This PR fixes the syntax of `grind` modifiers to use `patternIgnore` for
cases where both unicode and ascii variants are matched. This fixes an
issue where several variants of grind syntax weren't accepted (e.g.
`@[grind ← gen]`). Additionally, this reduces the chance that we get
another syntax matching bootstrap hell.
This PR generalizes the `a^(m+n)` grind normalizer to any semirings.
Example:
```
variable [Field R]
example (M : R) (h₀ : M ≠ 0) {n : Nat} (hn : n > 0) : M ^ n / M = M ^ (n - 1) := by
cases n <;> grind
```
This PR improves the “expected type mismatch” error message by omitting
the type's types when they are defeq, and putting them into separate
lines when not.
I found it rather tediuos to parse the error message when the expected
type is long, because I had to find the `:` in the middle of a large
expression somewhere. Also, when both are of sort `Prop` or `Type` it
doesn't add much value to print the sort (and it’s only one hover away
anyways).
This PR enables transforming nondependent `let`s into `have`s in a
number of contexts: the bodies of nonrecursive definitions, equation
lemmas, smart unfolding definitions, and types of theorems. A motivation
for this change is that when zeta reduction is disabled, `simp` can only
effectively rewrite `have` expressions (e.g. `split` uses `simp` with
zeta reduction disabled), and so we cache the nondependence calculations
by transforming `let`s to `have`s. The transformation can be disabled
using `set_option cleanup.letToHave false`.
Uses `Meta.letToHave`, introduced in #8954.
This PR uses the commutative ring module to normalize nonlinear
polynomials in `grind cutsat`. Examples:
```lean
example (a b : Nat) (h₁ : a + 1 ≠ a * b * a) (h₂ : a * a * b ≤ a + 1) : b * a^2 < a + 1 := by
grind
example (a b c : Int) (h₁ : a + 1 + c = b * a) (h₂ : c + 2*b*a = 0) : 6 * a * b - 2 * a ≤ 2 := by
grind
```
This PR implements support for the type class `LawfulEqCmp`. Examples:
```lean
example (a b c : Vector (List Nat) n)
: b = c → a.compareLex (List.compareLex compare) b = o → o = .eq → a = c := by
grind
example [Ord α] [Std.LawfulEqCmp (compare : α → α → Ordering)] (a b c : Array (Vector (List α) n))
: b = c → o = .eq → a.compareLex (Vector.compareLex (List.compareLex compare)) b = o → a = c := by
grind
```
This PR implements support for equations `<num> = 0` in rings and fields
of unknown characteristic. Examples:
```lean
example [Field α] (a : α) : (2 * a)⁻¹ = a⁻¹ / 2 := by grind
example [Field α] (a : α) : (2 : α) ≠ 0 → 1 / a + 1 / (2 * a) = 3 / (2 * a) := by grind
example [CommRing α] (a b : α) (h₁ : a + 2 = a) (h₂ : 2*b + a = 0) : a = 0 := by
grind
example [CommRing α] (a b : α) (h₁ : a + 6 = a) (h₂ : b + 9 = b) (h₂ : 3*b + a = 0) : a = 0 := by
grind
example [CommRing α] (a b : α) (h₁ : a + 6 = a) (h₂ : b + 9 = b) (h₂ : 3*b + a = 0) : a = 0 := by
grind
example [CommRing α] (a b : α) (h₁ : a + 2 = a) (h₂ : b = 0) : 4*a + b = 0 := by
grind
example [CommRing α] (a b c : α) (h₁ : a + 6 = a) (h₂ : c = c + 9) (h : b + 3*c = 0) : 27*a + b = 0 := by
grind
```
This PR introduces a simple variable-reordering heuristic for `cutsat`.
It is needed by the `ToInt` adapter to support finite types such as
`UInt64`. The current encoding into `Int` produces large coefficients,
which can enlarge the search space when an unfavorable variable order is
used. Example:
```lean
example (a b c : UInt64) : a ≤ 2 → b ≤ 3 → c - a - b = 0 → c ≤ 5 := by
grind
```
This PR implements support for equalities and disequalities in `grind
cutsat`. We still have to improve the encoding. Examples:
```lean
example (a b c : Fin 11) : a ≤ 2 → b ≤ 3 → c = a + b → c ≤ 5 := by
grind
example (a : Fin 2) : a ≠ 0 → a ≠ 1 → False := by
grind
```
This PR replaces all usages of `[:]` slice notation in `src` with the
new `[...]` notation in production code, tests and comments. The
underlying implementation of the `Subarray` functions stays the same.
Notation cheat sheet:
* `*...*` is the doubly-unbounded range.
* `*...a` or `*...<a` contains all elements that are less than `a`.
* `*...=a` contains all elements that are less than or equal to `a`.
