { ... })
The new infrastructure can be used to implement custom tactic_state
pretty printers.
Before this commit, if we declared an equational lemma using 'def',
then it would not be correctly added to the simp/hinst lemma set.
The new test exposes the problem.
@kha, I added autocompletion for ^. I try to elaborate the expression
before ^. using the local context provided by the parser.
The autocompletion only works if the type for the expression before ^. can be
inferred. This is a big limitation because this information cannot be
obtained in many cases. Here are examples that do not work:
meta def proof_for (e : expr) : smt_tactic expr :=
do cc ← to_cc_state, cc^.proof_for e
--^ does not work here
If we annotate cc with its type, it works:
meta def proof_for (e : expr) : smt_tactic expr :=
do cc : cc_state ← to_cc_state, cc^.proof_for e
--^ works
We don't have typing information on the equation lhs at
autocompletion time. So, the following will not work
private meta def mk_smt_goals_for (cfg : smt_config)
: list expr → list smt_goal → list expr → tactic (list smt_goal × list expr)
| [] sr tr := return (sr^.reverse, tr^.reverse)
--^ does not work since
| (tg::tgs) sr tr := ...
@kha The following three tests were broken on my machine. The new
produced output looked correct. So, I updated the expected output files.
Is this ok?
see #1296
