// Copyright 2023 The Go Authors. All rights reserved. // Use of this source code is governed by a BSD-style // license that can be found in the LICENSE file. /* Package rangefunc rewrites range-over-func to code that doesn't use range-over-funcs. Rewriting the construct in the front end, before noder, means the functions generated during the rewrite are available in a noder-generated representation for inlining by the back end. # Theory of Operation The basic idea is to rewrite for x := range f { ... } into f(func(x T) bool { ... }) But it's not usually that easy. # Range variables For a range not using :=, the assigned variables cannot be function parameters in the generated body function. Instead, we allocate fake parameters and start the body with an assignment. For example: for expr1, expr2 = range f { ... } becomes f(func(#p1 T1, #p2 T2) bool { expr1, expr2 = #p1, #p2 ... }) (All the generated variables have a # at the start to signal that they are internal variables when looking at the generated code in a debugger. Because variables have all been resolved to the specific objects they represent, there is no danger of using plain "p1" and colliding with a Go variable named "p1"; the # is just nice to have, not for correctness.) It can also happen that there are fewer range variables than function arguments, in which case we end up with something like f(func(x T1, _ T2) bool { ... }) or f(func(#p1 T1, #p2 T2, _ T3) bool { expr1, expr2 = #p1, #p2 ... }) # Return If the body contains a "break", that break turns into "return false", to tell f to stop. And if the body contains a "continue", that turns into "return true", to tell f to proceed with the next value. Those are the easy cases. If the body contains a return or a break/continue/goto L, then we need to rewrite that into code that breaks out of the loop and then triggers that control flow. In general we rewrite for x := range f { ... } into { var #next int f(func(x T1) bool { ... return true }) ... check #next ... } The variable #next is an integer code that says what to do when f returns. Each difficult statement sets #next and then returns false to stop f. A plain "return" rewrites to {#next = -1; return false}. The return false breaks the loop. Then when f returns, the "check #next" section includes if #next == -1 { return } which causes the return we want. Return with arguments is more involved, and has to deal with corner cases involving panic, defer, and recover. The results of the enclosing function or closure are rewritten to give them names if they don't have them already, and the names are assigned at the return site. func foo() (#rv1 A, #rv2 B) { { var ( #next int ) f(func(x T1) bool { ... { // return a, b #rv1, #rv2 = a, b #next = -1 return false } ... return true }) if #next == -1 { return } } # Checking To permit checking that an iterator is well-behaved -- that is, that it does not call the loop body again after it has returned false or after the entire loop has exited (it might retain a copy of the body function, or pass it to another goroutine) -- each generated loop has its own #stateK variable that is used to check for permitted call patterns to the yield function for a loop body. The state values are: abi.RF_DONE = 0 // body of loop has exited in a non-panic way abi.RF_READY = 1 // body of loop has not exited yet, is not running abi.RF_PANIC = 2 // body of loop is either currently running, or has panicked abi.RF_EXHAUSTED = 3 // iterator function call, e.g. f(func(x t){...}), returned so the sequence is "exhausted". abi.RF_MISSING_PANIC = 4 // used to report errors. The value of #stateK transitions (1) before calling the iterator function, var #stateN = abi.RF_READY (2) after the iterator function call returns, if #stateN == abi.RF_PANIC { panic(runtime.panicrangestate(abi.RF_MISSING_PANIC)) } #stateN = abi.RF_EXHAUSTED (3) at the beginning of the iteration of the loop body, if #stateN != abi.RF_READY { runtime.panicrangestate(#stateN) } #stateN = abi.RF_PANIC (4) when loop iteration continues, #stateN = abi.RF_READY [return true] (5) when control flow exits the loop body. #stateN = abi.RF_DONE [return false] For example: for x := range f { ... if ... { break } ... } becomes { var #state1 = abi.RF_READY f(func(x T1) bool { if #state1 != abi.RF_READY { runtime.panicrangestate(#state1) } #state1 = abi.RF_PANIC ... if ... { #state1 = abi.RF_DONE ; return false } ... #state1 = abi.RF_READY return true }) if #state1 == abi.RF_PANIC { // the code for the loop body did not return normally panic(runtime.panicrangestate(abi.RF_MISSING_PANIC)) } #state1 = abi.RF_EXHAUSTED } # Nested Loops So far we've only considered a single loop. If a function contains a sequence of loops, each can be translated individually. But loops can be nested. It would work to translate the innermost loop and then translate the loop around it, and so on, except that there'd be a lot of rewriting of rewritten code and the overall traversals could end up taking time quadratic in the depth of the nesting. To avoid all that, we use a single rewriting pass that