func fileTokens(tf *token.File) (toks []string, err error) {
src, err := ioutil.ReadFile(tf.Name())
if err != nil {
return nil, err
}
s := &scanner.Scanner{}
s.Init(tf, src, nil, 0)
tokmap := make(TokenSet)
for {
_, tok, lit := s.Scan()
if tok == token.EOF {
break
}
if tok == token.STRING {
// XXX: what if strings are misspelled?
lit = lit[1 : len(lit)-1]
}
tokmap[lit] = struct{}{}
}
for k, _ := range tokmap {
toks = append(toks, k)
}
return toks, nil
}
func (d *DIBuilder) getFile(file *token.File) llvm.Metadata {
if diFile := d.files[file]; diFile.C != nil {
return diFile
}
diFile := d.builder.CreateFile(d.remapFilePath(file.Name()), "")
d.files[file] = diFile
return diFile
}
func (d *DIBuilder) getFile(file *token.File) llvm.Value {
if diFile := d.files[file]; !diFile.IsNil() {
return diFile
}
diFile := d.builder.CreateFile(d.remapFilePath(file.Name()), "")
d.files[file] = diFile
return diFile
}
// findQueryPos searches fset for filename and translates the
// specified file-relative byte offsets into token.Pos form. It
// returns an error if the file was not found or the offsets were out
// of bounds.
//
func findQueryPos(fset *token.FileSet, filename string, startOffset, endOffset int) (start, end token.Pos, err error) {
var file *token.File
fset.Iterate(func(f *token.File) bool {
if sameFile(filename, f.Name()) {
// (f.Name() is absolute)
file = f
return false // done
}
return true // continue
})
if file == nil {
err = fmt.Errorf("couldn't find file containing position")
return
}
// Range check [start..end], inclusive of both end-points.
if 0 <= startOffset && startOffset <= file.Size() {
start = file.Pos(int(startOffset))
} else {
err = fmt.Errorf("start position is beyond end of file")
return
}
if 0 <= endOffset && endOffset <= file.Size() {
end = file.Pos(int(endOffset))
} else {
err = fmt.Errorf("end position is beyond end of file")
return
}
return
}
// parseQueryPos parses a string of the form "file:pos" or
// file:start,end" where pos, start, end match #%d and represent byte
// offsets, and returns the extent to which it refers.
//
// (Numbers without a '#' prefix are reserved for future use,
// e.g. to indicate line/column positions.)
//
func parseQueryPos(fset *token.FileSet, queryPos string) (start, end token.Pos, err error) {
if queryPos == "" {
err = fmt.Errorf("no source position specified (-pos flag)")
return
}
colon := strings.LastIndex(queryPos, ":")
if colon < 0 {
err = fmt.Errorf("invalid source position -pos=%q", queryPos)
return
}
filename, offset := queryPos[:colon], queryPos[colon+1:]
startOffset := -1
endOffset := -1
if hyphen := strings.Index(offset, ","); hyphen < 0 {
// e.g. "foo.go:#123"
startOffset = parseOctothorpDecimal(offset)
endOffset = startOffset
} else {
// e.g. "foo.go:#123,#456"
startOffset = parseOctothorpDecimal(offset[:hyphen])
endOffset = parseOctothorpDecimal(offset[hyphen+1:])
}
if startOffset < 0 || endOffset < 0 {
err = fmt.Errorf("invalid -pos offset %q", offset)
return
}
var file *token.File
fset.Iterate(func(f *token.File) bool {
if sameFile(filename, f.Name()) {
// (f.Name() is absolute)
file = f
return false // done
}
return true // continue
})
if file == nil {
err = fmt.Errorf("couldn't find file containing position -pos=%q", queryPos)
return
}
// Range check [start..end], inclusive of both end-points.
if 0 <= startOffset && startOffset <= file.Size() {
start = file.Pos(int(startOffset))
} else {
err = fmt.Errorf("start position is beyond end of file -pos=%q", queryPos)
return
}
if 0 <= endOffset && endOffset <= file.Size() {
end = file.Pos(int(endOffset))
} else {
err = fmt.Errorf("end position is beyond end of file -pos=%q", queryPos)
return
}
return
}
// createCompileUnit creates and returns debug metadata for the compile
// unit as a whole, using the first file in the file set as a representative
// (the choice of file is arbitrary).
func (d *DIBuilder) createCompileUnit() llvm.Metadata {
var file *token.File
d.fset.Iterate(func(f *token.File) bool {
file = f
return false
})
dir, err := os.Getwd()
if err != nil {
panic("could not get current directory: " + err.Error())
}
return d.builder.CreateCompileUnit(llvm.DICompileUnit{
Language: llvm.DW_LANG_Go,
File: d.remapFilePath(file.Name()),
Dir: dir,
Producer: "llgo",
})
}
// Init prepares the scanner S to tokenize the text src by setting the
// scanner at the beginning of src. The scanner uses the file set file
// for position information and it adds line information for each line.