* `a...*` contains all elements that are greater than or equal to `a`.
* `a...b` or `a...<b` contains all elements that are greater than or
equal to `a` and less than `b`.
* `a...=b` contains all elements that are greater than or equal to `a`
and less than or equal to `b`.
* `a<...*` contains all elements that are greater than `a`.
* `a<...b` or `a<...<b` contains all elements that are greater than `a`
and less than `b`.
* `a<...=b` contains all elements that are greater than `a` and less
than or equal to `b`.
Benchmarks have shown that importing the iterator-backed parts of the
polymorphic slice library in `Init` impacts build performance. This PR
avoids this problem by separating those parts of the library that do not
rely on iterators from those those that do. Whereever the new slice
notation is used, only the iterator-independent files are imported.
This PR implements support for strict inequalities in the `ToInt`
adapter used in `grind cutsat`. Example:
```lean
example (a b c : Fin 11) : c ≤ 9 → a ≤ b → b < c → a < c + 1 := by
grind
```
This PR implements support for (non strict) `ToInt` inequalities in
`grind cutsat`. `grind cutsat` can solve simple problems such as:
```lean
example (a b c : Fin 11) : a ≤ b → b ≤ c → a ≤ c := by
grind
example (a b c : Fin 11) : c ≤ 9 → a ≤ b → b ≤ c → a ≤ c + 1 := by
grind
example (a b c : UInt8) : a ≤ b → b ≤ c → a ≤ c := by
grind
example (a b c d : UInt32) : a ≤ b → b ≤ c → c ≤ d → a ≤ d := by
grind
```
Next step: strict inequalities, and equalities.
This PR adds the following features to `simp`:
- A routine for simplifying `have` telescopes in a way that avoids
quadratic complexity arising from locally nameless expression
representations, like what #6220 did for `letFun` telescopes.
Furthermore, simp converts `letFun`s into `have`s (nondependent lets),
and we remove the #6220 routine since we are moving away from `letFun`
encodings of nondependent lets.
- A `+letToHave` configuration option (enabled by default) that converts
lets into haves when possible, when `-zeta` is set. Previously Lean
would need to do a full typecheck of the bodies of `let`s, but the
`letToHave` procedure can skip checking some subexpressions, and it
modifies the `let`s in an entire expression at once rather than one at a
time.
- A `+zetaHave` configuration option, to turn off zeta reduction of
`have`s specifically. The motivation is that dependent `let`s can only
be dsimped by let, so zeta reducing just the dependent lets is a
reasonable way to make progress. The `+zetaHave` option is also added to
the meta configuration.
- When `simp` is zeta reducing, it now uses an algorithm that avoids
complexity quadratic in the depth of the let telescope.
- Additionally, the zeta reduction routines in `simp`, `whnf`, and
`isDefEq` now all are consistent with how they apply the `zeta`,
`zetaHave`, and `zetaUnused` configurations.
The `letToFun` option is addressing a TODO in `getSimpLetCase` ("handle
a block of nested let decls in a single pass if this becomes a
performance problem").
Performance should be compared to before #8804, which temporarily
disabled the #6220 optimizations for `letFun` telescopes.
Good kernel performance depends on carefully handling the `have`
encoding. Due to the way the kernel instantiates bvars (it does *not*
beta reduce when instantiating), we cannot use congruence theorems of
the form `(have x := v; f x) = (have x ;= v'; f' x)`, since the bodies
of the `have`s will not be syntactically equal, which triggers zeta
reduction in the kernel in `is_def_eq`. Instead, we work with `f v = f'
v'`, where `f` and `f'` are lambda expressions. There is still zeta
reduction, but only when converting between these two forms at the
outset of the generated proof.
This PR refactors the juggling of universes in the linear
`noConfusionType` construction: Instead of using `PUnit.{…} → ` in the
to get the branches of `withCtorType` to the same universe level, we use
`PULift`.
This fixes https://github.com/leanprover/lean4/issues/8962, although
probably doesn’t solve all issues of that kind while level equality
checking is incomplete.
This PR add instances showing that the Grothendieck (i.e. additive)
envelope of a semiring is an ordered ring if the original semiring is
ordered (and satisfies ExistsAddOfLE), and in this case the embedding is
monotone.
This PR improves the case splitting strategy used in `grind`, and
ensures `grind` also considers simple `match`-conditions for
case-splitting. Example:
```lean
example (x y : Nat)
: 0 < match x, y with
| 0, 0 => 1
| _, _ => x + y := by -- x or y must be greater than 0
grind
```
This PR adds a procedure that efficiently transforms `let` expressions
into `have` expressions (`Meta.letToHave`). This is exposed as the
`let_to_have` tactic.