handles a top-most range-over-func loop and all the range-over-func loops it contains at the same time. If we need to return from inside a doubly-nested loop, the rewrites above stay the same, but the check after the inner loop only says if #next < 0 { return false } to stop the outer loop so it can do the actual return. That is, for range f { for range g { ... return a, b ... } } becomes { var ( #next int ) var #state1 = abi.RF_READY f(func() bool { if #state1 != abi.RF_READY { runtime.panicrangestate(#state1) } #state1 = abi.RF_PANIC var #state2 = abi.RF_READY g(func() bool { if #state2 != abi.RF_READY { runtime.panicrangestate(#state2) } ... { // return a, b #rv1, #rv2 = a, b #next = -1 #state2 = abi.RF_DONE return false } ... #state2 = abi.RF_READY return true }) if #state2 == abi.RF_PANIC { panic(runtime.panicrangestate(abi.RF_MISSING_PANIC)) } #state2 = abi.RF_EXHAUSTED if #next < 0 { #state1 = abi.RF_DONE return false } #state1 = abi.RF_READY return true }) if #state1 == abi.RF_PANIC { panic(runtime.panicrangestate(abi.RF_MISSING_PANIC)) } #state1 = abi.RF_EXHAUSTED if #next == -1 { return } } # Labeled break/continue of range-over-func loops For a labeled break or continue of an outer range-over-func, we use positive #next values. Any such labeled break or continue really means "do N breaks" or "do N breaks and 1 continue". The positive #next value tells which level of loop N to target with a break or continue, where perLoopStep*N means break out of level N and perLoopStep*N-1 means continue into level N. The outermost loop has level 1, therefore #next == perLoopStep means to break from the outermost loop, and #next == perLoopStep-1 means to continue the outermost loop. Loops that might need to propagate a labeled break or continue add one or both of these to the #next checks: // N == depth of this loop, one less than the one just exited. if #next != 0 { if #next >= perLoopStep*N-1 { // break or continue this loop if #next >= perLoopStep*N+1 { // error checking // TODO reason about what exactly can appear // here given full or partial checking. runtime.panicrangestate(abi.RF_DONE) } rv := #next & 1 == 1 // code generates into #next&1 #next = 0 return rv } return false // or handle returns and gotos } For example (with perLoopStep == 2) F: for range f { // 1, 2 for range g { // 3, 4 for range h { ... break F ... ... continue F ... } } ... } becomes { var #next int var #state1 = abi.RF_READY f(func() { // 1,2 if #state1 != abi.RF_READY { runtime.panicrangestate(#state1) } #state1 = abi.RF_PANIC var #state2 = abi.RF_READY g(func() { // 3,4 if #state2 != abi.RF_READY { runtime.panicrangestate(#state2) } #state2 = abi.RF_PANIC var #state3 = abi.RF_READY h(func() { // 5,6 if #state3 != abi.RF_READY { runtime.panicrangestate(#state3) } #state3 = abi.RF_PANIC ... { // break F #next = 2 #state3 = abi.RF_DONE return false } ... { // continue F #next = 1 #state3 = abi.RF_DONE return false } ... #state3 = abi.RF_READY return true }) if #state3 == abi.RF_PANIC { panic(runtime.panicrangestate(abi.RF_MISSING_PANIC)) } #state3 = abi.RF_EXHAUSTED if #next != 0 { // no breaks or continues targeting this loop #state2 = abi.RF_DONE return false } return true }) if #state2 == abi.RF_PANIC { panic(runtime.panicrangestate(abi.RF_MISSING_PANIC)) } #state2 = abi.RF_EXHAUSTED if #next != 0 { // just exited g, test for break/continue applied to f/F if #next >= 1 { if #next >= 3 { runtime.panicrangestate(abi.RF_DONE) } // error rv := #next&1 == 1 #next = 0 return rv } #state1 = abi.RF_DONE return false } ... return true }) if #state1 == abi.RF_PANIC { panic(runtime.panicrangestate(abi.RF_MISSING_PANIC)) } #state1 = abi.RF_EXHAUSTED } Note that the post-h checks only consider a break, since no generated code tries to continue g. # Gotos and other labeled break/continue The final control flow translations are goto and break/continue of a non-range-over-func statement. In both cases, we may need to break out of one or more range-over-func loops before we can do the actual control flow statement. Each such break/continue/goto L statement is assigned a unique negative #next value (since -1 is return). Then the post-checks for a given loop test for the specific codes that refer to labels directly targetable from that block. Otherwise, the generic if #next < 0 { return false } check handles stopping the next loop to get one step closer to the label. For example Top: print("start\n") for range f { for range g { ... for range h { ... goto Top ... } } } becomes Top: print("start\n") { var #next int var #state1 = abi.RF_READY f(func() { if #state1 != abi.RF_READY{ runtime.panicrangestate(#state1) } #state1 = abi.RF_PANIC var #state2 = abi.RF_READY g(func() { if #state2 != abi.RF_READY { runtime.panicrangestate(#state2) } #state2 = abi.RF_PANIC ... var #state3 bool = abi.RF_READY h(func() { if #state3 != abi.RF_READY { runtime.panicrangestate(#state3) } #state3 = abi.RF_PANIC ... { // goto Top #next = -3 #state3 = abi.RF_DONE return false } ... #state3 = abi.RF_READY return true }) if #state3 == abi.RF_PANIC {runtime.panicrangestate(abi.RF_MISSING_PANIC)} #state3 = abi.RF_EXHAUSTED if #next < 0 { #state2 = abi.RF_DONE return false } #state2 = abi.RF_READY return true }) if #state2 == abi.RF_PANIC {runtime.panicrangestate(abi.RF_MISSING_PANIC)} #state2 = abi.RF_EXHAUSTED if #next < 0 { #state1 = abi.RF_DONE return false } #state1 = abi.RF_READY return true }) if #state1 == abi.RF_PANIC {runtime.panicrangestate(abi.RF_MISSING_PANIC)} #state1 = abi.RF_EXHAUSTED if #next == -3 { #next = 0 goto Top } } Labeled break/continue to non-range-over-funcs are handled the same way as goto. # Defers The last wrinkle is handling defer statements. If we have for range f { defer print("A") } we cannot rewrite that into f(func() { defer print("A") }) because the deferred code will run at the end of the iteration, not the end of the containing function. To fix that, the runtime provides a special hook that lets us obtain a defer "token" representing the outer function and then use it in a later defer to attach the deferred code to that outer function. Normally, defer print("A") compiles to runtime.deferproc(func() { print("A") }) This changes in a range-over-func. For example: for range f { defer print("A") } compiles to var #defers = runtime.deferrangefunc() f(func() { runtime.deferprocat(func() { print("A") }, #defers) }) For this rewriting phase, we insert the explicit initialization of #defers and then attach the #defers variable to the CallStmt representing the defer. That variable will be propagated to the backend and will cause the backend to compile the defer using deferprocat instead of an ordinary deferproc. TODO: Could call runtime.deferrangefuncend after f. */ package rangefunc import ( "cmd/compile/internal/base" "cmd/compile/internal/syntax" "cmd/compile/internal/types2" "fmt" "go/constant" "internal/abi" "os" ) // nopos is the zero syntax.Pos. var nopos syntax.Pos // A rewriter implements rewriting the range-over-funcs in a given function. type rewriter struct { pkg *types2.Package info *types2.Info sig *types2.Signature outer *syntax.FuncType body *syntax.BlockStmt // References to important types and values. any types2.Object bool types2.Object int types2.Object true types2.Object false types2.Object // Branch numbering, computed as needed. branchNext map[branch]int // branch -> #next value labelLoop map[string]*syntax.ForStmt // label -> innermost rangefunc loop it is declared inside (nil for no loop) // Stack of nodes being visited. stack []syntax.Node // all nodes forStack []*forLoop // range-over-func loops rewritten map[*syntax.ForStmt]syntax.Stmt // Declared variables in generated code for outermost loop. declStmt *syntax.DeclStmt nextVar types2.Object defers types2.Object stateVarCount int // stateVars are referenced from their respective loops bodyClosureCount int // to help the debugger, the closures generated for loop bodies get names rangefuncBodyClosures map[*syntax.FuncLit]bool } // A branch is a single labeled branch. type branch struct { tok syntax.Token label string } // A forLoop describes a single range-over-func loop being processed. type forLoop struct { nfor *syntax.ForStmt // actual syntax stateVar *types2.Var // #state variable for this loop stateVarDecl *syntax.VarDecl depth int // outermost loop has depth 1, otherwise depth = depth(parent)+1 checkRet bool // add check for "return" after loop checkBreak bool // add check for "break" after loop checkContinue bool // add check for "continue" after loop checkBranch []branch // add check for labeled branch after loop } type State int // Rewrite rewrites all the range-over-funcs in the files. // It returns the set of function literals generated from rangefunc loop bodies. // This allows for rangefunc loop bodies to be distingushed by debuggers. func Rewrite(pkg *types2.Package, info *types2.Info, files []*syntax.File) map[*syntax.FuncLit]bool { ri := make(map[*syntax.FuncLit]bool) for _, file := range files { syntax.Inspect(file, func(n syntax.Node) bool { switch n := n.(type) { case *syntax.FuncDecl: sig, _ := info.Defs[n.Name].Type().(*types2.Signature) rewriteFunc(pkg, info, n.Type, n.Body, sig, ri) return false case *syntax.FuncLit: sig, _ := info.Types[n].Type.(*types2.Signature) if sig == nil { tv := n.GetTypeInfo() sig = tv.Type.(*types2.Signature) } rewriteFunc(pkg, info, n.Type, n.Body, sig, ri) return false } return true }) } return ri } // rewriteFunc rewrites all the range-over-funcs in a single function (a top-level func or a func literal). // The typ and body are the function's type and body. func rewriteFunc(pkg *types2.Package, info *types2.Info, typ *syntax.FuncType, body *syntax.BlockStmt, sig *types2.Signature, ri map[*syntax.FuncLit]bool) { if body == nil { return } r := &rewriter{ pkg: pkg, info: info, outer: typ, body: body, sig: sig, rangefuncBodyClosures: ri, } syntax.Inspect(body, r.inspect) if (base.Flag.W != 0) && r.forStack != nil { syntax.Fdump(os.Stderr, body) } } // checkFuncMisuse reports whether to check for misuse of iterator callbacks functions. func (r *rewriter) checkFuncMisuse() bool { return base.Debug.RangeFuncCheck != 0 } // inspect is a callback for syntax.Inspect that drives the actual rewriting. // If it sees a func literal, it kicks off a separate rewrite for that literal. // Otherwise, it maintains a stack of range-over-func loops and // converts each in turn. func (r *rewriter) inspect(n syntax.Node) bool { switch n := n.