Matthew Dempsky ca2f85fd3f cmd/compile: add indexed export format
This CL introduces a new indexed data format for package export
data. This improves on the previous (sequential) binary format by
allowing the compiler to selectively (and lazily) load only the data
that's actually needed for compilation.

In large Go projects, the package export data can become very large
due to transitive type declaration dependencies and inline
function/method bodies. By lazily loading these declarations and
bodies as needed, we avoid wasting time and memory processing
unnecessary and/or redundant data.

In the benchmarks below, "old" is -iexport=false and "new" is
-iexport=true. The suffixes indicate the compiler concurrency (-c) and
inlining (-l) settings used for the build (using -gcflags=all=-foo).
Benchmarks were run on an HP Z620.

Juju is "go build -a github.com/juju/juju/cmd/...":

name          old real-time/op  new real-time/op  delta
Juju/c=1/l=0        44.0s ± 1%        38.7s ± 9%  -11.97%  (p=0.001 n=7+7)
Juju/c=1/l=4        53.7s ± 3%        45.3s ± 4%  -15.53%  (p=0.001 n=7+7)
Juju/c=4/l=0        39.7s ± 8%        32.0s ± 4%  -19.38%  (p=0.001 n=7+7)
Juju/c=4/l=4        46.3s ± 4%        38.0s ± 4%  -18.06%  (p=0.001 n=7+7)

name          old user-time/op  new user-time/op  delta
Juju/c=1/l=0         371s ± 1%         300s ± 0%  -19.07%  (p=0.001 n=7+6)
Juju/c=1/l=4         482s ± 0%         374s ± 1%  -22.37%  (p=0.001 n=7+7)
Juju/c=4/l=0         410s ± 1%         340s ± 1%  -17.19%  (p=0.001 n=7+7)
Juju/c=4/l=4         532s ± 1%         424s ± 1%  -20.26%  (p=0.001 n=7+7)

name          old sys-time/op   new sys-time/op   delta
Juju/c=1/l=0        33.4s ± 1%        28.4s ± 2%  -15.02%  (p=0.001 n=7+7)
Juju/c=1/l=4        40.7s ± 2%        32.8s ± 3%  -19.51%  (p=0.001 n=7+7)
Juju/c=4/l=0        39.8s ± 2%        34.4s ± 2%  -13.74%  (p=0.001 n=7+7)
Juju/c=4/l=4        48.4s ± 2%        40.4s ± 2%  -16.50%  (p=0.001 n=7+7)

Kubelet is "go build -a k8s.io/kubernetes/cmd/kubelet":

name             old real-time/op  new real-time/op  delta
Kubelet/c=1/l=0        42.0s ± 1%        34.8s ± 1%  -17.27%  (p=0.008 n=5+5)
Kubelet/c=1/l=4        55.4s ± 3%        45.4s ± 3%  -18.06%  (p=0.002 n=6+6)
Kubelet/c=4/l=0        37.4s ± 3%        29.9s ± 1%  -20.25%  (p=0.004 n=6+5)
Kubelet/c=4/l=4        48.1s ± 2%        39.0s ± 5%  -18.93%  (p=0.002 n=6+6)

name             old user-time/op  new user-time/op  delta
Kubelet/c=1/l=0         291s ± 1%         233s ± 1%  -19.96%  (p=0.002 n=6+6)
Kubelet/c=1/l=4         385s ± 1%         298s ± 1%  -22.51%  (p=0.002 n=6+6)
Kubelet/c=4/l=0         325s ± 0%         268s ± 1%  -17.48%  (p=0.004 n=5+6)
Kubelet/c=4/l=4         429s ± 1%         343s ± 1%  -20.08%  (p=0.002 n=6+6)

name             old sys-time/op   new sys-time/op   delta
Kubelet/c=1/l=0        25.1s ± 2%        20.9s ± 4%  -16.69%  (p=0.002 n=6+6)
Kubelet/c=1/l=4        31.2s ± 3%        24.4s ± 0%  -21.67%  (p=0.010 n=6+4)
Kubelet/c=4/l=0        30.2s ± 2%        25.6s ± 1%  -15.34%  (p=0.002 n=6+6)
Kubelet/c=4/l=4        37.3s ± 1%        30.9s ± 2%  -17.11%  (p=0.002 n=6+6)

Change-Id: Ie43eb3bbe1392cbb61c86792a17a57b33b9561f0
Reviewed-on: https://go-review.googlesource.com/106796
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
2018-04-24 01:05:27 +00:00

1503 lines
36 KiB
Go

// Copyright 2016 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 gc
import (
"fmt"
"os"
"path/filepath"
"runtime"
"strconv"
"strings"
"unicode/utf8"
"cmd/compile/internal/syntax"
"cmd/compile/internal/types"
"cmd/internal/objabi"
"cmd/internal/src"
)
func parseFiles(filenames []string) uint {
var noders []*noder
// Limit the number of simultaneously open files.
sem := make(chan struct{}, runtime.GOMAXPROCS(0)+10)
for _, filename := range filenames {
p := &noder{
basemap: make(map[*syntax.PosBase]*src.PosBase),
err: make(chan syntax.Error),
}
noders = append(noders, p)
go func(filename string) {
sem <- struct{}{}
defer func() { <-sem }()
defer close(p.err)
base := syntax.NewFileBase(filename)
f, err := os.Open(filename)
if err != nil {
p.error(syntax.Error{Pos: syntax.MakePos(base, 0, 0), Msg: err.Error()})
return
}
defer f.Close()
p.file, _ = syntax.Parse(base, f, p.error, p.pragma, syntax.CheckBranches) // errors are tracked via p.error
}(filename)
}
var lines uint
for _, p := range noders {
for e := range p.err {
p.yyerrorpos(e.Pos, "%s", e.Msg)
}
p.node()
lines += p.file.Lines
p.file = nil // release memory
if nsyntaxerrors != 0 {
errorexit()
}
// Always run testdclstack here, even when debug_dclstack is not set, as a sanity measure.
testdclstack()
}
localpkg.Height = myheight
if flagiexport {
// init.go requires all imported init functions to be
// fully resolved.