// It is ok to re-use the same file when re-scanning the same file as
// line information which is already present is ignored. Init causes a
// panic if the file size does not match the src size.
//
// Calls to Scan will use the error handler err if they encounter a
// syntax error and err is not nil. Also, for each error encountered,
// the Scanner field ErrorCount is incremented by one. The mode parameter
// determines how comments, illegal characters, and semicolons are handled.
//
// Note that Init may call err if there is an error in the first character
// of the file.
//
func (S *Scanner) Init(file *token.File, src []byte, err ErrorHandler, mode uint) {
// Explicitly initialize all fields since a scanner may be reused.
if file.Size() != len(src) {
panic("file size does not match src len")
}
S.file = file
S.dir, _ = filepath.Split(file.Name())
S.src = src
S.err = err
S.mode = mode
S.ch = ' '
S.offset = 0
S.rdOffset = 0
S.lineOffset = 0
S.insertSemi = false
S.ErrorCount = 0
S.next()
}
// Init prepares the scanner s to tokenize the text src by setting the
// scanner at the beginning of src. The scanner uses the file set file
// for position information and it adds line information for each line.
// It is ok to re-use the same file when re-scanning the same file as
// line information which is already present is ignored. Init causes a
// panic if the file size does not match the src size.
//
// Calls to Scan will invoke the error handler err if they encounter a
// syntax error and err is not nil. Also, for each error encountered,
// the Scanner field ErrorCount is incremented by one. The mode parameter
// determines how comments are handled.
//
// Note that Init may call err if there is an error in the first character
// of the file.
//
func (s *Scanner) Init(file *token.File, src []byte, err ErrorHandler, mode Mode) {
// Explicitly initialize all fields since a scanner may be reused.
if file.Size() != len(src) {
panic("file size does not match src len")
}
s.file = file
s.dir, _ = filepath.Split(file.Name())
s.src = src
s.err = err
s.mode = mode
s.ch = ' '
s.offset = 0
s.rdOffset = 0
s.lineOffset = 0
s.insertSemi = false
s.ErrorCount = 0
s.next()
}
func (d *debugInfo) getCompileUnit(file *token.File) *debug.CompileUnitDescriptor {
if d.cu == nil {
d.cu = make(map[*token.File]*debug.CompileUnitDescriptor)
}
cu := d.cu[file]
if cu == nil {
var path string
if file != nil {
path = d.Fset.File(file.Pos(0)).Name()
}
cu = &debug.CompileUnitDescriptor{
Language: debug.DW_LANG_Go,
Path: debug.FileDescriptor(path),
Producer: "llgo",
Runtime: LLGORuntimeVersion,
}
d.cu[file] = cu
}
return cu
}
// fileOffsetToPos translates the specified file-relative byte offsets
// into token.Pos form. It returns an error if the file was not found
// or the offsets were out of bounds.
//
func fileOffsetToPos(file *token.File, startOffset, endOffset int) (start, end token.Pos, err error) {
// Range check [start..end], inclusive of both end-points.