(type) { case *syntax.FuncLit: sig, _ := r.info.Types[n].Type.(*types2.Signature) if sig == nil { tv := n.GetTypeInfo() sig = tv.Type.(*types2.Signature) } rewriteFunc(r.pkg, r.info, n.Type, n.Body, sig, r.rangefuncBodyClosures) return false default: // Push n onto stack. r.stack = append(r.stack, n) if nfor, ok := forRangeFunc(n); ok { loop := &forLoop{nfor: nfor, depth: 1 + len(r.forStack)} r.forStack = append(r.forStack, loop) r.startLoop(loop) } case nil: // n == nil signals that we are done visiting // the top-of-stack node's children. Find it. n = r.stack[len(r.stack)-1] // If we are inside a range-over-func, // take this moment to replace any break/continue/goto/return // statements directly contained in this node. // Also replace any converted for statements // with the rewritten block. switch n := n.(type) { case *syntax.BlockStmt: for i, s := range n.List { n.List[i] = r.editStmt(s) } case *syntax.CaseClause: for i, s := range n.Body { n.Body[i] = r.editStmt(s) } case *syntax.CommClause: for i, s := range n.Body { n.Body[i] = r.editStmt(s) } case *syntax.LabeledStmt: n.Stmt = r.editStmt(n.Stmt) } // Pop n. if len(r.forStack) > 0 && r.stack[len(r.stack)-1] == r.forStack[len(r.forStack)-1].nfor { r.endLoop(r.forStack[len(r.forStack)-1]) r.forStack = r.forStack[:len(r.forStack)-1] } r.stack = r.stack[:len(r.stack)-1] } return true } // startLoop sets up for converting a range-over-func loop. func (r *rewriter) startLoop(loop *forLoop) { // For first loop in function, allocate syntax for any, bool, int, true, and false. if r.any == nil { r.any = types2.Universe.Lookup("any") r.bool = types2.Universe.Lookup("bool") r.int = types2.Universe.Lookup("int") r.true = types2.Universe.Lookup("true") r.false = types2.Universe.Lookup("false") r.rewritten = make(map[*syntax.ForStmt]syntax.Stmt) } if r.checkFuncMisuse() { // declare the state flag for this loop's body loop.stateVar, loop.stateVarDecl = r.stateVar(loop.nfor.Pos()) } } // editStmt returns the replacement for the statement x, // or x itself if it should be left alone. // This includes the for loops we are converting, // as left in x.rewritten by r.endLoop. func (r *rewriter) editStmt(x syntax.Stmt) syntax.Stmt { if x, ok := x.(*syntax.ForStmt); ok { if s := r.rewritten[x]; s != nil { return s } } if len(r.forStack) > 0 { switch x := x.(type) { case *syntax.BranchStmt: return r.editBranch(x) case *syntax.CallStmt: if x.Tok == syntax.Defer { return r.editDefer(x) } case *syntax.ReturnStmt: return r.editReturn(x) } } return x } // editDefer returns the replacement for the defer statement x. // See the "Defers" section in the package doc comment above for more context. func (r *rewriter) editDefer(x *syntax.CallStmt) syntax.Stmt { if r.defers == nil { // Declare and initialize the #defers token. init := &syntax.CallExpr{ Fun: runtimeSym(r.info, "deferrangefunc"), } tv := syntax.TypeAndValue{Type: r.any.Type()} tv.SetIsValue() init.SetTypeInfo(tv) r.defers = r.declOuterVar("#defers", r.any.Type(), init) } // Attach the token as an "extra" argument to the defer. x.DeferAt = r.useObj(r.defers) setPos(x.DeferAt, x.Pos()) return x } func (r *rewriter) stateVar(pos syntax.Pos) (*types2.Var, *syntax.VarDecl) { r.stateVarCount++ name := fmt.Sprintf("#state%d", r.stateVarCount) typ := r.int.Type() obj := types2.NewVar(pos, r.pkg, name, typ) n := syntax.NewName(pos, name) setValueType(n, typ) r.info.Defs[n] = obj return obj, &syntax.VarDecl{NameList: []*syntax.Name{n}, Values: r.stateConst(abi.RF_READY)} } // editReturn returns the replacement for the return statement x. // See the "Return" section in the package doc comment above for more context. func (r *rewriter) editReturn(x *syntax.ReturnStmt) syntax.Stmt { bl := &syntax.BlockStmt{} if x.Results != nil { // rewrite "return val" into "assign to named result; return" if len(r.outer.ResultList) > 0 { // Make sure that result parameters all have names for i, a := range r.outer.ResultList { if a.Name == nil || a.Name.Value == "_" { r.generateParamName(r.outer.ResultList, i) // updates a.Name } } } // Assign to named results results := []types2.Object{} for _, a := range r.outer.ResultList { results = append(results, r.info.Defs[a.Name]) } bl.List = append(bl.List, &syntax.AssignStmt{Lhs: r.useList(results), Rhs: x.Results}) x.Results = nil } next := -1 // return // Tell the loops along the way to check for a return. for _, loop := range r.forStack { loop.checkRet = true } // Set #next, and return false. bl.List = append(bl.List, &syntax.AssignStmt{Lhs: r.next(), Rhs: r.intConst(next)}) if r.checkFuncMisuse() { // mark this loop as exited, the others (which will be exited if iterators do not interfere) have not, yet. bl.List = append(bl.List, r.setState(abi.RF_DONE, x.Pos())) } bl.List = append(bl.List, &syntax.ReturnStmt{Results: r.useObj(r.false)}) setPos(bl, x.Pos()) return bl } // perLoopStep is part of the encoding of loop-spanning control flow // for function range iterators. Each multiple of two encodes a "return false" // passing control to an enclosing iterator; a terminal value of 1 encodes // "return true" (i.e., local continue) from the body function, and a terminal // value of 0 encodes executing the remainder of the body function. const perLoopStep = 2 // editBranch returns the replacement for the branch statement x, // or x itself if it should be left alone. // See the package doc comment above for more context. func (r *rewriter) editBranch(x *syntax.BranchStmt) syntax.Stmt { if x.Tok == syntax.Fallthrough { // Fallthrough is unaffected by the rewrite. return x } // Find target of break/continue/goto in r.forStack. // (The target may not be in r.forStack at all.) targ := x.Target i := len(r.forStack) - 1 if x.Label == nil && r.forStack[i].nfor != targ { // Unlabeled break or continue that's not nfor must be inside nfor. Leave alone. return x } for i >= 0 && r.forStack[i].nfor != targ { i-- } // exitFrom is the index of the loop interior to the target of the control flow, // if such a loop exists (it does not if i == len(r.forStack) - 1) exitFrom := i + 1 // Compute the value to assign to #next and the specific return to use. var next int var ret *syntax.ReturnStmt if x.Tok == syntax.Goto || i < 0 { // goto Label // or break/continue of labeled non-range-over-func loop (x.Label != nil). // We may be able to leave it alone, or we may have to break // out of one or more nested loops and then use #next to signal // to complete the break/continue/goto. // Figure out which range-over-func loop contains the label. r.computeBranchNext() nfor := r.forStack[len(r.forStack)-1].nfor label := x.Label.Value targ := r.labelLoop[label] if nfor == targ { // Label is in the innermost range-over-func loop; use it directly. return x } // Set #next to the code meaning break/continue/goto label. next = r.branchNext[branch{x.Tok, label}] // Break out of nested loops up to targ. i := len(r.forStack) - 1 for i >= 0 && r.forStack[i].nfor != targ { i-- } exitFrom = i + 1 // Mark loop we exit to get to targ to check for that branch. // When i==-1 / exitFrom == 0 that's the outermost func body. top := r.forStack[exitFrom] top.checkBranch = append(top.checkBranch, branch{x.Tok, label}) // Mark loops along the way to check for a plain return, so they break. for j := exitFrom + 1; j < len(r.forStack); j++ { r.forStack[j].checkRet = true } // In the innermost loop, use a plain "return false". ret = &syntax.ReturnStmt{Results: r.useObj(r.false)} } else { // break/continue of labeled range-over-func loop. if exitFrom == len(r.forStack) { // Simple break or continue. // Continue returns true, break returns false, optionally both adjust state, // neither modifies #next. var state abi.RF_State if x.Tok == syntax.Continue { ret = &syntax.ReturnStmt{Results: r.useObj(r.true)} state = abi.RF_READY } else { ret = &syntax.ReturnStmt{Results: r.useObj(r.false)} state = abi.RF_DONE } var stmts []syntax.Stmt if r.checkFuncMisuse() { stmts = []syntax.Stmt{r.setState(state, x.Pos()), ret} } else { stmts = []syntax.Stmt{ret} } bl := &syntax.BlockStmt{ List: stmts, } setPos(bl, x.Pos()) return bl } ret = &syntax.ReturnStmt{Results: r.useObj(r.false)} // The loop inside the one we are break/continue-ing // needs to make that happen when we break out of it. if x.Tok == syntax.Continue { r.forStack[exitFrom].checkContinue = true } else { exitFrom = i // exitFrom-- r.forStack[exitFrom].checkBreak = true } // The loops along the way just need to break. for j := exitFrom + 1; j < len(r.forStack); j++ { r.forStack[j].checkBreak = true } // Set next to break the appropriate number of times; // the final time may be a continue, not a break. next = perLoopStep * (i + 1) if x.Tok == syntax.Continue { next-- } } // Assign #next = next and do the return. as := &syntax.AssignStmt{Lhs: r.next(), Rhs: r.intConst(next)} bl := &syntax.BlockStmt{ List: []syntax.Stmt{as}, } if r.checkFuncMisuse() { // Set #stateK for this loop. // The exterior loops have not exited yet, and the iterator might interfere. bl.List = append(bl.List, r.setState(abi.RF_DONE, x.Pos())) } bl.List = append(bl.List, ret) setPos(bl, x.Pos()) return bl } // computeBranchNext computes the branchNext numbering // and determines which labels end up inside which range-over-func loop bodies. func (r *rewriter) computeBranchNext() { if r.labelLoop != nil { return } r.labelLoop = make(map[string]*syntax.ForStmt) r.branchNext = make(map[branch]int) var labels []string var stack []syntax.Node var forStack []*syntax.ForStmt forStack = append(forStack, nil) syntax.Inspect(r.body, func(n syntax.Node) bool { if n != nil { stack = append(stack, n) if nfor, ok := forRangeFunc(n); ok { forStack = append(forStack, nfor) } if n, ok := n.