// TODO(mdempsky): Can this be done elsewhere more cleanly?
for _, s := range types.InitSyms {
if n := asNode(s.Def); n != nil && s.Pkg != localpkg {
resolve(n)
}
}
}
return lines
}
// makeSrcPosBase translates from a *syntax.PosBase to a *src.PosBase.
func (p *noder) makeSrcPosBase(b0 *syntax.PosBase) *src.PosBase {
// fast path: most likely PosBase hasn't changed
if p.basecache.last == b0 {
return p.basecache.base
}
b1, ok := p.basemap[b0]
if !ok {
fn := b0.Filename()
if b0.IsFileBase() {
b1 = src.NewFileBase(fn, absFilename(fn))
} else {
// line directive base
p0 := b0.Pos()
p1 := src.MakePos(p.makeSrcPosBase(p0.Base()), p0.Line(), p0.Col())
b1 = src.NewLinePragmaBase(p1, fn, fileh(fn), b0.Line(), b0.Col())
}
p.basemap[b0] = b1
}
// update cache
p.basecache.last = b0
p.basecache.base = b1
return b1
}
func (p *noder) makeXPos(pos syntax.Pos) (_ src.XPos) {
return Ctxt.PosTable.XPos(src.MakePos(p.makeSrcPosBase(pos.Base()), pos.Line(), pos.Col()))
}
func (p *noder) yyerrorpos(pos syntax.Pos, format string, args ...interface{}) {
yyerrorl(p.makeXPos(pos), format, args...)
}
var pathPrefix string
// TODO(gri) Can we eliminate fileh in favor of absFilename?
func fileh(name string) string {
return objabi.AbsFile("", name, pathPrefix)
}
func absFilename(name string) string {
return objabi.AbsFile(Ctxt.Pathname, name, pathPrefix)
}
// noder transforms package syntax's AST into a Node tree.
type noder struct {
basemap map[*syntax.PosBase]*src.PosBase
basecache struct {
last *syntax.PosBase
base *src.PosBase
}
file *syntax.File
linknames []linkname
pragcgobuf [][]string
err chan syntax.Error
scope ScopeID
// scopeVars is a stack tracking the number of variables declared in the
// current function at the moment each open scope was opened.
scopeVars []int
lastCloseScopePos syntax.Pos
}
func (p *noder) funcBody(fn *Node, block *syntax.BlockStmt) {
oldScope := p.scope
p.scope = 0
funchdr(fn)
if block != nil {
body := p.stmts(block.List)
if body == nil {
body = []*Node{nod(OEMPTY, nil, nil)}
}
fn.Nbody.Set(body)
lineno = p.makeXPos(block.Rbrace)
fn.Func.Endlineno = lineno
}
funcbody()
p.scope = oldScope
}
func (p *noder) openScope(pos syntax.Pos) {
types.Markdcl()
if trackScopes {
Curfn.Func.Parents = append(Curfn.Func.Parents, p.scope)
p.scopeVars = append(p.scopeVars, len(Curfn.Func.Dcl))
p.scope = ScopeID(len(Curfn.Func.Parents))
p.markScope(pos)
}
}
func (p *noder) closeScope(pos syntax.Pos) {
p.lastCloseScopePos = pos
types.Popdcl()
if trackScopes {
scopeVars := p.scopeVars[len(p.scopeVars)-1]
p.scopeVars = p.scopeVars[:len(p.scopeVars)-1]
if scopeVars == len(Curfn.Func.Dcl) {
// no variables were declared in this scope, so we can retract it.
if int(p.scope) != len(Curfn.Func.Parents) {
Fatalf("scope tracking inconsistency, no variables declared but scopes were not retracted")
}
p.scope = Curfn.Func.Parents[p.scope-1]
Curfn.Func.Parents = Curfn.Func.Parents[:len(Curfn.Func.Parents)-1]
nmarks := len(Curfn.Func.Marks)
Curfn.Func.Marks[nmarks-1].Scope = p.scope
prevScope := ScopeID(0)
if nmarks >= 2 {
prevScope = Curfn.Func.Marks[nmarks-2].Scope
}
if Curfn.Func.Marks[nmarks-1].Scope == prevScope {
Curfn.Func.Marks = Curfn.Func.Marks[:nmarks-1]
}
return
}
p.scope = Curfn.Func.Parents[p.scope-1]
p.markScope(pos)
}
}
func (p *noder) markScope(pos syntax.Pos) {
xpos := p.makeXPos(pos)
if i := len(Curfn.Func.Marks); i > 0 && Curfn.Func.Marks[i-1].Pos == xpos {
Curfn.Func.Marks[i-1].Scope = p.scope
} else {
Curfn.Func.Marks = append(Curfn.Func.Marks, Mark{xpos, p.scope})
}
}
// closeAnotherScope is like closeScope, but it reuses the same mark
// position as the last closeScope call. This is useful for "for" and
// "if" statements, as their implicit blocks always end at the same
// position as an explicit block.
func (p *noder) closeAnotherScope() {
p.closeScope(p.lastCloseScopePos)
}
// linkname records a //go:linkname directive.
type linkname struct {
pos syntax.Pos
local string
remote string
}
func (p *noder) node() {
types.Block = 1
imported_unsafe = false
p.lineno(p.file.PkgName)
mkpackage(p.file.PkgName.Value)
xtop = append(xtop, p.decls(p.file.DeclList)...)
for _, n := range p.linknames {
if imported_unsafe {
lookup(n.local).Linkname = n.remote
} else {
p.yyerrorpos(n.pos, "//go:linkname only allowed in Go files that import \"unsafe\"")
}
}
pragcgobuf = append(pragcgobuf, p.pragcgobuf...)
lineno = src.NoXPos
clearImports()
}
func (p *noder) decls(decls []syntax.Decl) (l []*Node) {
var cs constState
for _, decl := range decls {
p.lineno(decl)
switch decl := decl.(type) {
case *syntax.ImportDecl:
p.importDecl(decl)
case *syntax.VarDecl:
l = append(l, p.varDecl(decl)...)