if 0 <= startOffset && startOffset <= file.Size() {
start = file.Pos(int(startOffset))
} else {
err = fmt.Errorf("start position is beyond end of file")
return
}
if 0 <= endOffset && endOffset <= file.Size() {
end = file.Pos(int(endOffset))
} else {
err = fmt.Errorf("end position is beyond end of file")
return
}
return
}
func (s *Scanner) Init(file *token.File, src []byte, err ErrorHandler, mode Mode) {
if file.Size() != len(src) {
panic(fmt.Sprintf("file size (%d) does not match src len (%d)", file.Size(), len(src)))
}
s.tokScanner = scan.Scanner{Matcher: getTokenMatcher()}
s.errScanner = scan.Scanner{Matcher: getErrorMatcher()}
s.src = skipBOM(src)
s.tokScanner.SetSource(s.src)
s.errScanner.SetSource(s.src)
s.file = file
s.fileBase = s.file.Base()
s.dir, _ = filepath.Split(file.Name())
s.err = err
s.mode = mode
s.ErrorCount = 0
s.preSemi = false
s.semiPos = 0
}
func (s *Scanner) Init(file *token.File, src []byte, err ErrorHandler, mode Mode) {
//fmt.Println("Init src", strconv.Quote(string(src)), mode)
if file.Size() != len(src) {
panic(fmt.Sprintf("file size (%d) does not match src len (%d)", file.Size(), len(src)))
}
s.gombiScanner = newGombiScanner()
s.SetSource(skipBOM(src))
s.file = file
s.dir, _ = filepath.Split(file.Name())
s.err = err
s.mode = mode
s.ErrorCount = 0
s.lastIsPreSemi = false
s.commentAfterPreSemi = false
s.endOfLinePos = 0
s.endOfLine = 0
s.commentQueue.reset()
}
// Init prepares the scanner s to tokenize the text src by setting the
// scanner at the beginning of src. The scanner uses the file set file
// for position information and it adds line information for each line.
// It is ok to re-use the same file when re-scanning the same file as
// line information which is already present is ignored. Init causes a
// panic if the file size does not match the src size.
//
// Calls to Scan will invoke the error handler err if they encounter a
// syntax error and err is not nil. Also, for each error encountered,
// the Scanner field ErrorCount is incremented by one. The mode parameter
// determines how comments are handled.
//
// Note that Init may call err if there is an error in the first character
// of the file.
//
func (s *Scanner) Init(file *token.File, src []byte, err ErrorHandler, mode Mode) {
// Explicitly initialize all fields since a scanner may be reused.
if file.Size() != len(src) {
panic(fmt.Sprintf("file size (%d) does not match src len (%d)", file.Size(), len(src)))
}
s.file = file
s.dir, _ = filepath.Split(file.Name())
s.src = src
s.err = err
s.mode = mode
s.ch = ' '
s.offset = 0
s.rdOffset = 0
s.lineOffset = 0
s.insertSemi = false
s.ErrorCount = 0
s.next()
if s.ch == bom {
s.next() // ignore BOM at file beginning
}
}
func linenum(f *token.File, p token.Pos) int32 {
return int32(f.Line(p))
}
// TODO(adonovan): make this a method: func (*token.File) Contains(token.Pos)
func tokenFileContainsPos(f *token.File, pos token.Pos) bool {
p := int(pos)
base := f.Base()
return base <= p && p < base+f.Size()
}
func process(output io.Writer, input io.Reader) {
in := bufio.NewReader(input)
out := bufio.NewWriter(output)
var r rune
done := false
regex := make([]rune, 0, 8)
read := func() {
var er error
r, _, er = in.ReadRune()
if er == io.EOF {
done = true
} else if er != nil {
panic(er.Error())
}
}
skipws := func() {
for {
read()
if done {
break
}
if strings.IndexRune(" \n\t\r", r) == -1 {
break
}
}
}
var rules []*rule
usercode := false
familyn := 1
id := 0
newRule := func(family, index int) *rule {
x := new(rule)
rules = append(rules, x)
x.family = family
x.id = id
x.index = index
id++
return x
}
buf := make([]rune, 0, 8)
readCode := func() string {
if '{' != r {
panic("expected {")
}
buf = buf[:0]
