This PR makes the eager lambda lifting heuristic more predictable by
blocking it from lifting from
any kind of inlineable function, not just `@[inline]`. It also adapts
the doc-string to describe
what is actually going on.
This PR inlines several Decidable instances for performance reasons.
Unlike the previous #10934 it does not attempt to also simplify the
Decidable instance system as
that has proven to have non trivial performance impact.
Co-authored-by: Rob23oba <152706811+Rob23oba@users.noreply.github.com>
This PR fixes a memleak caused by the Lean based `IO.waitAny`
implementation by reverting it.
This the faulty Lean implementation:
```lean
def IO.waitAny (tasks : @& List (Task α)) (h : tasks.length > 0 := by exact Nat.zero_lt_succ _) :
BaseIO α := do
have : Nonempty α := ⟨tasks[0].get⟩
let promise : IO.Promise α ← IO.Promise.new
tasks.forM <| fun t => BaseIO.chainTask (sync := true) t promise.resolve
return promise.result!.get
```
In a situation where we call this function repeatedly in a loop with a
pair of tasks `[t1, t2]`
where `t2` is a long lived task that we pass every time and `t1` is
fresh a short lived task, `t2` will
accumlate more and more children from `BaseIO.chainTask` that fill
memory over time. The old C++
implementation did not have this issue so we are reverting.
This PR defines `String.Slice.replace` and redefines `String.replace` to
use the `Slice` version.
The new implementation is generic in the pattern, so it supports things
like `"education".replace isVowel "☃!" = "☃!d☃!c☃!t☃!☃!n"`. Since it
uses the `ForwardSearcher` infrastructure, `String` patterns are
searched using KMP, unlike the previous implementation which had
quadratic runtime. As a side effect, the behavior when replacing an
empty string now matches that of most other programming languages,
namely `"abc".replace "" "k" = "kakbkck"`.
This PR fixes the KMP implementation, which did incorrect bookkeeping of
the backtracking process, leading to incorrect starting ranges of
matches.
The new implementation does not require `partial` anywhere.
This PR adds support for specifying anchors to restrict the search space
in `grind` when using `grind only`. Anchors can limit which case splits
are performed and which local lemmas are instantiated.
This PR adds the `set_config` tactic for setting `grind` configuration
options. It uses the same syntax used for setting configuration options
in the `grind` main tactic.
This PR tries to preserve names of pattern variables in match
alternatives in `decreasing_by`, by telescoping into the concrete
alternative rather than the type of the matcher's alt. Fixes#10976.
This PR ensures that searching for an empty string returns the expected
pattern of alternating size-zero matches and size-one rejects.
In particular, splitting by an empty string returns an array formed of
the empty string, all of the string's characters as singleton strings,
followed by another empty string. This matches the [Rust
behavior](https://doc.rust-lang.org/std/primitive.str.html#method.split),
for example.
This PR adds inline annotations to several `Decidable` instances.
Additionally, it removes the `Decidable` instance for `p → q` which is
made redundant by `forall_prop_decidable`.
This PR changes the closure allocator to use the general allocator
instead of the small object one.
This is because users may create closures with a gigantic amount of
closed variables which in turn
boost the size of the closure beyond the small object threshold.
This issue was uncovered by #10979. Detecting that the small object
threshold is at fault requires
building mimalloc in debug mode at which point it yields:
```
mimalloc: assertion failed: at "/home/henrik/lean4/build/debug/mimalloc/src/mimalloc/src/alloc.c":132, mi_heap_malloc_small_zero
assertion: "size <= MI_SMALL_SIZE_MAX"
```
The generated code at fault here looks as follows:
```c
LEAN_EXPORT lean_object* l_initExec___at___00res_spec__0(lean_object* x_1) {
_start:
{
lean_object* x_2; lean_object* x_3; lean_object* x_4; lean_object* x_5; lean_object* x_6; lean_object* x_7; lean_object* x_8; lean_object* x_9; lean_object* x_10; lean_object* x_11; lean_object* x_12; lean_object* x_13; lean_object* x_14;
x_2 = lean_alloc_closure((void*)(l_initializer_ext___at___00initExec___at___00res_spec__0_spec__0___lam__0___boxed), 3, 0);
x_3 = l_initExec___redArg___closed__0;
x_4 = l_initExec___redArg___closed__1;
x_5 = l_instMonadLiftNonDetT___closed__0;
x_6 = l_initExec___redArg___closed__2;
x_7 = l_initExec___at___00res_spec__0___closed__0;
lean_inc_ref(x_2);
x_8 = lean_alloc_closure((void*)(l_initExec___at___00res_spec__0___lam__29___boxed), 213, 212);
lean_closure_set(x_8, 0, x_3);
lean_closure_set(x_8, 1, x_2);
lean_closure_set(x_8, 2, x_4);
lean_closure_set(x_8, 3, x_3);
lean_closure_set(x_8, 4, x_4);
lean_closure_set(x_8, 5, x_3);
lean_closure_set(x_8, 6, x_4);
lean_closure_set(x_8, 7, x_3);
lean_closure_set(x_8, 8, x_4);
lean_closure_set(x_8, 9, x_3);
lean_closure_set(x_8, 10, x_4);
lean_closure_set(x_8, 11, x_3);
lean_closure_set(x_8, 12, x_4);
lean_closure_set(x_8, 13, x_3);
lean_closure_set(x_8, 14, x_4);
lean_closure_set(x_8, 15, x_5);
lean_closure_set(x_8, 16, x_6);
lean_closure_set(x_8, 17, x_5);
lean_closure_set(x_8, 18, x_5);
lean_closure_set(x_8, 19, x_5);
lean_closure_set(x_8, 20, x_5);
lean_closure_set(x_8, 21, x_5);
lean_closure_set(x_8, 22, x_5);
...