case *syntax.ConstDecl:
l = append(l, p.constDecl(decl, &cs)...)
case *syntax.TypeDecl:
l = append(l, p.typeDecl(decl))
case *syntax.FuncDecl:
l = append(l, p.funcDecl(decl))
default:
panic("unhandled Decl")
}
}
return
}
func (p *noder) importDecl(imp *syntax.ImportDecl) {
val := p.basicLit(imp.Path)
ipkg := importfile(&val)
if ipkg == nil {
if nerrors == 0 {
Fatalf("phase error in import")
}
return
}
ipkg.Direct = true
var my *types.Sym
if imp.LocalPkgName != nil {
my = p.name(imp.LocalPkgName)
} else {
my = lookup(ipkg.Name)
}
pack := p.nod(imp, OPACK, nil, nil)
pack.Sym = my
pack.Name.Pkg = ipkg
switch my.Name {
case ".":
importdot(ipkg, pack)
return
case "init":
yyerrorl(pack.Pos, "cannot import package as init - init must be a func")
return
case "_":
return
}
if my.Def != nil {
redeclare(pack.Pos, my, "as imported package name")
}
my.Def = asTypesNode(pack)
my.Lastlineno = pack.Pos
my.Block = 1 // at top level
}
func (p *noder) varDecl(decl *syntax.VarDecl) []*Node {
names := p.declNames(decl.NameList)
typ := p.typeExprOrNil(decl.Type)
var exprs []*Node
if decl.Values != nil {
exprs = p.exprList(decl.Values)
}
p.lineno(decl)
return variter(names, typ, exprs)
}
// constState tracks state between constant specifiers within a
// declaration group. This state is kept separate from noder so nested
// constant declarations are handled correctly (e.g., issue 15550).
type constState struct {
group *syntax.Group
typ *Node
values []*Node
iota int64
}
func (p *noder) constDecl(decl *syntax.ConstDecl, cs *constState) []*Node {
if decl.Group == nil || decl.Group != cs.group {
*cs = constState{
group: decl.Group,
}
}
names := p.declNames(decl.NameList)
typ := p.typeExprOrNil(decl.Type)
var values []*Node
if decl.Values != nil {
values = p.exprList(decl.Values)
cs.typ, cs.values = typ, values
} else {
if typ != nil {
yyerror("const declaration cannot have type without expression")
}
typ, values = cs.typ, cs.values
}
var nn []*Node
for i, n := range names {
if i >= len(values) {
yyerror("missing value in const declaration")
break
}
v := values[i]
if decl.Values == nil {
v = treecopy(v, n.Pos)
}
n.Op = OLITERAL
declare(n, dclcontext)
n.Name.Param.Ntype = typ
n.Name.Defn = v
n.SetIota(cs.iota)
nn = append(nn, p.nod(decl, ODCLCONST, n, nil))
}
if len(values) > len(names) {
yyerror("extra expression in const declaration")
}
cs.iota++
return nn
}
func (p *noder) typeDecl(decl *syntax.TypeDecl) *Node {
n := p.declName(decl.Name)
n.Op = OTYPE
declare(n, dclcontext)
// decl.Type may be nil but in that case we got a syntax error during parsing
typ := p.typeExprOrNil(decl.Type)
param := n.Name.Param
param.Ntype = typ
param.Pragma = decl.Pragma
param.Alias = decl.Alias
if param.Alias && param.Pragma != 0 {
yyerror("cannot specify directive with type alias")
param.Pragma = 0
}
return p.nod(decl, ODCLTYPE, n, nil)
}
func (p *noder) declNames(names []*syntax.Name) []*Node {
var nodes []*Node
for _, name := range names {
nodes = append(nodes, p.declName(name))
}
return nodes
}
func (p *noder) declName(name *syntax.Name) *Node {
return p.setlineno(name, dclname(p.name(name)))
}
func (p *noder) funcDecl(fun *syntax.FuncDecl) *Node {
name := p.name(fun.Name)
t := p.signature(fun.Recv, fun.Type)
f := p.nod(fun, ODCLFUNC, nil, nil)
if fun.Recv == nil {
if name.Name == "init" {
name = renameinit()
if t.List.Len() > 0 || t.Rlist.Len() > 0 {
yyerrorl(f.Pos, "func init must have no arguments and no return values")
}
}
if localpkg.Name == "main" && name.Name == "main" {
if t.List.Len() > 0 || t.Rlist.Len() > 0 {
yyerrorl(f.Pos, "func main must have no arguments and no return values")
}
}
} else {
f.Func.Shortname = name
name = nblank.Sym // filled in by typecheckfunc
}
f.Func.Nname = p.setlineno(fun.Name, newfuncname(name))
f.Func.Nname.Name.Defn = f
f.Func.Nname.Name.Param.Ntype = t
pragma := fun.Pragma
f.Func.Pragma = fun.Pragma
f.SetNoescape(pragma&Noescape != 0)
if pragma&Systemstack != 0 && pragma&Nosplit != 0 {
yyerrorl(f.Pos, "go:nosplit and go:systemstack cannot be combined")
}
if fun.Recv == nil {
declare(f.Func.Nname, PFUNC)
}
p.funcBody(f, fun.Body)
if fun.Body != nil {
if f.Noescape() {
yyerrorl(f.Pos, "can only use //go:noescape with external func implementations")
}
} else {
if pure_go || strings.HasPrefix(f.funcname(), "init.") {
yyerrorl(f.Pos, "missing function body")
}
}
return f
}
func (p *noder) signature(recv *syntax.Field, typ *syntax.FuncType) *Node {
n := p.nod(typ, OTFUNC, nil, nil)
if recv != nil {
n.Left = p.param(recv, false, false)
}
n.List.Set(p.params(typ.ParamList, true))
n.Rlist.Set(p.params(typ.ResultList, false))
return n
}
func (p *noder) params(params []*syntax.Field, dddOk bool) []*Node {
var nodes []*Node
for i, param := range params {
p.lineno(param)