nesting := 1
for {
buf = append(buf, r)
read()
if done {
panic("unmatched {")
}
if '{' == r {
nesting++
}
if '}' == r {
nesting--
if 0 == nesting {
break
}
}
}
buf = append(buf, r)
return string(buf)
}
var decls string
var parse func(int)
parse = func(family int) {
rulen := 0
declvar := func() {
decls += fmt.Sprintf("var a%d [%d]dfa\n", family, rulen)
}
for !done {
skipws()
if done {
break
}
regex = regex[:0]
if '>' == r {
if 0 == family {
panic("unmatched >")
}
x := newRule(family, -1)
x.code = "yylex = yylex.pop()\n"
declvar()
skipws()
x.code += readCode()
return
}
delim := r
read()
if done {
panic("unterminated pattern")
}
for {
if r == delim && (len(regex) == 0 || regex[len(regex)-1] != '\\') {
break
}
if '\n' == r {
panic("regex interrupted by newline")
}
regex = append(regex, r)
read()
if done {
panic("unterminated pattern")
}
}
if "" == string(regex) {
usercode = true
break
}
skipws()
if done {
panic("last pattern lacks action")
}
x := newRule(family, rulen)
rulen++
x.regex = make([]rune, len(regex))
copy(x.regex, regex)
nested := false
if '<' == r {
skipws()
if done {
panic("'<' lacks action")
}
x.code = fmt.Sprintf("yylex = yylex.push(%d)\n", familyn)
nested = true
}
x.code += readCode()
if nested {
familyn++
parse(familyn - 1)
}
}
if 0 != family {
panic("unmatched <")
}
x := newRule(family, -1)
x.code = "// [END]\n"
declvar()
}
parse(0)
if !usercode {
return
}
skipws()
buf = buf[:0]
for !done {
buf = append(buf, r)
read()
}
fs := token.NewFileSet()
t, err := parser.ParseFile(fs, "", string(buf), parser.ImportsOnly)
if err != nil {
panic(err.Error())
}
printer.Fprint(out, fs, t)
var file *token.File
fs.Iterate(func(f *token.File) bool {
file = f
return true
})
for m := file.LineCount(); m > 1; m-- {
i := 0
for '\n' != buf[i] {
i++
}
buf = buf[i+1:]
}
fmt.Fprintf(out, `import ("bufio";"io";"strings")
type dfa struct {
acc []bool
f []func(rune) int
id int
}
type family struct {
a []dfa
endcase int
}
`)
out.WriteString(decls)
out.WriteString("var a []family\n")
out.WriteString("func init() {\n")
fmt.Fprintf(out, "a = make([]family, %d)\n", familyn)
for _, x := range rules {
gen(out, x)
}
for i := 0; i < familyn; i++ {
fmt.Fprintf(out, "a[%d].a = a%d[:]\n", i, i)
}
out.WriteString(`}
func getAction(c *frame) int {
if -1 == c.match { return -1 }
c.action = c.fam.a[c.match].id
c.match = -1
return c.action
}
type frame struct {
atEOF bool
action, match, matchn, n int
buf []rune
text string
in *bufio.Reader
state []int
fam family
}
func newFrame(in *bufio.Reader, index int) *frame {
f := new(frame)
f.buf = make([]rune, 0, 128)
f.in = in
f.match = -1
f.fam = a[index]
f.state = make([]int, len(f.fam.a))
return f
}
type Lexer []*frame
func NewLexer(in io.Reader) Lexer {
stack := make([]*frame, 0, 4)
stack = append(stack, newFrame(bufio.NewReader(in), 0))
return stack
}
func (stack Lexer) isDone() bool {
return 1 == len(stack) && stack[0].atEOF
}
func (stack Lexer) nextAction() int {
c := stack[len(stack) - 1]
for {
if c.atEOF { return c.fam.endcase }
if c.n == len(c.buf) {
r,_,er := c.in.ReadRune()
switch er {
case nil: c.buf = append(c.buf, r)
case io.EOF:
c.atEOF = true
if c.n > 0 {
c.text = string(c.buf)
return getAction(c)
}
return c.fam.endcase
default: panic(er.Error())
}
}
jammed := true
r := c.buf[c.n]
for i, x := range c.fam.a {
if -1 == c.state[i] { continue }
c.state[i] = x.f[c.state[i]](r)
if -1 == c.state[i] { continue }
jammed = false
if x.acc[c.state[i]] {
if -1 == c.match || c.matchn < c.n+1 || c.match > i {
c.match = i
c.matchn = c.n+1
}
}
}
if jammed {
a := getAction(c)
if -1 == a { c.matchn = c.n + 1 }
c.n = 0
for i, _ := range c.state { c.state[i] = 0 }
c.text = string(c.buf[:c.matchn])
