chore: get cache - #42756
Draft
thorimur wants to merge 42 commits into
Draft
Conversation
thorimur
commented
Aug 14, 2026
thorimur
commented
Contributor
…via `by as_aux_lemma` Delegating to `by as_aux_lemma => exact @$t` had three problems: * `by` only leaves the exporting context when `backward.proofsInPublic` is `false` (`Lean.Elab.Term.runTactic`), so under that option `private` silently became a no-op and failed to resolve private declarations at all. This contradicted the existing comment stating we deliberately do not check that option. * `@$t` only means "no implicit lambda" for non-identifiers; for an identifier it takes the `elabAtom` branch and makes implicit arguments *explicit*, so `private` broke any lemma with leading implicit arguments. * `by` already abstracts proofs into an auxiliary theorem itself, so `as_aux_lemma` emitted a redundant second one. Elaborate the term outside the exporting context and call `mkAuxTheorem` ourselves instead. This matches what `by exact` does (`exact` is `elabTermEnsuringType` against the goal type, and `runTactic` wraps using the expected type rather than re-inferring), minus the three problems above. `withSynthesize` is required so that synthetic metavariables created while elaborating the term are solved before abstracting; otherwise `mkValueTypeClosure` abstracts them into parameters of the auxiliary theorem and the proof is hoisted back out into the exporting context. Also skip the wrapper for a local hypothesis, pass `cache := !e.hasSorry`, and make the module docstring an actual docstring (`/-!`), which silences a `linter.style.header` warning on every build. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Adds a temporary linter which finds uses of `set_option backward.privateInPublic true` that are caused by a bare term proof appearing in a position elaborated in an exporting environment, and suggests wrapping that proof in the `private` term elaborator from `Mathlib.Util.PrivateProof` instead. Detection uses a stateful linter to track the declarations of the current module, and then looks for references to the private ones which were elaborated in an exporting environment, discarding those inside `by` blocks with `Prop` goals. Suggestions are produced by slicing the original text out of the `FileMap` using the (canonical) infotree positions, so nothing is re-printed. This linter is intended to drive a one-off sweep, and is to be removed afterwards. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The linter now skips references enclosed in the `private` term elaborator or in `private_decl%`, which is what `field := private ..` in a structure instance expands to. Suggestions in structure instance fields are deliberately left unparenthesized, so that they are taken as the structure instance `private` modifier. Also restricts references to identifier syntax: wrapper nodes such as `(..)` carry the referenced constant as their `TermInfo.expr` too, and were being reported as references to it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…otal The metavariable context available to the linter is the one saved at the nearest enclosing infotree context node, not the one in force when a given term was elaborated, so `instantiateMVars` can produce expressions mentioning free variables outside the term's local context. `inferType` then fails with "unknown free variable", which the stateful linter framework reports as a linter failure. Both checks now catch such failures. Also fixes `isProofByBlock`, which ran the `Prop` check in an empty local context rather than the goal's own: the goal's type generally mentions the local hypotheses of the enclosing declaration. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
When every reference in a command can be wrapped in `private` and nothing is reported as unfixable, the linter now also walks the chain of `set_option .. in`s wrapping the command and suggests deleting each one which sets `backward.privateInPublic` to `true` or `backward.privateInPublic.warn` to `false`. Each deletion is its own message, one per `set_option` rather than one per wrap, and the suggested range runs from the `set_option` keyword through the trailing whitespace of the `in`, so applying it removes the whole line. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5 noreply@anthropic.com
github-actions
Bot
added
the
large-import
PR summary 782e4ee4e4Import changes exceeding 2%
|
| Files | Import difference |
|---|---|
| ../mathlib-ci/scripts/pr_summary/import_trans_difference.sh all | |
| There are 8269 files with changed transitive imports taking up over 361009 characters: this is too many to display! | |
You can run this locally from your mathlib4 directory: |
git clone https://github.com/leanprover-community/mathlib-ci.git ../mathlib-ci
Declarations diff (regex)
+ F
+ Foo
+ MyPreorder
+ Occurrence
+ S
+ Wrap
+ aPriv
+ aPrivBothErrors
+ aPrivSilenced
+ bPriv
+ bPrivSilenced
+ chainInner
+ chainOuter
+ check
+ commandAt?