nodes = append(nodes, p.param(param, dddOk, i+1 == len(params)))
}
return nodes
}
func (p *noder) param(param *syntax.Field, dddOk, final bool) *Node {
var name *types.Sym
if param.Name != nil {
name = p.name(param.Name)
}
typ := p.typeExpr(param.Type)
n := p.nodSym(param, ODCLFIELD, typ, name)
// rewrite ...T parameter
if typ.Op == ODDD {
if !dddOk {
yyerror("cannot use ... in receiver or result parameter list")
} else if !final {
yyerror("can only use ... with final parameter in list")
}
typ.Op = OTARRAY
typ.Right = typ.Left
typ.Left = nil
n.SetIsddd(true)
if n.Left != nil {
n.Left.SetIsddd(true)
}
}
return n
}
func (p *noder) exprList(expr syntax.Expr) []*Node {
if list, ok := expr.(*syntax.ListExpr); ok {
return p.exprs(list.ElemList)
}
return []*Node{p.expr(expr)}
}
func (p *noder) exprs(exprs []syntax.Expr) []*Node {
var nodes []*Node
for _, expr := range exprs {
nodes = append(nodes, p.expr(expr))
}
return nodes
}
func (p *noder) expr(expr syntax.Expr) *Node {
p.lineno(expr)
switch expr := expr.(type) {
case nil, *syntax.BadExpr:
return nil
case *syntax.Name:
return p.mkname(expr)
case *syntax.BasicLit:
return p.setlineno(expr, nodlit(p.basicLit(expr)))
case *syntax.CompositeLit:
n := p.nod(expr, OCOMPLIT, nil, nil)
if expr.Type != nil {
n.Right = p.expr(expr.Type)
}
l := p.exprs(expr.ElemList)
for i, e := range l {
l[i] = p.wrapname(expr.ElemList[i], e)
}
n.List.Set(l)
lineno = p.makeXPos(expr.Rbrace)
return n
case *syntax.KeyValueExpr:
// use position of expr.Key rather than of expr (which has position of ':')
return p.nod(expr.Key, OKEY, p.expr(expr.Key), p.wrapname(expr.Value, p.expr(expr.Value)))
case *syntax.FuncLit:
return p.funcLit(expr)
case *syntax.ParenExpr:
return p.nod(expr, OPAREN, p.expr(expr.X), nil)
case *syntax.SelectorExpr:
// parser.new_dotname
obj := p.expr(expr.X)
if obj.Op == OPACK {
obj.Name.SetUsed(true)
return oldname(restrictlookup(expr.Sel.Value, obj.Name.Pkg))
}
return p.setlineno(expr, nodSym(OXDOT, obj, p.name(expr.Sel)))
case *syntax.IndexExpr:
return p.nod(expr, OINDEX, p.expr(expr.X), p.expr(expr.Index))
case *syntax.SliceExpr:
op := OSLICE
if expr.Full {
op = OSLICE3
}
n := p.nod(expr, op, p.expr(expr.X), nil)
var index [3]*Node
for i, x := range expr.Index {
if x != nil {
index[i] = p.expr(x)
}
}
n.SetSliceBounds(index[0], index[1], index[2])
return n
case *syntax.AssertExpr:
return p.nod(expr, ODOTTYPE, p.expr(expr.X), p.typeExpr(expr.Type))
case *syntax.Operation:
if expr.Op == syntax.Add && expr.Y != nil {
return p.sum(expr)
}
x := p.expr(expr.X)
if expr.Y == nil {
if expr.Op == syntax.And {
x = unparen(x) // TODO(mdempsky): Needed?
if x.Op == OCOMPLIT {
// Special case for &T{...}: turn into (*T){...}.
// TODO(mdempsky): Switch back to p.nod after we
// get rid of gcCompat.
x.Right = nod(OIND, x.Right, nil)
x.Right.SetImplicit(true)
return x
}
}
return p.nod(expr, p.unOp(expr.Op), x, nil)
}
return p.nod(expr, p.binOp(expr.Op), x, p.expr(expr.Y))
case *syntax.CallExpr:
n := p.nod(expr, OCALL, p.expr(expr.Fun), nil)
n.List.Set(p.exprs(expr.ArgList))
n.SetIsddd(expr.HasDots)
return n
case *syntax.ArrayType:
var len *Node
if expr.Len != nil {
len = p.expr(expr.Len)
} else {
len = p.nod(expr, ODDD, nil, nil)
}
return p.nod(expr, OTARRAY, len, p.typeExpr(expr.Elem))
case *syntax.SliceType:
return p.nod(expr, OTARRAY, nil, p.typeExpr(expr.Elem))
case *syntax.DotsType:
return p.nod(expr, ODDD, p.typeExpr(expr.Elem), nil)
case *syntax.StructType:
return p.structType(expr)
case *syntax.InterfaceType:
return p.interfaceType(expr)
case *syntax.FuncType:
return p.signature(nil, expr)
case *syntax.MapType:
return p.nod(expr, OTMAP, p.typeExpr(expr.Key), p.typeExpr(expr.Value))
case *syntax.ChanType:
n := p.nod(expr, OTCHAN, p.typeExpr(expr.Elem), nil)
n.SetTChanDir(p.chanDir(expr.Dir))
return n
case *syntax.TypeSwitchGuard:
n := p.nod(expr, OTYPESW, nil, p.expr(expr.X))
if expr.Lhs != nil {
n.Left = p.declName(expr.Lhs)
if n.Left.isBlank() {
yyerror("invalid variable name %v in type switch", n.Left)
}
}
return n
}
panic("unhandled Expr")
}
// sum efficiently handles very large summation expressions (such as
// in issue #16394). In particular, it avoids left recursion and
// collapses string literals.
func (p *noder) sum(x syntax.Expr) *Node {
// While we need to handle long sums with asymptotic
// efficiency, the vast majority of sums are very small: ~95%
// have only 2 or 3 operands, and ~99% of string literals are
// never concatenated.
adds := make([]*syntax.Operation, 0, 2)
for {
add, ok := x.(*syntax.Operation)
if !ok || add.Op != syntax.Add || add.Y == nil {
break
}
adds = append(adds, add)
x = add.X
}
// nstr is the current rightmost string literal in the
// summation (if any), and chunks holds its accumulated
// substrings.