copy(c.buf, c.buf[c.matchn:])
c.buf = c.buf[:len(c.buf) - c.matchn]
return a
}
c.n++
}
panic("unreachable")
}
func (stack Lexer) push(index int) Lexer {
c := stack[len(stack) - 1]
return append(stack,
newFrame(bufio.NewReader(strings.NewReader(c.text)), index))
}
func (stack Lexer) pop() Lexer {
return stack[:len(stack) - 1]
}
func (stack Lexer) Text() string {
c := stack[len(stack) - 1]
return c.text
}
`)
if !*standalone {
writeLex(out, rules)
out.WriteString(string(buf))
out.Flush()
return
}
m := 0
const funmac = "NN_FUN"
for m < len(buf) {
m++
if funmac[:m] != string(buf[:m]) {
out.WriteString(string(buf[:m]))
buf = buf[m:]
m = 0
} else if funmac == string(buf[:m]) {
writeNNFun(out, rules)
buf = buf[m:]
m = 0
}
}
out.WriteString(string(buf))
out.Flush()
}
func getRangeLinesAtLeastOne(f *token.File, Pos, End token.Pos, fileSize int) (lines []int, firstLineNum int) {
lines = []int{}
firstLineNum = -1
l := f.Line(Pos)
for p := Pos; p <= End; p++ {
if f.Line(p) > l {
l = f.Line(p)
if firstLineNum == -1 {
firstLineNum = l
}
lines = append(lines, f.Offset(p))
}
}
if (int(End) == fileSize+f.Base()-1) || f.Line(End+1) > l {
lines = append(lines, f.Offset(End+1))
if firstLineNum == -1 {
firstLineNum = f.Line(End + 1)
}
}
if firstLineNum < 0 {
for p := End; ; p++ {
if f.Line(p) > l {
firstLineNum = l
lines = append(lines, f.Offset(p))
break
}
}
}
return
}
func process(output io.Writer, input io.Reader) error {
lineno := 1
in := bufio.NewReader(input)
out := bufio.NewWriter(output)
var r rune
read := func() bool {
var err error
r, _, err = in.ReadRune()
if err == io.EOF {
return true
}
if err != nil {
panic(err)
}
if r == '\n' {
lineno++
}
return false
}
skipws := func() bool {
for !read() {
if strings.IndexRune(" \n\t\r", r) == -1 {
return false
}
}
return true
}
var buf []rune
readCode := func() string {
if '{' != r {
panic(ErrExpectedLBrace)
}
buf = []rune{r}
nesting := 1
for {
if read() {
panic(ErrUnmatchedLBrace)
}
buf = append(buf, r)
if '{' == r {
nesting++
} else if '}' == r {
nesting--
if 0 == nesting {
break
}
}
}
return string(buf)
}
var root rule
needRootRAngle := false
var parse func(*rule) error
parse = func(node *rule) error {
for {
panicIf(skipws, ErrUnexpectedEOF)
if '<' == r {
if node != &root || len(node.kid) > 0 {
panic(ErrUnexpectedLAngle)
}
panicIf(skipws, ErrUnexpectedEOF)
node.startCode = readCode()
needRootRAngle = true
continue
} else if '>' == r {
if node == &root {
if !needRootRAngle {
panic(ErrUnmatchedRAngle)
}
}
if skipws() {
return ErrUnexpectedEOF
}
node.endCode = readCode()
return nil
}
delim := r
panicIf(read, ErrUnexpectedEOF)
var regex []rune
for {
if r == delim && (len(regex) == 0 || regex[len(regex)-1] != '\\') {
break
}
if '\n' == r {
return ErrUnexpectedNewline
}
regex = append(regex, r)
panicIf(read, ErrUnexpectedEOF)
}
if "" == string(regex) {
break
}
panicIf(skipws, ErrUnexpectedEOF)
x := new(rule)
x.id = fmt.Sprintf("%d", lineno)
node.kid = append(node.kid, x)
x.regex = make([]rune, len(regex))
copy(x.regex, regex)
if '<' == r {
panicIf(skipws, ErrUnexpectedEOF)
x.startCode = readCode()
parse(x)
} else {
x.code = readCode()
}
}
return nil
}
err := parse(&root)
if err != nil {
return err
}
buf = nil
for done := skipws(); !done; done = read() {
buf = append(buf, r)
}
fs := token.NewFileSet()
// Append a blank line to make things easier when there are only package and
// import declarations.
t, err := parser.ParseFile(fs, "", string(buf)+"\n", parser.ImportsOnly)
if err != nil {
panic(err)
}
printer.Fprint(out, fs, t)
var file *token.File
fs.Iterate(func(f *token.File) bool {
file = f
return true
})
// Skip over package and import declarations. This is why we appended a blank
// line above.