(*syntax.LabeledStmt); ok { l := n.Label.Value labels = append(labels, l) f := forStack[len(forStack)-1] r.labelLoop[l] = f } } else { n := stack[len(stack)-1] stack = stack[:len(stack)-1] if n == forStack[len(forStack)-1] { forStack = forStack[:len(forStack)-1] } } return true }) // Assign numbers to all the labels we observed. used := -1 // returns use -1 for _, l := range labels { used -= 3 r.branchNext[branch{syntax.Break, l}] = used r.branchNext[branch{syntax.Continue, l}] = used + 1 r.branchNext[branch{syntax.Goto, l}] = used + 2 } } // endLoop finishes the conversion of a range-over-func loop. // We have inspected and rewritten the body of the loop and can now // construct the body function and rewrite the for loop into a call // bracketed by any declarations and checks it requires. func (r *rewriter) endLoop(loop *forLoop) { // Pick apart for range X { ... } nfor := loop.nfor start, end := nfor.Pos(), nfor.Body.Rbrace // start, end position of for loop rclause := nfor.Init.(*syntax.RangeClause) rfunc := types2.CoreType(rclause.X.GetTypeInfo().Type).(*types2.Signature) // type of X - func(func(...)bool) if rfunc.Params().Len() != 1 { base.Fatalf("invalid typecheck of range func") } ftyp := types2.CoreType(rfunc.Params().At(0).Type()).(*types2.Signature) // func(...) bool if ftyp.Results().Len() != 1 { base.Fatalf("invalid typecheck of range func") } // Give the closure generated for the body a name, to help the debugger connect it to its frame, if active. r.bodyClosureCount++ clo := r.bodyFunc(nfor.Body.List, syntax.UnpackListExpr(rclause.Lhs), rclause.Def, ftyp, start, end) cloDecl, cloVar := r.declSingleVar(fmt.Sprintf("#yield%d", r.bodyClosureCount), clo.GetTypeInfo().Type, clo) setPos(cloDecl, start) // Build X(bodyFunc) call := &syntax.ExprStmt{ X: &syntax.CallExpr{ Fun: rclause.X, ArgList: []syntax.Expr{ r.useObj(cloVar), }, }, } setPos(call, start) // Build checks based on #next after X(bodyFunc) checks := r.checks(loop, end) // Rewrite for vars := range X { ... } to // // { // r.declStmt // call // checks // } // // The r.declStmt can be added to by this loop or any inner loop // during the creation of r.bodyFunc; it is only emitted in the outermost // converted range loop. block := &syntax.BlockStmt{Rbrace: end} setPos(block, start) if len(r.forStack) == 1 && r.declStmt != nil { setPos(r.declStmt, start) block.List = append(block.List, r.declStmt) } // declare the state variable here so it has proper scope and initialization if r.checkFuncMisuse() { stateVarDecl := &syntax.DeclStmt{DeclList: []syntax.Decl{loop.stateVarDecl}} setPos(stateVarDecl, start) block.List = append(block.List, stateVarDecl) } // iteratorFunc(bodyFunc) block.List = append(block.List, cloDecl, call) if r.checkFuncMisuse() { // iteratorFunc has exited, check for swallowed panic, and set body state to abi.RF_EXHAUSTED nif := &syntax.IfStmt{ Cond: r.cond(syntax.Eql, r.useObj(loop.stateVar), r.stateConst(abi.RF_PANIC)), Then: &syntax.BlockStmt{ List: []syntax.Stmt{r.callPanic(start, r.stateConst(abi.RF_MISSING_PANIC))}, }, } setPos(nif, end) block.List = append(block.List, nif) block.List = append(block.List, r.setState(abi.RF_EXHAUSTED, end)) } block.List = append(block.List, checks...) if len(r.forStack) == 1 { // ending an outermost loop r.declStmt = nil r.nextVar = nil r.defers = nil } r.rewritten[nfor] = block } func (r *rewriter) cond(op syntax.Operator, x, y syntax.Expr) *syntax.Operation { cond := &syntax.Operation{Op: op, X: x, Y: y} tv := syntax.TypeAndValue{Type: r.bool.Type()} tv.SetIsValue() cond.SetTypeInfo(tv) return cond } func (r *rewriter) setState(val abi.RF_State, pos syntax.Pos) *syntax.AssignStmt { ss := r.setStateAt(len(r.forStack)-1, val) setPos(ss, pos) return ss } func (r *rewriter) setStateAt(index int, stateVal abi.RF_State) *syntax.AssignStmt { loop := r.forStack[index] return &syntax.AssignStmt{ Lhs: r.useObj(loop.stateVar), Rhs: r.stateConst(stateVal), } } // bodyFunc converts the loop body (control flow has already been updated) // to a func literal that can be passed to the range function. // // vars is the range variables from the range statement. // def indicates whether this is a := range statement. // ftyp is the type of the function we are creating // start and end are the syntax positions to use for new nodes // that should be at the start or end of the loop. func (r *rewriter) bodyFunc(body []syntax.Stmt, lhs []syntax.Expr, def bool, ftyp *types2.Signature, start, end syntax.Pos) *syntax.FuncLit { // Starting X(bodyFunc); build up bodyFunc first. var params, results []*types2.Var results = append(results, types2.NewVar(start, nil, "#r", r.bool.Type())) bodyFunc := &syntax.FuncLit{ // Note: Type is ignored but needs to be non-nil to avoid panic in syntax.Inspect. Type: &syntax.FuncType{}, Body: &syntax.BlockStmt{ List: []syntax.Stmt{}, Rbrace: end, }, } r.rangefuncBodyClosures[bodyFunc] = true setPos(bodyFunc, start) for i := 0; i < ftyp.Params().Len(); i++ { typ := ftyp.Params().At(i).Type() var paramVar *types2.Var if i < len(lhs) && def { // Reuse range variable as parameter. x := lhs[i] paramVar = r.info.Defs[x.(*syntax.Name)].