//
// Consider the expression x + "a" + "b" + "c" + y. When we
// reach the string literal "a", we assign nstr to point to
// its corresponding Node and initialize chunks to {"a"}.
// Visiting the subsequent string literals "b" and "c", we
// simply append their values to chunks. Finally, when we
// reach the non-constant operand y, we'll join chunks to form
// "abc" and reassign the "a" string literal's value.
//
// N.B., we need to be careful about named string constants
// (indicated by Sym != nil) because 1) we can't modify their
// value, as doing so would affect other uses of the string
// constant, and 2) they may have types, which we need to
// handle correctly. For now, we avoid these problems by
// treating named string constants the same as non-constant
// operands.
var nstr *Node
chunks := make([]string, 0, 1)
n := p.expr(x)
if Isconst(n, CTSTR) && n.Sym == nil {
nstr = n
chunks = append(chunks, nstr.Val().U.(string))
}
for i := len(adds) - 1; i >= 0; i-- {
add := adds[i]
r := p.expr(add.Y)
if Isconst(r, CTSTR) && r.Sym == nil {
if nstr != nil {
// Collapse r into nstr instead of adding to n.
chunks = append(chunks, r.Val().U.(string))
continue
}
nstr = r
chunks = append(chunks, nstr.Val().U.(string))
} else {
if len(chunks) > 1 {
nstr.SetVal(Val{U: strings.Join(chunks, "")})
}
nstr = nil
chunks = chunks[:0]
}
n = p.nod(add, OADD, n, r)
}
if len(chunks) > 1 {
nstr.SetVal(Val{U: strings.Join(chunks, "")})
}
return n
}
func (p *noder) typeExpr(typ syntax.Expr) *Node {
// TODO(mdempsky): Be stricter? typecheck should handle errors anyway.
return p.expr(typ)
}
func (p *noder) typeExprOrNil(typ syntax.Expr) *Node {
if typ != nil {
return p.expr(typ)
}
return nil
}
func (p *noder) chanDir(dir syntax.ChanDir) types.ChanDir {
switch dir {
case 0:
return types.Cboth
case syntax.SendOnly:
return types.Csend
case syntax.RecvOnly:
return types.Crecv
}
panic("unhandled ChanDir")
}
func (p *noder) structType(expr *syntax.StructType) *Node {
var l []*Node
for i, field := range expr.FieldList {
p.lineno(field)
var n *Node
if field.Name == nil {
n = p.embedded(field.Type)
} else {
n = p.nodSym(field, ODCLFIELD, p.typeExpr(field.Type), p.name(field.Name))
}
if i < len(expr.TagList) && expr.TagList[i] != nil {
n.SetVal(p.basicLit(expr.TagList[i]))
}
l = append(l, n)
}
p.lineno(expr)
n := p.nod(expr, OTSTRUCT, nil, nil)
n.List.Set(l)
return n
}
func (p *noder) interfaceType(expr *syntax.InterfaceType) *Node {
var l []*Node
for _, method := range expr.MethodList {
p.lineno(method)
var n *Node
if method.Name == nil {
n = p.nodSym(method, ODCLFIELD, oldname(p.packname(method.Type)), nil)
} else {
mname := p.name(method.Name)
sig := p.typeExpr(method.Type)
sig.Left = fakeRecv()
n = p.nodSym(method, ODCLFIELD, sig, mname)
ifacedcl(n)
}
l = append(l, n)
}
n := p.nod(expr, OTINTER, nil, nil)
n.List.Set(l)
return n
}
func (p *noder) packname(expr syntax.Expr) *types.Sym {
switch expr := expr.(type) {
case *syntax.Name:
name := p.name(expr)
if n := oldname(name); n.Name != nil && n.Name.Pack != nil {
n.Name.Pack.Name.SetUsed(true)
}
return name
case *syntax.SelectorExpr:
name := p.name(expr.X.(*syntax.Name))
var pkg *types.Pkg
if asNode(name.Def) == nil || asNode(name.Def).Op != OPACK {
yyerror("%v is not a package", name)
pkg = localpkg
} else {
asNode(name.Def).Name.SetUsed(true)
pkg = asNode(name.Def).Name.Pkg
}
return restrictlookup(expr.Sel.Value, pkg)
}
panic(fmt.Sprintf("unexpected packname: %#v", expr))
}
func (p *noder) embedded(typ syntax.Expr) *Node {
op, isStar := typ.(*syntax.Operation)
if isStar {
if op.Op != syntax.Mul || op.Y != nil {
panic("unexpected Operation")
}
typ = op.X
}
sym := p.packname(typ)
n := p.nodSym(typ, ODCLFIELD, oldname(sym), lookup(sym.Name))
n.SetEmbedded(true)
if isStar {
n.Left = p.nod(op, OIND, n.Left, nil)
}
return n
}
func (p *noder) stmts(stmts []syntax.Stmt) []*Node {
return p.stmtsFall(stmts, false)
}
func (p *noder) stmtsFall(stmts []syntax.Stmt, fallOK bool) []*Node {
var nodes []*Node
for i, stmt := range stmts {
s := p.stmtFall(stmt, fallOK && i+1 == len(stmts))
if s == nil {
} else if s.Op == OBLOCK && s.Ninit.Len() == 0 {
nodes = append(nodes, s.List.Slice()...)
