diff options
Diffstat (limited to 'backends/lean/Base/Progress')
| -rw-r--r-- | backends/lean/Base/Progress/Progress.lean | 77 |
1 files changed, 46 insertions, 31 deletions
diff --git a/backends/lean/Base/Progress/Progress.lean b/backends/lean/Base/Progress/Progress.lean index 1c509775..c8f94e9e 100644 --- a/backends/lean/Base/Progress/Progress.lean +++ b/backends/lean/Base/Progress/Progress.lean @@ -23,7 +23,8 @@ inductive ProgressError | Error (msg : MessageData) deriving Inhabited -def progressWith (fExpr : Expr) (th : TheoremOrLocal) (keep : Option Name) (ids : Array Name) +def progressWith (fExpr : Expr) (th : TheoremOrLocal) + (keep : Option Name) (ids : Array Name) (splitPost : Bool) (asmTac : TacticM Unit) : TacticM ProgressError := do /- Apply the theorem We try to match the theorem with the goal @@ -112,24 +113,31 @@ def progressWith (fExpr : Expr) (th : TheoremOrLocal) (keep : Option Name) (ids mgoal.tryClearMany #[hEq.fvarId!] setGoals (mgoal :: (← getUnsolvedGoals)) trace[Progress] "Goal after splitting eq and post and simplifying the target: {mgoal}" - -- Continue splitting following the ids provided by the user - if ¬ ids.isEmpty then - let hPost ← - match hPost with - | none => do return (.Error m!"Too many ids provided ({ids}): there is no postcondition to split") - | some hPost => pure hPost - let rec splitPost (prevId : Name) (hPost : Expr) (ids : List Name) : TacticM ProgressError := do + -- Continue splitting following the post following the user's instructions + match hPost with + | none => + -- Sanity check + if ¬ ids.isEmpty then + return (.Error m!"Too many ids provided ({ids}): there is no postcondition to split") + else return .Ok + | some hPost => do + let rec splitPostWithIds (prevId : Name) (hPost : Expr) (ids : List Name) : TacticM ProgressError := do match ids with - | [] => pure .Ok -- Stop + | [] => + /- We used all the user provided ids. + Split the remaining conjunctions by using fresh ids if the user + instructed to fully split the post-condition, otherwise stop -/ + if splitPost then + splitFullConjTac hPost (λ _ => pure .Ok) + else pure .Ok | nid :: ids => do trace[Progress] "Splitting post: {hPost}" -- Split if ← isConj (← inferType hPost) then - splitConjTac hPost (some (prevId, nid)) (λ _ nhPost => splitPost nid nhPost ids) + splitConjTac hPost (some (prevId, nid)) (λ _ nhPost => splitPostWithIds nid nhPost ids) else return (.Error m!"Too many ids provided ({nid :: ids}) not enough conjuncts to split in the postcondition") let curPostId := (← hPost.fvarId!.getDecl).userName - splitPost curPostId hPost ids - else return .Ok + splitPostWithIds curPostId hPost ids match res with | .Error _ => return res -- Can we get there? We're using "return" | .Ok => @@ -160,7 +168,8 @@ def getFirstArg (args : Array Expr) : Option Expr := do /- Helper: try to lookup a theorem and apply it, or continue with another tactic if it fails -/ -def tryLookupApply (keep : Option Name) (ids : Array Name) (asmTac : TacticM Unit) (fExpr : Expr) +def tryLookupApply (keep : Option Name) (ids : Array Name) (splitPost : Bool) + (asmTac : TacticM Unit) (fExpr : Expr) (kind : String) (th : Option TheoremOrLocal) (x : TacticM Unit) : TacticM Unit := do let res ← do match th with @@ -172,7 +181,7 @@ def tryLookupApply (keep : Option Name) (ids : Array Name) (asmTac : TacticM Uni -- Apply the theorem let res ← do try - let res ← progressWith fExpr th keep ids asmTac + let res ← progressWith fExpr th keep ids splitPost asmTac pure (some res) catch _ => none match res with @@ -181,7 +190,8 @@ def tryLookupApply (keep : Option Name) (ids : Array Name) (asmTac : TacticM Uni | none => x -- The array of ids are identifiers to use when introducing fresh variables -def progressAsmsOrLookupTheorem (keep : Option Name) (withTh : Option TheoremOrLocal) (ids : Array Name) (asmTac : TacticM Unit) : TacticM Unit := do +def progressAsmsOrLookupTheorem (keep : Option Name) (withTh : Option TheoremOrLocal) + (ids : Array Name) (splitPost : Bool) (asmTac : TacticM Unit) : TacticM Unit := do withMainContext do -- Retrieve the goal let mgoal ← Tactic.getMainGoal @@ -197,7 +207,7 @@ def progressAsmsOrLookupTheorem (keep : Option Name) (withTh : Option TheoremOrL -- Otherwise, lookup one. match withTh with | some th => do - match ← progressWith fExpr th keep ids asmTac with + match ← progressWith fExpr th keep ids splitPost asmTac with | .Ok => return () | .Error msg => throwError