This PR completes the review of `@[grind]` annotations without a sigil
(e.g. `=` or `←`), replacing most of them with more specific annotations
or patterns.
---------
Co-authored-by: Leonardo de Moura <leomoura@amazon.com>
This PR implements a new E-matching pattern inference procedure that is
faithful to the behavior documented in the reference manual regarding
minimal indexable subexpressions. The old inference procedure was
failing to enforce this condition. For example, the manual documents
`[grind ->]` as follows
`[@grind →]` selects a multi-pattern from the hypotheses of the theorem.
In other words, `grind` will use the theorem for forwards reasoning.
To generate a pattern, it traverses the hypotheses of the theorem from
left to right. Each time it encounters a **minimal indexable
subexpression** which covers an argument which was not previously
covered, it adds that subexpression as a pattern, until all arguments
have been covered.
That said, the new procedure is currently disabled, and the following
option must be used to enable it.
```
set_option backward.grind.inferPattern false
```
Users can inspect differences between the old a new procedures using the
option
```
set_option backward.grind.checkInferPatternDiscrepancy true
```
Example:
```lean
/--
warning: found discrepancy between old and new `grind` pattern inference procedures, old:
[@List.length #2 (@toList _ #1#0)]
new:
[@toList #2#1#0]
use `set_option backward.grind.inferPattern true` to force old procedure
-/
#guard_msgs in
set_option backward.grind.checkInferPatternDiscrepancy true in
@[grind] theorem Vector.length_toList' (xs : Vector α n) : xs.toList.length = n := by sorry
```
This PR moves the definitions and basic facts about `Function.Injective`
and `Function.Surjective` up from Mathlib. We can do a better job of
arguing via injectivity in `grind` if these are available.
This PR updates `@[grind]` annotations which should be `@[grind =]`, for
robustness (and, presumably, in some fraction of cases the existing
heuristic for `@[grind]` is already too liberal).
This PR shares common functionality relate to equalities between same
constructors, and when these are type-correct. In particular it uses the
more complete logic from `mkInjectivityThm` also in other places, such
as `CasesOnSameCtor` and the deriving code for `BEq`, `DecidableEq`,
`Ord`, for more consistency and better error messages.
This PR implements `mkNoConfusionImp` in Lean rather than in C. This
reduces our reliance on C, and may bring performance benefits from not
reducing `noConfusionType` during elaboration time (it still gets
reduced by the kernel when type-checking).
This PR makes it possible to write custom interpolation notation which
treats interpolated `String`s specially.
Sometimes it is desirable for `let w := "world"; foo!"hello {w}"` and
`foo!"hello world"` to mean different things; for instance, if debugging
and wanting to show all interpolands with `repr`. The current approach
forces `hello` to also be rendered with `repr`, which is not desirable.
This doesn't modify any existing formatters.
Requested in [#lean4 > ✔ dbg_trace should use `Repr` instance @
💬](https://leanprover.zulipchat.com/#narrow/channel/270676-lean4/topic/.E2.9C.94.20dbg_trace.20should.20use.20.60Repr.60.20instance/near/495082575)
---------
Co-authored-by: Sebastian Ullrich <sebasti@nullri.ch>
This PR upstreams the Verso parser and adds preliminary support for
Verso in docstrings. This will allow the compiler to check examples and
cross-references in documentation.
After a `stage0` update, a follow-up PR will add the appropriate
attributes that allow the feature to be used. The parser tests from
Verso also remain to be upstreamed, and user-facing documentation will
be added once the feature has been used on more internals.
This PR fixes a performance issue in `grind linarith`. It was creating
unnecessary `NatModule`/`IntModule` structures for commutative rings
without an order. This kind of type should be handled by `grind ring`
only.