} else {
nodes = append(nodes, s)
}
}
return nodes
}
func (p *noder) stmt(stmt syntax.Stmt) *Node {
return p.stmtFall(stmt, false)
}
func (p *noder) stmtFall(stmt syntax.Stmt, fallOK bool) *Node {
p.lineno(stmt)
switch stmt := stmt.(type) {
case *syntax.EmptyStmt:
return nil
case *syntax.LabeledStmt:
return p.labeledStmt(stmt, fallOK)
case *syntax.BlockStmt:
l := p.blockStmt(stmt)
if len(l) == 0 {
// TODO(mdempsky): Line number?
return nod(OEMPTY, nil, nil)
}
return liststmt(l)
case *syntax.ExprStmt:
return p.wrapname(stmt, p.expr(stmt.X))
case *syntax.SendStmt:
return p.nod(stmt, OSEND, p.expr(stmt.Chan), p.expr(stmt.Value))
case *syntax.DeclStmt:
return liststmt(p.decls(stmt.DeclList))
case *syntax.AssignStmt:
if stmt.Op != 0 && stmt.Op != syntax.Def {
n := p.nod(stmt, OASOP, p.expr(stmt.Lhs), p.expr(stmt.Rhs))
n.SetImplicit(stmt.Rhs == syntax.ImplicitOne)
n.SetSubOp(p.binOp(stmt.Op))
return n
}
n := p.nod(stmt, OAS, nil, nil) // assume common case
rhs := p.exprList(stmt.Rhs)
lhs := p.assignList(stmt.Lhs, n, stmt.Op == syntax.Def)
if len(lhs) == 1 && len(rhs) == 1 {
// common case
n.Left = lhs[0]
n.Right = rhs[0]
} else {
n.Op = OAS2
n.List.Set(lhs)
n.Rlist.Set(rhs)
}
return n
case *syntax.BranchStmt:
var op Op
switch stmt.Tok {
case syntax.Break:
op = OBREAK
case syntax.Continue:
op = OCONTINUE
case syntax.Fallthrough:
if !fallOK {
yyerror("fallthrough statement out of place")
}
op = OFALL
case syntax.Goto:
op = OGOTO
default:
panic("unhandled BranchStmt")
}
n := p.nod(stmt, op, nil, nil)
if stmt.Label != nil {
n.Left = p.newname(stmt.Label)
}
return n
case *syntax.CallStmt:
var op Op
switch stmt.Tok {
case syntax.Defer:
op = ODEFER
case syntax.Go:
op = OPROC
default:
panic("unhandled CallStmt")
}
return p.nod(stmt, op, p.expr(stmt.Call), nil)
case *syntax.ReturnStmt:
var results []*Node
if stmt.Results != nil {
results = p.exprList(stmt.Results)
}
n := p.nod(stmt, ORETURN, nil, nil)
n.List.Set(results)
if n.List.Len() == 0 && Curfn != nil {
for _, ln := range Curfn.Func.Dcl {
if ln.Class() == PPARAM {
continue
}
if ln.Class() != PPARAMOUT {
break
}
if asNode(ln.Sym.Def) != ln {
yyerror("%s is shadowed during return", ln.Sym.Name)
}
}
}
return n
case *syntax.IfStmt:
return p.ifStmt(stmt)
case *syntax.ForStmt:
return p.forStmt(stmt)
case *syntax.SwitchStmt:
return p.switchStmt(stmt)
case *syntax.SelectStmt:
return p.selectStmt(stmt)
}
panic("unhandled Stmt")
}
func (p *noder) assignList(expr syntax.Expr, defn *Node, colas bool) []*Node {
if !colas {
return p.exprList(expr)
}
defn.SetColas(true)
var exprs []syntax.Expr
if list, ok := expr.(*syntax.ListExpr); ok {
exprs = list.ElemList
} else {
exprs = []syntax.Expr{expr}
}
res := make([]*Node, len(exprs))
seen := make(map[*types.Sym]bool, len(exprs))
newOrErr := false
for i, expr := range exprs {
p.lineno(expr)
res[i] = nblank
name, ok := expr.(*syntax.Name)
if !ok {
p.yyerrorpos(expr.Pos(), "non-name %v on left side of :=", p.expr(expr))
newOrErr = true
continue
}
sym := p.name(name)
if sym.IsBlank() {
continue
}
if seen[sym] {
p.yyerrorpos(expr.Pos(), "%v repeated on left side of :=", sym)
newOrErr = true
continue
}
seen[sym] = true
if sym.Block == types.Block {
res[i] = oldname(sym)
continue
}
newOrErr = true
n := newname(sym)
declare(n, dclcontext)
n.Name.Defn = defn
defn.Ninit.Append(nod(ODCL, n, nil))
res[i] = n
}
if !newOrErr {
yyerrorl(defn.Pos, "no new variables on left side of :=")
}
return res
}
func (p *noder) blockStmt(stmt *syntax.BlockStmt) []*Node {
p.openScope(stmt.Pos())
nodes := p.stmts(stmt.List)
p.closeScope(stmt.Rbrace)
return nodes
}
func (p *noder) ifStmt(stmt *syntax.IfStmt) *Node {
p.openScope(stmt.Pos())
n := p.nod(stmt, OIF, nil, nil)
if stmt.Init != nil {
n.Ninit.Set1(p.stmt(stmt.Init))
}
if stmt.Cond != nil {
n.Left = p.expr(stmt.Cond)
}
n.Nbody.Set(p.blockStmt(stmt.Then))
if stmt.Else != nil {
e := p.stmt(stmt.Else)
if e.Op == OBLOCK && e.Ninit.Len() == 0 {
n.Rlist.Set(e.List.Slice())
} else {
n.Rlist.Set1(e)
}
}
p.closeAnotherScope()
return n
}
func (p *noder) forStmt(stmt *syntax.ForStmt) *Node {
p.openScope(stmt.Pos())
var n *Node
if r, ok := stmt.Init.(*syntax.RangeClause); ok {
if stmt.Cond != nil || stmt.Post != nil {
panic("unexpected RangeClause")
}
n = p.nod(r, ORANGE, nil, p.expr(r.X))
if r.Lhs != nil {
n.List.Set(p.assignList(r.Lhs, n, r.Def))
}
} else {
n = p.nod(stmt, OFOR, nil, nil)
if stmt.Init != nil {
n.Ninit.Set1(p.stmt(stmt.Init))
}
if stmt.Cond != nil {
n.Left = p.expr(stmt.Cond)
}
if stmt.Post != nil {
n.Right = p.stmt(stmt.Post)
}
}
n.Nbody.Set(p.blockStmt(stmt.Body))
p.closeAnotherScope()
return n
}
func (p *noder) switchStmt(stmt *syntax.SwitchStmt) *Node {
p.openScope(stmt.Pos())
n := p.nod(stmt, OSWITCH, nil, nil)
if stmt.Init != nil {
n.Ninit.Set1(p.stmt(stmt.Init))
}
if stmt.Tag != nil {
n.Left = p.expr(stmt.Tag)
}
tswitch := n.Left
if tswitch != nil && tswitch.Op != OTYPESW {
tswitch = nil
}
n.List.Set(p.caseClauses(stmt.Body, tswitch, stmt.Rbrace))
p.closeScope(stmt.Rbrace)
return n
}
func (p *noder) caseClauses(clauses []*syntax.CaseClause, tswitch *Node, rbrace syntax.Pos) []*Node {
var nodes []*Node
for i, clause := range clauses {
p.lineno(clause)
if i > 0 {
p.closeScope(clause.Pos())
}
p.openScope(clause.Pos())
n := p.nod(clause, OXCASE, nil, nil)
if clause.Cases != nil {
n.List.Set(p.exprList(clause.Cases))
}
if tswitch != nil && tswitch.Left != nil {
nn := newname(tswitch.Left.Sym)
declare(nn, dclcontext)
n.Rlist.Set1(nn)
// keep track of the instances for reporting unused
nn.Name.Defn = tswitch
}
// Trim trailing empty statements. We omit them from
// the Node AST anyway, and it's easier to identify
// out-of-place fallthrough statements without them.