It uses the `withTrackingZetaDelta` technique: the expression is
typechecked, and any `let` variables that don't enter the zeta delta set
are nondependent. The procedure uses a number of heuristics to limit the
amount of typechecking performed. For example, it is ok to skip
subexpressions that do not contain fvars, mvars, or `let`s.
This PR implements support for normalization for commutative semirings
that do not implement `AddRightCancel`. Examples:
```lean
variable (R : Type u) [CommSemiring R]
example (a b c : R) : a * (b + c) = a * c + b * a := by grind
example (a b : R) : (a + b)^2 = a^2 + 2 * a * b + b^2 := by grind
example (a b : R) : (a + 2 * b)^2 = a^2 + 4 * a * b + 4 * b^2 := by grind
example (a b : R) : (a + 2 * b)^2 = 4 * b^2 + b * 4 * a + a^2 := by grind
```
This PR allows `simp` to recognize and warn about simp lemmas that are
likely looping in the current simp set. It does so automatically
whenever simplification fails with the dreaded “max recursion depth”
error fails, but it can be made to do it always with `set_option
linter.loopingSimpArgs true`. This check is not on by default because it
is somewhat costly, and can warn about simp calls that still happen to
work.
This closes#5111. In the end, this implemented much simpler logic than
described there (and tried in the abandoned #8688; see that PR
description for more background information), but it didn’t work as well
as I thought. The current logic is:
“Simplify the RHS of the simp theorem, complain if that fails”.
It is a reasonable policy for a Lean project to say that all simp
invocation should be so that this linter does not complain. Often it is
just a matter of explicitly disabling some simp theorems from the
default simp set, to make it clear and robust that in this call, we do
not want them to trigger. But given that often such simp call happen to
work, it’s too pedantic to impose it on everyone.
This PR implements first-class support for nondependent let expressions
in the elaborator; recall that a let expression `let x : t := v; b` is
called *nondependent* if `fun x : t => b` typechecks, and the notation
for a nondependent let expression is `have x := v; b`. Previously we
encoded `have` using the `letFun` function, but now we make use of the
`nondep` flag in the `Expr.letE` constructor for the encoding. This has
been given full support throughout the metaprogramming interface and the
elaborator. Key changes to the metaprogramming interface:
- Local context `ldecl`s with `nondep := true` are generally treated as
`cdecl`s. This is because in the body of a `have` expression the
variable is opaque. Functions like `LocalDecl.isLet` by default return
`false` for nondependent `ldecl`s. In the rare case where it is needed,
they take an additional optional `allowNondep : Bool` flag (defaults to
`false`) if the variable is being processed in a context where the value
is relevant.
- Functions such as `mkLetFVars` by default generalize nondependent let
variables and create lambda expressions for them. The
`generalizeNondepLet` flag (default true) can be set to false if `have`
expressions should be produced instead. **Breaking change:** Uses of
`letLambdaTelescope`/`mkLetFVars` need to use `generalizeNondepLet :=
false`. See the next item.
- There are now some mapping functions to make telescoping operations
more convenient. See `mapLetTelescope` and `mapLambdaLetTelescope`.
There is also `mapLetDecl` as a counterpart to `withLetDecl` for
creating `let`/`have` expressions.
- Important note about the `generalizeNondepLet` flag: it should only be
used for variables in a local context that the metaprogram "owns". Since
nondependent let variables are treated as constants in most cases, the
`value` field might refer to variables that do not exist, if for example
those variables were cleared or reverted. Using `mapLetDecl` is always
fine.
- The simplifier will cache its let dependence calculations in the
nondep field of let expressions.
- The `intro` tactic still produces *dependent* local variables. Given
that the simplifier will transform lets into haves, it would be
surprising if that would prevent `intro` from creating a local variable
whose value cannot be used.
Note that nondependence of lets is not checked by the kernel. To
external checker authors: If the elaborator gets the nondep flag wrong,
we consider this to be an elaborator error. Feel free to typecheck `letE
n t v b true` as if it were `app (lam n t b default) v` and please
report issues.
This PR follows up from #8751, which made sure the nondep flag was
preserved in the C++ interface.
This PR adds a linter (`linter.unusedSimpArgs`) that complains when a
simp argument (`simp [foo]`) is unused. It should do the right thing if
the `simp` invocation is run multiple times, e.g. inside `all_goals`. It
does not trigger when the `simp` call is inside a macro. The linter
message contains a clickable hint to remove the simp argument.
I chose to display a separate warning for each unused argument. This
means that the user has to click multiple times to remove all of them
(and wait for re-elaboration in between). But this just means multiple
endorphine kicks, and the main benefit over a single warning that would
have to span the whole argument list is that already the squigglies tell
the users about unused arguments.
This closes#4483.