This PR implements model-based theory combination for types `A` which
implement the `ToInt` interface. Examples:
```lean
example {C : Type} (h : Fin 4 → C) (x : Fin 4)
: 3 ≤ x → x ≤ 3 → h x = h (-1) := by
grind
example {C : Type} (h : UInt8 → C) (x y z w : UInt8)
: y + 1 + w ≤ x + w → x + w ≤ z → z ≤ y + w + 1 → h (x + w) = h (y + w + 1) := by
grind
example {C : Type} (h : Fin 8 → C) (x y w r : Fin 8)
: y + 1 + w ≤ r → r ≤ y + w + x → x = 1 → h r = h (y + w + 1) := by
grind
```
This PR removes `grind →` annotations that fire too often, unhelpfully.
It would be nice for `grind` to instantiate these lemmas, but only if
they already see `xs ++ ys` and `#[]` in the same equivalence class, not
just as soon as it sees `xs ++ ys`.
In the meantime, let's see what is using these.
This PR introduces limited functionality frontends `cutsat` and
`grobner` for `grind`. We disable theorem instantiation (and case
splitting for `grobner`), and turn off all other solvers. Both still
allow `grind` configuration options, so for example one can use `cutsat
+ring` (or `grobner +cutsat`) to solve problems that require both.
For `cutsat`, it is helpful to instantiate a limited set of theorems
(e.g. `Nat.max_def`). Currently this isn't supported, but we intend to
add this later.
This PR fixes the `grind` canonicalizer for `OfNat.ofNat` applications.
Example:
```lean
example {C : Type} (h : Fin 2 → C) :
-- `0` in the first `OfNat.ofNat` is not a raw literal
h (@OfNat.ofNat (Fin (1 + 1)) 0 Fin.instOfNat) = h 0 := by
grind
```
This PR ensures that the auxiliary temporary metavariable IDs created by
the E-matching module used in `grind` are not affected by what has been
executed before invoking `grind`. The goal is to increase `grind`’s
robustness.
For example, in the E-matching module we use `Expr.quickLt` to sort
candidates. `Expr.quickLt` depends on the `Expr` hash code, which in
turn depends on metavariable IDs. Thus, before this change, the initial
next metavariable ID at the time of `grind` invocation could affect the
order in which instances were generated, and consequently the `grind`
search.
This PR changes the string interpolation procedure to omit redundant
empty parts. For example `s!"{1}{2}"` previously elaborated to `toString
"" ++ toString 1 ++ toString "" ++ toString 2 ++ toString ""` and now
elaborates to `toString 1 ++ toString 2`.
- [x] Updated docstrings for `simp!`, `simp_all!`, `dsimp!` to use
user-friendly language
- [x] Updated docstrings for `autoUnfold` fields to use user-friendly
language
- [x] Fixed broken test by updating expected output for simp! hover
documentation
- [x] Replaced technical terms with clear language: "will unfold
applications of functions defined by pattern matching, when one of the
patterns applies"
<!-- START COPILOT CODING AGENT TIPS -->
---
💡 You can make Copilot smarter by setting up custom instructions,
customizing its development environment and configuring Model Context
Protocol (MCP) servers. Learn more [Copilot coding agent
tips](https://gh.io/copilot-coding-agent-tips) in the docs.
---------
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: nomeata <148037+nomeata@users.noreply.github.com>
This PR adds missing the lemmas `ofList_eq_insertMany_empty`,
`get?_eq_some_iff`, `getElem?_eq_some_iff` and `getKey?_eq_some_iff` to
all container types.
This PR adds missing `grind` normalization rules for `natCast` and
`intCast` Examples:
```
open Lean.Grind
variable (R : Type) (a b : R)
section CommSemiring
variable [CommSemiring R]
example (m n : Nat) : (m + n) • a = m • a + n • a := by grind
example (m n : Nat) : (m * n) • a = m • (n • a) := by grind
end CommSemiring
section CommRing
variable [CommRing R]
example (m n : Nat) : (m + n) • a = m • a + n • a := by grind
example (m n : Nat) : (m * n) • a = m • (n • a) := by grind
example (m n : Int) : (m * n) • (a * b) = (m • a) * (n • b) := by grind
end CommRing
```
This PR makes `IO.RealWorld` opaque. It also adds a new compiler -only
`lcRealWorld` constant to represent this type within the compiler. By
default, an opaque type definition is treated like `lcAny`, whereas we
want a more efficient representation. At the moment, this isn't a big
difference, but in the future we would like to completely erase
`IO.RealWorld` at runtime.