for m := file.LineCount(); m > 1; m-- {
i := 0
for '\n' != buf[i] {
i++
}
buf = buf[i+1:]
}
prefixReplacer.WriteString(out, lexertext)
for _, kid := range root.kid {
gen(out, kid)
}
prefixReplacer.WriteString(out, lexeroutro)
if !standalone {
writeLex(out, root)
out.WriteString(string(buf))
out.Flush()
if len(outFilename) > 0 {
gofmt()
}
return nil
}
m := 0
const funmac = "NN_FUN"
for m < len(buf) {
m++
if funmac[:m] != string(buf[:m]) {
out.WriteString(string(buf[:m]))
buf = buf[m:]
m = 0
} else if funmac == string(buf[:m]) {
writeNNFun(out, root)
buf = buf[m:]
m = 0
}
}
out.WriteString(string(buf))
out.Flush()
if len(outFilename) > 0 {
gofmt()
}
return nil
}
func process(output io.Writer, input io.Reader) {
lineno := 1
in := bufio.NewReader(input)
out := bufio.NewWriter(output)
var r rune
read := func() bool {
var err error
r, _, err = in.ReadRune()
if err == io.EOF {
return true
}
if err != nil {
panic(err)
}
if r == '\n' {
lineno++
}
return false
}
skipws := func() bool {
for !read() {
if strings.IndexRune(" \n\t\r", r) == -1 {
return false
}
}
return true
}
panicIf := func(f func() bool, err Error) {
if f() {
panic(err)
}
}
var buf []rune
readCode := func() string {
if '{' != r {
panic(ErrExpectedLBrace)
}
buf = []rune{r}
nesting := 1
for {
if read() {
panic(ErrUnmatchedLBrace)
}
buf = append(buf, r)
if '{' == r {
nesting++
} else if '}' == r {
nesting--
if 0 == nesting {
break
}
}
}
return string(buf)
}
var root rule
needRootRAngle := false
var parse func(*rule)
parse = func(node *rule) {
for {
panicIf(skipws, ErrUnexpectedEOF)
if '<' == r {
if node != &root || len(node.kid) > 0 {
panic(ErrUnexpectedLAngle)
}
panicIf(skipws, ErrUnexpectedEOF)
node.startCode = readCode()
needRootRAngle = true
continue
} else if '>' == r {
if node == &root {
if !needRootRAngle {
panic(ErrUnmatchedRAngle)
}
}
panicIf(skipws, ErrUnexpectedEOF)
node.endCode = readCode()
return
}
delim := r
panicIf(read, ErrUnexpectedEOF)
var regex []rune
for {
if r == delim && (len(regex) == 0 || regex[len(regex)-1] != '\\') {
break
}
if '\n' == r {
panic(ErrUnexpectedNewline)
}
regex = append(regex, r)
panicIf(read, ErrUnexpectedEOF)
}
if "" == string(regex) {
break
}
panicIf(skipws, ErrUnexpectedEOF)
x := new(rule)
x.id = fmt.Sprintf("%d", lineno)
node.kid = append(node.kid, x)
x.regex = make([]rune, len(regex))
copy(x.regex, regex)
if '<' == r {
panicIf(skipws, ErrUnexpectedEOF)
x.startCode = readCode()
parse(x)
} else {
x.code = readCode()
}
}
}
parse(&root)
buf = nil
for done := skipws(); !done; done = read() {
buf = append(buf, r)
}
fs := token.NewFileSet()
// Append a blank line to make things easier when there are only package and
// import declarations.
t, err := parser.ParseFile(fs, "", string(buf)+"\n", parser.ImportsOnly)
if err != nil {
panic(err)
}
printer.Fprint(out, fs, t)
var file *token.File
fs.Iterate(func(f *token.File) bool {
file = f
return true
})
// Skip over package and import declarations. This is why we appended a blank
// line above.
for m := file.LineCount(); m > 1; m-- {
i := 0
for '\n' != buf[i] {
i++
}
buf = buf[i+1:]
}
out.WriteString(`import ("bufio";"io";"strings")
type frame struct {
i int
s string
line, column int
}
type Lexer struct {
// The lexer runs in its own goroutine, and communicates via channel 'ch'.
ch chan frame
// We record the level of nesting because the action could return, and a
// subsequent call expects to pick up where it left off. In other words,
// we're simulating a coroutine.
// TODO: Support a channel-based variant that compatible with Go's yacc.
stack []frame
stale bool
// The 'l' and 'c' fields were added for
// https://github.com/wagerlabs/docker/blob/65694e801a7b80930961d70c69cba9f2465459be/buildfile.nex
// Since then, I introduced the built-in Line() and Column() functions.
l, c int
parseResult interface{}
// The following line makes it easy for scripts to insert fields in the
// generated code.