body := clause.Body
for len(body) > 0 {
if _, ok := body[len(body)-1].(*syntax.EmptyStmt); !ok {
break
}
body = body[:len(body)-1]
}
n.Nbody.Set(p.stmtsFall(body, true))
if l := n.Nbody.Len(); l > 0 && n.Nbody.Index(l-1).Op == OFALL {
if tswitch != nil {
yyerror("cannot fallthrough in type switch")
}
if i+1 == len(clauses) {
yyerror("cannot fallthrough final case in switch")
}
}
nodes = append(nodes, n)
}
if len(clauses) > 0 {
p.closeScope(rbrace)
}
return nodes
}
func (p *noder) selectStmt(stmt *syntax.SelectStmt) *Node {
n := p.nod(stmt, OSELECT, nil, nil)
n.List.Set(p.commClauses(stmt.Body, stmt.Rbrace))
return n
}
func (p *noder) commClauses(clauses []*syntax.CommClause, rbrace syntax.Pos) []*Node {
var nodes []*Node
for i, clause := range clauses {
p.lineno(clause)
if i > 0 {
p.closeScope(clause.Pos())
}
p.openScope(clause.Pos())
n := p.nod(clause, OXCASE, nil, nil)
if clause.Comm != nil {
n.List.Set1(p.stmt(clause.Comm))
}
n.Nbody.Set(p.stmts(clause.Body))
nodes = append(nodes, n)
}
if len(clauses) > 0 {
p.closeScope(rbrace)
}
return nodes
}
func (p *noder) labeledStmt(label *syntax.LabeledStmt, fallOK bool) *Node {
lhs := p.nod(label, OLABEL, p.newname(label.Label), nil)
var ls *Node
if label.Stmt != nil { // TODO(mdempsky): Should always be present.
ls = p.stmtFall(label.Stmt, fallOK)
}
lhs.Name.Defn = ls
l := []*Node{lhs}
if ls != nil {
if ls.Op == OBLOCK && ls.Ninit.Len() == 0 {
l = append(l, ls.List.Slice()...)
} else {
l = append(l, ls)
}
}
return liststmt(l)
}
var unOps = [...]Op{
syntax.Recv: ORECV,
syntax.Mul: OIND,
syntax.And: OADDR,
syntax.Not: ONOT,
syntax.Xor: OCOM,
syntax.Add: OPLUS,
syntax.Sub: OMINUS,
}
func (p *noder) unOp(op syntax.Operator) Op {
if uint64(op) >= uint64(len(unOps)) || unOps[op] == 0 {
panic("invalid Operator")
}
return unOps[op]
}
var binOps = [...]Op{
syntax.OrOr: OOROR,
syntax.AndAnd: OANDAND,
syntax.Eql: OEQ,
syntax.Neq: ONE,
syntax.Lss: OLT,
syntax.Leq: OLE,
syntax.Gtr: OGT,
syntax.Geq: OGE,
syntax.Add: OADD,
syntax.Sub: OSUB,
syntax.Or: OOR,
syntax.Xor: OXOR,
syntax.Mul: OMUL,
syntax.Div: ODIV,
syntax.Rem: OMOD,
syntax.And: OAND,
syntax.AndNot: OANDNOT,
syntax.Shl: OLSH,
syntax.Shr: ORSH,
}
func (p *noder) binOp(op syntax.Operator) Op {
if uint64(op) >= uint64(len(binOps)) || binOps[op] == 0 {
panic("invalid Operator")
}
return binOps[op]
}
func (p *noder) basicLit(lit *syntax.BasicLit) Val {
// TODO: Don't try to convert if we had syntax errors (conversions may fail).
// Use dummy values so we can continue to compile. Eventually, use a
// form of "unknown" literals that are ignored during type-checking so
// we can continue type-checking w/o spurious follow-up errors.
switch s := lit.Value; lit.Kind {
case syntax.IntLit:
x := new(Mpint)
x.SetString(s)
return Val{U: x}
case syntax.FloatLit:
x := newMpflt()
x.SetString(s)
return Val{U: x}
case syntax.ImagLit:
x := new(Mpcplx)
x.Imag.SetString(strings.TrimSuffix(s, "i"))
return Val{U: x}
case syntax.RuneLit:
var r rune
if u, err := strconv.Unquote(s); err == nil && len(u) > 0 {
// Package syntax already reported any errors.