This PR offers an alternative `noConfusion` construction for the
off-diagonal use (i.e. for different constructors), based on comparing
the `.ctorIdx`. This should lead to faster type checking, as the kernel
only has to reduce `.ctorIdx` twice, instead of the complicate
`noConfusionType` construction.
This PR changes the "declaration uses 'sorry'" error to pretty print an
actual `sorry` expression in the message. The effect is that the `sorry`
is hoverable and, if it's labeled, you can "go to definition" to see
where it came from.
The implementation prefers reporting synthetic sorries. These can appear
even if there are no error messages if a declaration refers to a
declaration that has elaboration errors. Users should focus on
elaboration errors before worrying about user-written `sorry`s.
In the future we could have some more precise logic for sorry reporting.
All the sorries in a declaration should be considered to be reported,
and we should not re-report sorries in later declarations. Some
elaborators use `warn.sorry` to avoid re-reporting sorries in auxiliary
declarations.
This PR changes the implementation of a function `unfoldPredRel` used in
(co)inductive predicate machinery, that unfolds pointwise order on
predicates to quantifications and implications. Previous implementation
relied on `withDeclsDND` that could not deal with types which depend on
each other. This caused the following example to fail:
```lean4
inductive infSeq_functor1.{u} {α : Type u} (r : α → α → Prop) (call : {α : Type u} → (r : α → α → Prop) → α → Prop) : α → Prop where
| step : r a b → infSeq_functor1 r call b → infSeq_functor1 r call a
def infSeq1 (r : α → α → Prop) : α → Prop := infSeq_functor1 r (infSeq1)
coinductive_fixpoint monotonicity by sorry
#check infSeq1.coinduct
```
Closes#10234.
This PR completes the `grind` solver extension design and ports the
`grind ac` solver to the new framework. Future PRs will document the API
and port the remaining solvers. An additional benefit of the new design
is faster build times.
The proof of the instWPMonad instance relies on the equality of any two
terms of type `IO.RealWorld`, which is only a side effect of the current
transparent definition. Ignoring the questions around the utility of
proving things about programs in `IO`, the semantic validity of this
instance in the intended model of the IO monad is also unclear.
I tried a few things to axiomatize this instance so it could be put into
the test file to preserve the one test section that relies on it, but I
was unsuccessful; everything I attempted caused errors.
This PR adds infrastructure for registering new `grind` solvers. `grind`
already includes many solvers, and this PR is the first step toward
modularizing the design and supporting user-defined solvers.
This PR prepares for a future reorganization of the import hierarchy so
that `Init.Data.String.Basic` can import `Init.Data.UInt.Bitwise` and
`Init.Data.Array.Lemmas`.
This PR moves `String.utf8EncodeChar` to the prelude to prepare for the
imminent redefinition of `String`.
The definition in the prelude uses modulo and division operations on
natural numbers. In `String.Extra`, a `csimp` lemma is provided, showing
that the new definition is equal to the previous one (which is now
called `utf8EncodeCharFast`) which uses bitwise operations on `UInt8`.
This PR implements diagnostic information for the `grind ac` module. It
now displays the basis, normalized disequalities, and additional
properties detected for each associative operator.
This PR improves the counterexamples produced by `grind linarith` for
`NatModule`s. `grind` now hides occurrences of the auxiliary function
`Grind.IntModule.OfNatModule.toQ`.
This PR implements `NatModule` normalization when the `AddRightCancel`
instance is not available. Note that in this case, the embedding into
`IntModule` is not injective. Therefore, we use a custom normalizer,
similar to the `CommSemiring` normalizer used in the `grind ring`
module. Example:
```lean
open Lean Grind
example [NatModule α] (a b c : α)
: 2•a + 2•(b + 2•c) + 3•a = 4•a + c + 2•b + 3•c + a := by
grind
```