(*types2.Var) } else { // Declare new parameter and assign it to range expression. paramVar = types2.NewVar(start, r.pkg, fmt.Sprintf("#p%d", 1+i), typ) if i < len(lhs) { x := lhs[i] as := &syntax.AssignStmt{Lhs: x, Rhs: r.useObj(paramVar)} as.SetPos(x.Pos()) setPos(as.Rhs, x.Pos()) bodyFunc.Body.List = append(bodyFunc.Body.List, as) } } params = append(params, paramVar) } tv := syntax.TypeAndValue{ Type: types2.NewSignatureType(nil, nil, nil, types2.NewTuple(params...), types2.NewTuple(results...), false), } tv.SetIsValue() bodyFunc.SetTypeInfo(tv) loop := r.forStack[len(r.forStack)-1] if r.checkFuncMisuse() { bodyFunc.Body.List = append(bodyFunc.Body.List, r.assertReady(start, loop)) bodyFunc.Body.List = append(bodyFunc.Body.List, r.setState(abi.RF_PANIC, start)) } // Original loop body (already rewritten by editStmt during inspect). bodyFunc.Body.List = append(bodyFunc.Body.List, body...) // end of loop body, set state to abi.RF_READY and return true to continue iteration if r.checkFuncMisuse() { bodyFunc.Body.List = append(bodyFunc.Body.List, r.setState(abi.RF_READY, end)) } ret := &syntax.ReturnStmt{Results: r.useObj(r.true)} ret.SetPos(end) bodyFunc.Body.List = append(bodyFunc.Body.List, ret) return bodyFunc } // checks returns the post-call checks that need to be done for the given loop. func (r *rewriter) checks(loop *forLoop, pos syntax.Pos) []syntax.Stmt { var list []syntax.Stmt if len(loop.checkBranch) > 0 { did := make(map[branch]bool) for _, br := range loop.checkBranch { if did[br] { continue } did[br] = true doBranch := &syntax.BranchStmt{Tok: br.tok, Label: &syntax.Name{Value: br.label}} list = append(list, r.ifNext(syntax.Eql, r.branchNext[br], true, doBranch)) } } curLoop := loop.depth - 1 curLoopIndex := curLoop - 1 if len(r.forStack) == 1 { if loop.checkRet { list = append(list, r.ifNext(syntax.Eql, -1, false, retStmt(nil))) } } else { // Idealized check, implemented more simply for now. // // N == depth of this loop, one less than the one just exited. // if #next != 0 { // if #next >= perLoopStep*N-1 { // this loop // if #next >= perLoopStep*N+1 { // error checking // runtime.panicrangestate(abi.RF_DONE) // } // rv := #next & 1 == 1 // code generates into #next&1 // #next = 0 // return rv // } // return false // or handle returns and gotos // } if loop.checkRet { // Note: next < 0 also handles gotos handled by outer loops. // We set checkRet in that case to trigger this check. if r.checkFuncMisuse() { list = append(list, r.ifNext(syntax.Lss, 0, false, r.setStateAt(curLoopIndex, abi.RF_DONE), retStmt(r.useObj(r.false)))) } else { list = append(list, r.ifNext(syntax.Lss, 0, false, retStmt(r.useObj(r.false)))) } } depthStep := perLoopStep * (curLoop) if r.checkFuncMisuse() { list = append(list, r.ifNext(syntax.Gtr, depthStep, false, r.callPanic(pos, r.stateConst(abi.RF_DONE)))) } else { list = append(list, r.ifNext(syntax.Gtr, depthStep, true)) } if r.checkFuncMisuse() { if loop.checkContinue { list = append(list, r.ifNext(syntax.Eql, depthStep-1, true, r.setStateAt(curLoopIndex, abi.RF_READY), retStmt(r.useObj(r.true)))) } if loop.checkBreak { list = append(list, r.ifNext(syntax.Eql, depthStep, true, r.setStateAt(curLoopIndex, abi.RF_DONE), retStmt(r.useObj(r.false)))) } if loop.checkContinue || loop.checkBreak { list = append(list, r.ifNext(syntax.Gtr, 0, false, r.setStateAt(curLoopIndex, abi.RF_DONE), retStmt(r.useObj(r.false)))) } } else { if loop.checkContinue { list = append(list, r.ifNext(syntax.Eql, depthStep-1, true, retStmt(r.useObj(r.true)))) } if loop.checkBreak { list = append(list, r.ifNext(syntax.Eql, depthStep, true, retStmt(r.useObj(r.false)))) } if loop.checkContinue || loop.checkBreak { list = append(list, r.ifNext(syntax.Gtr, 0, false, retStmt(r.useObj(r.false)))) } } } for _, j := range list { setPos(j, pos) } return list } // retStmt returns a return statement returning the given return values. func retStmt(results syntax.Expr) *syntax.ReturnStmt { return &syntax.ReturnStmt{Results: results} } // ifNext returns the statement: // // if #next op c { [#next = 0;] thens... } func (r *rewriter) ifNext(op syntax.Operator, c int, zeroNext bool, thens ...syntax.Stmt) syntax.Stmt { var thenList []syntax.Stmt if zeroNext { clr := &syntax.AssignStmt{ Lhs: r.next(), Rhs: r.intConst(0), } thenList = append(thenList, clr) } for _, then := range thens { thenList = append(thenList, then) } nif := &syntax.IfStmt{ Cond: r.cond(op, r.next(), r.intConst(c)), Then: &syntax.BlockStmt{ List: thenList, }, } return nif } // setValueType marks x as a value with type typ. func setValueType(x syntax.Expr, typ syntax.Type) { tv := syntax.TypeAndValue{Type: typ} tv.SetIsValue() x.SetTypeInfo(tv) } // assertReady returns the statement: // // if #stateK != abi.RF_READY { runtime.panicrangestate(#stateK) } // // where #stateK is the state variable for loop. func (r *rewriter) assertReady(start syntax.Pos, loop *forLoop) syntax.Stmt { nif := &syntax.IfStmt{ Cond: r.cond(syntax.Neq, r.useObj(loop.stateVar), r.stateConst(abi.RF_READY)), Then: &syntax.BlockStmt{ List: []syntax.Stmt{r.callPanic(start, r.useObj(loop.stateVar))}, }, } setPos(nif, start) return nif } func (r *rewriter) callPanic(start syntax.Pos, arg syntax.Expr) syntax.Stmt { callPanicExpr := &syntax.CallExpr{ Fun: runtimeSym(r.info, "panicrangestate"), ArgList: []syntax.Expr{arg}, } setValueType(callPanicExpr, nil) // no result type return &syntax.ExprStmt{X: callPanicExpr} } // next returns a reference to the #next variable. func (r *rewriter) next() *syntax.Name { if r.nextVar == nil { r.nextVar = r.declOuterVar("#next", r.int.Type(), nil) } return r.useObj(r.nextVar) } // forRangeFunc checks whether n is a range-over-func. // If so, it returns n.