```
With the crash happening in `lean_alloc_closure` where we
unconditionally invoke the small allocator
which cannot cope with closures this large. Hopefully changing this to
the general purpose allocator
doesn't have too much of an impact on performance.
Closes: #10979
This PR adds the basic infrastructure to perform termination proofs
about `String.ValidPos` and `String.Slice.Pos`.
We choose approach where the intended way to do termination arguments is
to argue about the position itself rather than some projection of it
like `remainingBytes`.
The types `String.ValidPos` and `String.Slice.Pos` are equipped with a
`WellFoundedRelation` instance given by the greater-than relation. This
means that if a function takes a position `p` and performs a recursive
call on `q`, then the decreasing obligation will be `p < q`. This works
well in the common case where `q` is `p.next h`, in which case the goal
`p < p.next h` is solved by the simplifier.
For stepping through a string backwards, we introduce a type synonym
with a `WellFoundedRelation` instance given by the less-than relation.
This means that if a function takes a position `p` and performs a
recursive call on `q` and specifies `termination_by p.down`, then the
decreasing obligation will be `q < p`. This works well in the case where
`q` is `p.prev h`, in which case the goal `p.prev h < p` is solved by
the simplifier.
For termination arguments invoving multiple strings, the lower-level
primitive `p.remainingBytes` (landing in `Nat`) is also available.
In a future PR, we will additionally provide the necessary typeclasses
instances to register `String.ValidPos` and `String.Slice.Pos` with
`grind` to make complex termination arguments more convenient in user
code.
This PR performs more widening in ElimDeadBranches in an attempt to
improve performance in situations with a lot of local precision.
While this is not enough to make the compilation instant it pushes
compilation time from 12s to 3s for the example in #10857 and barely
introduces regressions so it seems like a good first step in this
direction.
Closes: #10857
This PR implements the following `grind` improvements:
1. `set_option` can now be used to set `grind` configuration options in
the interactive mode.
2. Fixes a bug in the repeated theorem instantiation detection.
3. Adds the macro `use [...]` as a shorthand for `instantiate only
[...]`.
This PR adds the combinator ` · t_1 ... t_n` to the `grind` interactive
mode. The `finish?` tactic now generates scripts using this combinator
to conform to Mathlib coding standards. The new format is also more
compact. Example:
```lean
/--
info: Try this:
[apply] ⏎
instantiate only [= mem_indices_of_mem, insert, = getElem_def]
instantiate only [= getElem?_neg, = getElem?_pos]
cases #f590
· cases #ffdf
· instantiate only
instantiate only [= Array.getElem_set]
· instantiate only
instantiate only [size, = HashMap.mem_insert, = HashMap.getElem_insert, = Array.getElem_push]
· instantiate only [= mem_indices_of_mem, = getElem_def]
instantiate only [usr getElem_indices_lt]
instantiate only [size]
cases #ffdf
· instantiate only [=_ WF]
instantiate only [= getElem?_neg, = getElem?_pos, = Array.getElem_set]
instantiate only [WF']
· instantiate only
instantiate only [= HashMap.mem_insert, = HashMap.getElem_insert, = Array.getElem_push]
-/
#guard_msgs in
example (m : IndexMap α β) (a a' : α) (b : β) (h : a' ∈ m.insert a b) :
(m.insert a b)[a'] = if h' : a' == a then b else m[a'] := by
grind => finish?
```
This PR ensures that model-based theory combination in `grind cutsat`
considers nonlinear terms. Nonlinear multiplications such as `x * y` are
treated as uninterpreted symbols in `cutsat`.
Closes#10885
This PR adds support for scientific literals for `Rat` in `grind`.
`grind` does not yet add support for this kind of literal in arbitrary
fields.
closes#10489
This PR fixes a bug in the equality propagation procedure in
`grind.order`. Specifically, it affects the procedure that asserts
equalities in the `grind` core state that are implied by (ring)
inequalities in the `grind.order` module.
closes#10622
This is a guard against #10705; if a kernel error is raised when the
return value of this function is eventually checked, it is often
silenced downstream, making it hard to spot the failure.
If we panic here via `assert!`, then the diagnostic cannot be missed.
This PR adds the `mbtc` tactic to the `grind` interactive mode. It
implements model-based theory combination. It also ensures `finish?` is
capable of generating it.
This PR ensures that solver propagation steps are necessary in the
generated tactic script to close the goal.
It produces more compact proof scripts, but this is not just an
optimization, if we include an unnecessary step, we may fail to replay
the generated script when `cases` steps are pruned using
non-chronological backtracking (NCB). For example, when executing
`finish?`, we may have performed a `cases #<anchor>` step that enabled
`ring` to propagate a new fact. If this fact is not used in the final
proof, and the corresponding `cases #<anchor>` step is pruned by NCB,
the `ring` step will fail during replay.