Making Init and Std clean wrt to this linter revealed close to 1000
unused simp args, a pleasant experience for anyone enjoying tidying
things: #8905
This PR modifies `let` and `have` term syntaxes to be consistent with
each other. Adds configuration options; for example, `have` is
equivalent to `let +nondep`, for *nondependent* lets. Other options
include `+usedOnly` (for `let_tmp`), `+zeta` (for `letI`/`haveI`), and
`+postponeValue` (for `let_delayed)`. There is also `let (eq := h) x :=
v; b` for introducing `h : x = v` when elaborating `b`. The `eq` option
works for pattern matching as well, for example `let (eq := h) (x, y) :=
p; b`.
Future PRs will add these options to tactic syntax, once a stage0 update
has been done.
This PR implements support for (commutative) semirings in `grind`. It
uses the Grothendieck completion to construct a (commutative) ring
`Lean.Grind.Ring.OfSemiring.Q α` from a (commutative) semiring `α`. This
construction is mostly useful for semirings that implement
`AddRightCancel α`. Otherwise, the function `toQ` is not injective.
Examples:
```lean
example (x y : Nat) : x^2*y = 1 → x*y^2 = y → y*x = 1 := by
grind
example [CommSemiring α] [AddRightCancel α] (x y : α) : x^2*y = 1 → x*y^2 = y → y*x = 1 := by
grind
example (a b : Nat) : 3 * a * b = a * b * 3 := by grind
example (k z : Nat) : k * (z * 2 * (z * 2 + 1)) = z * (k * (2 * (z * 2 + 1))) := by grind
example [CommSemiring α] [AddRightCancel α] [IsCharP α 0] (x y : α)
: x^2*y = 1 → x*y^2 = y → x + y = 1 → False := by
grind
```
This PR makes `simp` consult its own cache more often, to avoid
replicating work.
Before, the simp cache was checked upon entry of `simpImpl` only, which
then calls `simpLoop`, which recursively iterates the `pre`-lemmas,
without checking the cache again.
Now, `simpLoop` itself checks the cache. This seems more principled,
given that `simpLoop` is actually putting entries into the cache for
each of its calls, so it’s more uniform if it checks the cache itself.
This avoids repeated rewrites. For example given
```
theorem ab : a = b := testSorry
theorem bc : b = c := testSorry
example (h : P c) : P b ∧ P a := by simp [ab, bc, h]
```
simp would rewrite `b ==> c` twice (once as part of `b ==> c` and then
again as part of `a ==> b ==> c`). And it’d be order dependent: With
```
example (h : P c) : P a ∧ P b := by simp [ab, bc, h]
```
the `a ==> b ==> c` chain would insert `b ==> c` into the cache, and
picked up by `simpImpl` when rewriting `P b`.
With this change, `b ==> c` is performed only once in both examples.
Instruction counts on stdlib and mathlib both show a mild improvement
across the board (0.5%), with individual modules improving by up to 4%
in stdlib and even more in mathlib.
(This does not check the cache before applying `post`, which explains
where there are still some repeated rewrites in the trace logs. But I’m
less sure about inserting a cache check here and so I am treading
carefully here. It’s also going to be at most one `post` application
that’s duplicated, because if `post` returns `.visit`, we go back to
`pre` and thus a cache check.)
This PR refactors the way simp arguments are elaborated: Instead of
changing the `SimpTheorems` structure as we go, this elaborates each
argument to a more declarative description of what it does, and then
apply those. This enables more interesting checks of simp arguments that
need to happen in the context of the eventually constructed simp context
(the checks in #8688), or after simp has run (unused argument linter
#8901).
The new data structure describing an elaborated simp argument isn’t the
most elegant, but follows from the code.
While I am at it, move handling of `[*]` into `elabSimpArgs`. Downstream
adaption branches exist (but may not be fully up to date because of the
permission changes).
While I am at it, I cleaned up `SimpTheorems.lean` file a bit (sorting
declarations, mild renaming) and added documentation.
This PR make sure that the local instance cache calculation applies more
reductions. In #2199 there was an issue where metavariables could
prevent local variables from being considered as local instances. We use
a slightly different approach that ensures that, for example, `let`s at
the ends of telescopes do not cause similar problems. These reductions
were already being calculated, so this does not require any additional
work to be done.
Metaprogramming interface addition: the various forall telescope
functions that do reduction now have a `whnfType` flag (default false).
If it's true, then the callback `k` is given the WHNF of the type. This
is a free operation, since the telescope function already computes it.
This PR refactors `Lean.Grind.NatModule/IntModule/Ring.IsOrdered`.
We ensure the the diamond from `Ring` to `NatModule` via either
`Semiring` or `IntModule` is defeq, which was not previously the case.
---------
Co-authored-by: Leonardo de Moura <leomoura@amazon.com>