// [NEX_END_OF_LEXER_STRUCT]
}
// NewLexerWithInit creates a new Lexer object, runs the given callback on it,
// then returns it.
func NewLexerWithInit(in io.Reader, initFun func(*Lexer)) *Lexer {
type dfa struct {
acc []bool // Accepting states.
f []func(rune) int // Transitions.
startf, endf []int // Transitions at start and end of input.
nest []dfa
}
yylex := new(Lexer)
if initFun != nil {
initFun(yylex)
}
yylex.ch = make(chan frame)
var scan func(in *bufio.Reader, ch chan frame, family []dfa, line, column int)
scan = func(in *bufio.Reader, ch chan frame, family []dfa, line, column int) {
// Index of DFA and length of highest-precedence match so far.
matchi, matchn := 0, -1
var buf []rune
n := 0
checkAccept := func(i int, st int) bool {
// Higher precedence match? DFAs are run in parallel, so matchn is at most len(buf), hence we may omit the length equality check.
if family[i].acc[st] && (matchn < n || matchi > i) {
matchi, matchn = i, n
return true
}
return false
}
var state [][2]int
for i := 0; i < len(family); i++ {
mark := make([]bool, len(family[i].startf))
// Every DFA starts at state 0.
st := 0
for {
state = append(state, [2]int{i, st})
mark[st] = true
// As we're at the start of input, follow all ^ transitions and append to our list of start states.
st = family[i].startf[st]
if -1 == st || mark[st] { break }
// We only check for a match after at least one transition.
checkAccept(i, st)
}
}
atEOF := false
for {
if n == len(buf) && !atEOF {
r,_,err := in.ReadRune()
switch err {
case io.EOF: atEOF = true
case nil: buf = append(buf, r)
default: panic(err)
}
}
if !atEOF {
r := buf[n]
n++
var nextState [][2]int
for _, x := range state {
x[1] = family[x[0]].f[x[1]](r)
if -1 == x[1] { continue }
nextState = append(nextState, x)
checkAccept(x[0], x[1])
}
state = nextState
} else {
dollar: // Handle $.
for _, x := range state {
mark := make([]bool, len(family[x[0]].endf))
for {
mark[x[1]] = true
x[1] = family[x[0]].endf[x[1]]
if -1 == x[1] || mark[x[1]] { break }
if checkAccept(x[0], x[1]) {
// Unlike before, we can break off the search. Now that we're at the end, there's no need to maintain the state of each DFA.
break dollar
}
}
}
state = nil
}
if state == nil {
lcUpdate := func(r rune) {
if r == '\n' {
line++
column = 0
} else {
column++
}
}
// All DFAs stuck. Return last match if it exists, otherwise advance by one rune and restart all DFAs.
if matchn == -1 {
if len(buf) == 0 { // This can only happen at the end of input.
break
}
lcUpdate(buf[0])
buf = buf[1:]
} else {
text := string(buf[:matchn])
buf = buf[matchn:]
matchn = -1
ch <- frame{matchi, text, line, column}
if len(family[matchi].nest) > 0 {
scan(bufio.NewReader(strings.NewReader(text)), ch, family[matchi].nest, line, column)
}
if atEOF {
break
}
for _, r := range text {
lcUpdate(r)
}
}
n = 0
for i := 0; i < len(family); i++ {
state = append(state, [2]int{i, 0})
}
}
}
ch <- frame{-1, "", line, column}
}
go scan(bufio.NewReader(in), yylex.ch, []dfa{`)
for _, kid := range root.kid {
gen(out, kid)
}
out.WriteString(`}, 0, 0)
return yylex
}
func NewLexer(in io.Reader) *Lexer {
return NewLexerWithInit(in, nil)
}
// Text returns the matched text.
func (yylex *Lexer) Text() string {
return yylex.stack[len(yylex.stack) - 1].s
}
// Line returns the current line number.
// The first line is 0.
func (yylex *Lexer) Line() int {
return yylex.stack[len(yylex.stack) - 1].line
}
// Column returns the current column number.