msg | none => @@ -206,7 +216,7 @@ def progressAsmsOrLookupTheorem (keep : Option Name) (withTh : Option TheoremOrL let decls ← ctx.getDecls for decl in decls.reverse do trace[Progress] "Trying assumption: {decl.userName} : {decl.type}" - let res ← do try progressWith fExpr (.Local decl) keep ids asmTac catch _ => continue + let res ← do try progressWith fExpr (.Local decl) keep ids splitPost asmTac catch _ => continue match res with | .Ok => return () | .Error msg => throwError msg @@ -216,7 +226,7 @@ def progressAsmsOrLookupTheorem (keep : Option Name) (withTh : Option TheoremOrL let pspec ← do let thName ← pspecAttr.find? fName pure (thName.map fun th => .Theorem th) - tryLookupApply keep ids asmTac fExpr "pspec theorem" pspec do + tryLookupApply keep ids splitPost asmTac fExpr "pspec theorem" pspec do -- It failed: try to lookup a *class* expr spec theorem (those are more -- specific than class spec theorems) let pspecClassExpr ← do @@ -225,7 +235,7 @@ def progressAsmsOrLookupTheorem (keep : Option Name) (withTh : Option TheoremOrL | some arg => do let thName ← pspecClassExprAttr.find? fName arg pure (thName.map fun th => .Theorem th) - tryLookupApply keep ids asmTac fExpr "pspec class expr theorem" pspecClassExpr do + tryLookupApply keep ids splitPost asmTac fExpr "pspec class expr theorem" pspecClassExpr do -- It failed: try to lookup a *class* spec theorem let pspecClass ← do match ← getFirstArgAppName args with @@ -233,7 +243,7 @@ def progressAsmsOrLookupTheorem (keep : Option Name) (withTh : Option TheoremOrL | some argName => do let thName ← pspecClassAttr.find? fName argName pure (thName.map fun th => .Theorem th) - tryLookupApply keep ids asmTac fExpr "pspec class theorem" pspecClass do + tryLookupApply keep ids splitPost asmTac fExpr "pspec class theorem" pspecClass do -- Try a recursive call - we try the assumptions of kind "auxDecl" let ctx ← Lean.MonadLCtx.getLCtx let decls ← ctx.getAllDecls @@ -241,14 +251,14 @@ def progressAsmsOrLookupTheorem (keep : Option Name) (withTh : Option TheoremOrL | .default | .implDetail => false | .auxDecl => true) for decl in decls.reverse do trace[Progress] "Trying recursive assumption: {decl.userName} : {decl.type}" - let res ← do try progressWith fExpr (.Local decl) keep ids asmTac catch _ => continue + let res ← do try progressWith fExpr (.Local decl) keep ids splitPost asmTac catch _ => continue match res with | .Ok => return () | .Error msg => throwError msg -- Nothing worked: failed throwError "Progress failed" -syntax progressArgs := ("keep" ("as" (ident))?)? ("with" ident)? ("as" " ⟨ " (ident)+ " ⟩")? +syntax progressArgs := ("keep" ("as" (ident))?)? ("with" ident)? ("as" " ⟨ " ident,* " .."? " ⟩")? def evalProgress (args : TSyntax `Progress.progressArgs) : TacticM Unit := do let args := args.raw @@ -263,8 +273,8 @@ def evalProgress (args : TSyntax `Progress.progressArgs) : TacticM Unit := do else do pure (some (← mkFreshUserName `h)) else pure none trace[Progress] "Keep: {keep}" - let withArg := (args.get! 1).getArgs let withArg ← do + let withArg := (args.get! 1).getArgs if withArg.size > 0 then let id := withArg.get! 1 trace[Progress] "With arg: {id}" @@ -283,20 +293,25 @@ def evalProgress (args : TSyntax `Progress.progressArgs) : TacticM Unit := do | id :: _ => pure (some (.Theorem id)) else pure none - let args := (args.get! 2).getArgs - let args := (args.get! 2).getArgs - let ids := args.map Syntax.getId + let ids := + let args := (args.get! 2).getArgs + let args := (args.get! 2).getSepArgs + args.map Syntax.getId trace[Progress] "User-provided ids: {ids}" - progressAsmsOrLookupTheorem keep withArg ids (firstTac [assumptionTac, Arith.scalarTac]) + let splitPost : Bool := + let args := (args.get! 2).getArgs + (args.get! 3).getArgs.size > 0 + trace[Progress] "Split post: {splitPost}" + progressAsmsOrLookupTheorem keep withArg ids splitPost (firstTac [assumptionTac, Arith.scalarTac]) elab "progress" args:progressArgs : tactic => evalProgress args -/- -namespace Test +/-namespace Test open Primitives Result set_option trace.Progress true + set_option pp.rawOnError true #eval showStoredPSpec #eval showStoredPSpecClass @@ -306,9 +321,9 @@ namespace Test (hmax : x.val + y.val ≤ Scalar.max ty) : ∃ z, x + y = ret z ∧ z.val = x.val + y.val := by -- progress keep as h with Scalar.add_spec as ⟨ z ⟩ - progress keep as h + progress keep as h as ⟨ z, h1 .. ⟩ simp [*] -end Test -/ +end Test-/ end Progress |