// Check for them again though because 0 is a
// better fallback value for invalid rune
// literals than 0xFFFD.
if len(u) == 1 {
r = rune(u[0])
} else {
r, _ = utf8.DecodeRuneInString(u)
}
}
x := new(Mpint)
x.SetInt64(int64(r))
x.Rune = true
return Val{U: x}
case syntax.StringLit:
if len(s) > 0 && s[0] == '`' {
// strip carriage returns from raw string
s = strings.Replace(s, "\r", "", -1)
}
// Ignore errors because package syntax already reported them.
u, _ := strconv.Unquote(s)
return Val{U: u}
default:
panic("unhandled BasicLit kind")
}
}
func (p *noder) name(name *syntax.Name) *types.Sym {
return lookup(name.Value)
}
func (p *noder) mkname(name *syntax.Name) *Node {
// TODO(mdempsky): Set line number?
return mkname(p.name(name))
}
func (p *noder) newname(name *syntax.Name) *Node {
// TODO(mdempsky): Set line number?
return newname(p.name(name))
}
func (p *noder) wrapname(n syntax.Node, x *Node) *Node {
// These nodes do not carry line numbers.
// Introduce a wrapper node to give them the correct line.
switch x.Op {
case OTYPE, OLITERAL:
if x.Sym == nil {
break
}
fallthrough
case ONAME, ONONAME, OPACK:
x = p.nod(n, OPAREN, x, nil)
x.SetImplicit(true)
}
return x
}
func (p *noder) nod(orig syntax.Node, op Op, left, right *Node) *Node {
return p.setlineno(orig, nod(op, left, right))
}
func (p *noder) nodSym(orig syntax.Node, op Op, left *Node, sym *types.Sym) *Node {
return p.setlineno(orig, nodSym(op, left, sym))
}
func (p *noder) setlineno(src_ syntax.Node, dst *Node) *Node {
pos := src_.Pos()
if !pos.IsKnown() {
// TODO(mdempsky): Shouldn't happen. Fix package syntax.
return dst
}
dst.Pos = p.makeXPos(pos)
return dst
}
func (p *noder) lineno(n syntax.Node) {
if n == nil {
return
}
pos := n.Pos()
if !pos.IsKnown() {
// TODO(mdempsky): Shouldn't happen. Fix package syntax.
return
}
lineno = p.makeXPos(pos)
}
// error is called concurrently if files are parsed concurrently.
func (p *noder) error(err error) {
p.err <- err.(syntax.Error)
}
// pragmas that are allowed in the std lib, but don't have
// a syntax.Pragma value (see lex.go) associated with them.
var allowedStdPragmas = map[string]bool{
"go:cgo_export_static": true,
"go:cgo_export_dynamic": true,
"go:cgo_import_static": true,
"go:cgo_import_dynamic": true,
"go:cgo_ldflag": true,
"go:cgo_dynamic_linker": true,
"go:generate": true,
}
// pragma is called concurrently if files are parsed concurrently.
func (p *noder) pragma(pos syntax.Pos, text string) syntax.Pragma {
switch {
case strings.HasPrefix(text, "line "):
// line directives are handled by syntax package
panic("unreachable")
case strings.HasPrefix(text, "go:linkname "):
f := strings.Fields(text)
if len(f) != 3 {
p.error(syntax.Error{Pos: pos, Msg: "usage: //go:linkname localname linkname"})
break
}
p.linknames = append(p.linknames, linkname{pos, f[1], f[2]})
case strings.HasPrefix(text, "go:cgo_import_dynamic "):
// This is permitted for general use because Solaris
// code relies on it in golang.org/x/sys/unix and others.
fields := pragmaFields(text)
if len(fields) >= 4 {
lib := strings.Trim(fields[3], `"`)
if lib != "" && !safeArg(lib) && !isCgoGeneratedFile(pos) {
p.error(syntax.Error{Pos: pos, Msg: fmt.Sprintf("invalid library name %q in cgo_import_dynamic directive", lib)})
}
p.pragcgo(pos, text)
return pragmaValue("go:cgo_import_dynamic")
}
fallthrough
case strings.HasPrefix(text, "go:cgo_"):
// For security, we disallow //go:cgo_* directives other
// than cgo_import_dynamic outside cgo-generated files.
// Exception: they are allowed in the standard library, for runtime and syscall.
if !isCgoGeneratedFile(pos) && !compiling_std {
p.error(syntax.Error{Pos: pos, Msg: fmt.Sprintf("//%s only allowed in cgo-generated code", text)})
}
p.pragcgo(pos, text)
fallthrough // because of //go:cgo_unsafe_args
default:
verb := text
if i := strings.Index(text, " "); i >= 0 {
verb = verb[:i]
}
prag := pragmaValue(verb)
const runtimePragmas = Systemstack | Nowritebarrier | Nowritebarrierrec | Yeswritebarrierrec
if !compiling_runtime && prag&runtimePragmas != 0 {
p.error(syntax.Error{Pos: pos, Msg: fmt.Sprintf("//%s only allowed in runtime", verb)})
}
if prag == 0 && !allowedStdPragmas[verb] && compiling_std {
p.error(syntax.Error{Pos: pos, Msg: fmt.Sprintf("//%s is not allowed in the standard library", verb)})
}
return prag
}
return 0
}
// isCgoGeneratedFile reports whether pos is in a file
// generated by cgo, which is to say a file with name
// beginning with "_cgo_". Such files are allowed to
// contain cgo directives, and for security reasons
// (primarily misuse of linker flags), other files are not.
// See golang.org/issue/23672.
func isCgoGeneratedFile(pos syntax.Pos) bool {
return strings.HasPrefix(filepath.Base(filepath.Clean(fileh(pos.Base().Filename()))), "_cgo_")
}
// safeArg reports whether arg is a "safe" command-line argument,
// meaning that when it appears in a command-line, it probably
// doesn't have some special meaning other than its own name.
// This is copied from SafeArg in cmd/go/internal/load/pkg.go.
func safeArg(name string) bool {
if name == "" {
return false
}
c := name[0]
return '0' <= c && c <= '9' || 'A' <= c && c <= 'Z' || 'a' <= c && c <= 'z' || c == '.' || c == '_' || c == '/' || c >= utf8.RuneSelf
}
func mkname(sym *types.Sym) *Node {
n := oldname(sym)
if n.Name != nil && n.Name.Pack != nil {
n.Name.Pack.Name.SetUsed(true)
}
return n
}
func unparen(x *Node) *Node {
for x.Op == OPAREN {
x = x.Left
}
return x
}