(*syntax.ForStmt), true. // Otherwise it returns nil, false. func forRangeFunc(n syntax.Node) (*syntax.ForStmt, bool) { nfor, ok := n.(*syntax.ForStmt) if !ok { return nil, false } nrange, ok := nfor.Init.(*syntax.RangeClause) if !ok { return nil, false } _, ok = types2.CoreType(nrange.X.GetTypeInfo().Type).(*types2.Signature) if !ok { return nil, false } return nfor, true } // intConst returns syntax for an integer literal with the given value. func (r *rewriter) intConst(c int) *syntax.BasicLit { lit := &syntax.BasicLit{ Value: fmt.Sprint(c), Kind: syntax.IntLit, } tv := syntax.TypeAndValue{Type: r.int.Type(), Value: constant.MakeInt64(int64(c))} tv.SetIsValue() lit.SetTypeInfo(tv) return lit } func (r *rewriter) stateConst(s abi.RF_State) *syntax.BasicLit { return r.intConst(int(s)) } // useObj returns syntax for a reference to decl, which should be its declaration. func (r *rewriter) useObj(obj types2.Object) *syntax.Name { n := syntax.NewName(nopos, obj.Name()) tv := syntax.TypeAndValue{Type: obj.Type()} tv.SetIsValue() n.SetTypeInfo(tv) r.info.Uses[n] = obj return n } // useList is useVar for a list of decls. func (r *rewriter) useList(vars []types2.Object) syntax.Expr { var new []syntax.Expr for _, obj := range vars { new = append(new, r.useObj(obj)) } if len(new) == 1 { return new[0] } return &syntax.ListExpr{ElemList: new} } func (r *rewriter) makeVarName(pos syntax.Pos, name string, typ types2.Type) (*types2.Var, *syntax.Name) { obj := types2.NewVar(pos, r.pkg, name, typ) n := syntax.NewName(pos, name) tv := syntax.TypeAndValue{Type: typ} tv.SetIsValue() n.SetTypeInfo(tv) r.info.Defs[n] = obj return obj, n } func (r *rewriter) generateParamName(results []*syntax.Field, i int) { obj, n := r.sig.RenameResult(results, i) r.info.Defs[n] = obj } // declOuterVar declares a variable with a given name, type, and initializer value, // in the same scope as the outermost loop in a loop nest. func (r *rewriter) declOuterVar(name string, typ types2.Type, init syntax.Expr) *types2.Var { if r.declStmt == nil { r.declStmt = &syntax.DeclStmt{} } stmt := r.declStmt obj, n := r.makeVarName(stmt.Pos(), name, typ) stmt.DeclList = append(stmt.DeclList, &syntax.VarDecl{ NameList: []*syntax.Name{n}, // Note: Type is ignored Values: init, }) return obj } // declSingleVar declares a variable with a given name, type, and initializer value, // and returns both the declaration and variable, so that the declaration can be placed // in a specific scope. func (r *rewriter) declSingleVar(name string, typ types2.Type, init syntax.Expr) (*syntax.DeclStmt, *types2.Var) { stmt := &syntax.DeclStmt{} obj, n := r.makeVarName(stmt.Pos(), name, typ) stmt.DeclList = append(stmt.DeclList, &syntax.VarDecl{ NameList: []*syntax.Name{n}, // Note: Type is ignored Values: init, }) return stmt, obj } // runtimePkg is a fake runtime package that contains what we need to refer to in package runtime. var runtimePkg = func() *types2.Package { var nopos syntax.Pos pkg := types2.NewPackage("runtime", "runtime") anyType := types2.Universe.Lookup("any").Type() intType := types2.Universe.Lookup("int").Type() // func deferrangefunc() unsafe.Pointer obj := types2.NewFunc(nopos, pkg, "deferrangefunc", types2.NewSignatureType(nil, nil, nil, nil, types2.NewTuple(types2.NewParam(nopos, pkg, "extra", anyType)), false)) pkg.Scope().Insert(obj) // func panicrangestate() obj = types2.NewFunc(nopos, pkg, "panicrangestate", types2.NewSignatureType(nil, nil, nil, types2.NewTuple(types2.NewParam(nopos, pkg, "state", intType)), nil, false)) pkg.Scope().Insert(obj) return pkg }() // runtimeSym returns a reference to a symbol in the fake runtime package. func runtimeSym(info *types2.Info, name string) *syntax.Name { obj := runtimePkg.Scope().Lookup(name) n := syntax.NewName(nopos, "runtime."+name) tv := syntax.TypeAndValue{Type: obj.Type()} tv.SetIsValue() tv.SetIsRuntimeHelper() n.SetTypeInfo(tv) info.Uses[n] = obj return n } // setPos walks the top structure of x that has no position assigned // and assigns it all to have position pos. // When setPos encounters a syntax node with a position assigned, // setPos does not look inside that node. // setPos only needs to handle syntax we create in this package; // all other syntax should have positions assigned already. func setPos(x syntax.Node, pos syntax.Pos) { if x == nil { return } syntax.Inspect(x, func(n syntax.Node) bool { if n == nil || n.Pos() != nopos { return false } n.SetPos(pos) switch n := n.(type) { case *syntax.BlockStmt: if n.Rbrace == nopos { n.Rbrace = pos } } return true }) }