+ eq0
+ eq0'
+ exportedPrivateNames
+ f
+ f'
+ fImplicit
+ fImplicit'
+ fLocal
+ fLocalSilenced
+ fNestedBy
+ fProofsInPublic
+ findWrap
+ fixed1
+ fixed2
+ fixed3
+ fixed4
+ fixed5
+ fixed6
+ fixed7
+ foo
+ fooLeRefl
+ fooPub
+ fooThm
+ fα
+ fα'
+ fα''
+ fαPub'
+ fαPub''
+ gImplicit
+ implicitThm
+ inst1
+ isOptionValue
+ isPrivateWrapper
+ isProofByBlock
+ isWrappableProof
+ plain
+ precZeroOpeners
+ precedingTokenIsOpener
+ publicPrivateNames
+ publicUses
+ run
+ stacked
+ stacked'
+ sup_aux
+ thm1
+ truncatedSup
+ typ1
+ unused
+ usePriv
+ used
+ usesIt
+ usesPlain
+ usesS
+ val1
+ val2
+ val3
+ val4
+ val5
+ val6
+ whose
+++ FEq
You can run this locally as follows
## from your `mathlib4` directory:
git clone https://github.com/leanprover-community/mathlib-ci.git ../mathlib-ci
## summary with just the declaration names:
../mathlib-ci/scripts/pr_summary/declarations_diff.sh <optional_commit>
## more verbose report:
../mathlib-ci/scripts/pr_summary/declarations_diff.sh long <optional_commit>The doc-module for scripts/pr_summary/declarations_diff.sh in the mathlib-ci repository contains some details about this script.
Declarations diff (Lean -- pending)
Computed after the build finishes.
Increase in strong tech debt: (relative, absolute) = (0.72, 0.73)
| Current number | Change | Type (strong) |
|---|---|---|
| 5130 | 1 | backward.isDefEq.respectTransparency |
| 2677 | 1 | backward.isDefEq.respectTransparency.types |
| 649 | -35 | backward.privateInPublic |
| 272 | -60 | backward.privateInPublic.warn |
| 1 | 1 | backward.proofsInPublic |
Current commit 782e4ee4e4
Reference commit 4d476f6d44
This script lives in the mathlib-ci repository. To run it locally, from your mathlib4 directory:
git clone https://github.com/leanprover-community/mathlib-ci.git ../mathlib-ci
../mathlib-ci/scripts/reporting/technical-debt-metrics.sh pr_summary
- The
relativevalue is the weighted sum of the differences with weight given by the inverse of the current value of the statistic. - The
absolutevalue is therelativevalue divided by the total sum of the inverses of the current values (i.e. the weighted average of the differences).
…f dot notation (data is being constructed classically)
Module linters run on the terminal command, so their messages cannot be captured by `#guard_msgs` in the usual way. The test invokes the linter by hand at the end of the file, on the commands obtained by re-parsing the file out of the elaborator's own `FileMap`, so that the commands carry exactly the positions the declarations were recorded with. Also updates the `privateProof` test for `val4`: now that a deletion is suggested whenever nothing is left unfixable, a command whose only private references sit inside a `Prop`-goal `by` block has its `set_option` reported too, which is correct — it never needed the option. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…f public values Two statements in the module docstring were wrong. The exported view of a declaration is not simply "definitions with a body": bodies are always recorded in the file, but `Environment.setExporting true` replaces the value of an unexposed definition with an axiom, even in the current module, while declarations exported only because of `backward.privateInPublic` are exported as is and so keep their bodies regardless of exposure. Also records why theorem values are not followed: an exported theorem whose proof mentions an unexported private declaration is accepted. The caveat about which commands might still need the option for their own elaboration named private `abbrev`s and `@[expose] private def`s. Neither is exposed: a command all of whose declared names are private is elaborated in a non-exporting environment throughout, per the `withExporting` call in `Lean.Elab.MutualDef.elabMutualDef`, and `@[expose]` warns that it is meaningless there. Only a command declaring a public name alongside a private one is affected. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…hen we want to use it in definitions that are not fully ready to remove `backward.privateInPublic`. Note: we also insist on wrapping the proof anyway now (unless it's a free variable), since the user probably wants to see if that'll work.
This reverts commit bdd3888.
…t is used A single command can declare several candidates — a `mutual` block, or a structure and its projections, constructors and recursors — and the linter suggested deleting its `set_option` as soon as any one of them was unused, which would have unexported the others along with it. The `set_option`s of commands declaring a used candidate are now collected as safeguarded, and deletions of them suppressed. The test case added here is a private structure whose type is used in a public signature: the command declares sixteen candidates, of which only the structure itself is used. Without the safeguard the linter proposes deleting the `set_option` that exports it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.