// The first column is 0.
func (yylex *Lexer) Column() int {
return yylex.stack[len(yylex.stack) - 1].column
}
func (yylex *Lexer) next(lvl int) int {
if lvl == len(yylex.stack) {
l, c := 0, 0
if lvl > 0 {
l, c = yylex.stack[lvl - 1].line, yylex.stack[lvl - 1].column
}
yylex.stack = append(yylex.stack, frame{0, "", l, c})
}
if lvl == len(yylex.stack) - 1 {
p := &yylex.stack[lvl]
*p = <-yylex.ch
yylex.stale = false
} else {
yylex.stale = true
}
return yylex.stack[lvl].i
}
func (yylex *Lexer) pop() {
yylex.stack = yylex.stack[:len(yylex.stack) - 1]
}
`)
if !*standalone {
writeLex(out, root)
out.WriteString(string(buf))
out.Flush()
return
}
m := 0
const funmac = "NN_FUN"
for m < len(buf) {
m++
if funmac[:m] != string(buf[:m]) {
out.WriteString(string(buf[:m]))
buf = buf[m:]
m = 0
} else if funmac == string(buf[:m]) {
writeNNFun(out, root)
buf = buf[m:]
m = 0
}
}
out.WriteString(string(buf))
out.Flush()
}
func GetRangeLines(f *token.File, Pos, End token.Pos, fileSize int) (lines []int, firstLineNum int) {
lines = []int{}
firstLineNum = -1
l := f.Line(Pos)
for p := Pos; p <= End; p++ {
if f.Line(p) > l {
l = f.Line(p)
if firstLineNum == -1 {
firstLineNum = l
}
lines = append(lines, f.Offset(p))
}
}
print(End)
print(" -> ")
println(fileSize + f.Base() - 1)
if (int(End) == fileSize+f.Base()-1) || f.Line(End+1) > l {
lines = append(lines, f.Offset(End+1))
if firstLineNum == -1 {
firstLineNum = f.Line(End + 1)
}
}
return
}
func printDecls(tf *token.File, f *ast.File) {
for _, d := range f.Decls {
fmt.Printf("> %d %d\n", tf.Offset(d.Pos()), tf.Offset(d.End()))
}
}
func GetLines(f *token.File) []int {
lines := make([]int, 0, 20)
l := -1
for i := f.Base(); i < f.Base()+f.Size(); i++ {
if f.Line(token.Pos(i)) > l {
l = f.Line(token.Pos(i))
lines = append(lines, f.Offset(token.Pos(i)))
}
}
return lines
}
func process(output io.Writer, input io.Reader) {
lineno := 1
in := bufio.NewReader(input)
out := bufio.NewWriter(output)
var r rune
read := func() bool {
var err error
r, _, err = in.ReadRune()
if err == io.EOF {
return true
}
if err != nil {
panic(err)
}
if r == '\n' {
lineno++
}
return false
}
skipws := func() bool {
for !read() {
if strings.IndexRune(" \n\t\r", r) == -1 {
return false
}
}
return true
}
panicIf := func(f func() bool, err Error) {
if f() {
panic(err)
}
}
var buf []rune
readCode := func() string {
if '{' != r {
panic(ErrExpectedLBrace)
}
buf = []rune{r}
nesting := 1
for {
if read() {
panic(ErrUnmatchedLBrace)
}
buf = append(buf, r)
if '{' == r {
nesting++
} else if '}' == r {
nesting--
if 0 == nesting {
break
}
}
}
return string(buf)
}
var root rule
needRootRAngle := false
var parse func(*rule)
parse = func(node *rule) {
for {
panicIf(skipws, ErrUnexpectedEOF)
if '<' == r {
if node != &root || len(node.kid) > 0 {
panic(ErrUnexpectedLAngle)
}
panicIf(skipws, ErrUnexpectedEOF)
node.startCode = readCode()
needRootRAngle = true
continue
} else if '>' == r {
if node == &root {
if !needRootRAngle {
panic(ErrUnmatchedRAngle)
}
}
panicIf(skipws, ErrUnexpectedEOF)
node.endCode = readCode()
return
}
delim := r
panicIf(read, ErrUnexpectedEOF)
var regex []rune
for {
if r == delim && (len(regex) == 0 || regex[len(regex)-1] != '\\') {
break
}
if '\n' == r {
panic(ErrUnexpectedNewline)
}
regex = append(regex, r)
panicIf(read, ErrUnexpectedEOF)
}
if "" == string(regex) {
break
}
panicIf(skipws, ErrUnexpectedEOF)
x := new(rule)
x.id = fmt.Sprintf("%d", lineno)
node.kid = append(node.kid, x)
x.regex = make([]rune, len(regex))
copy(x.regex, regex)
if '<' == r {
panicIf(skipws, ErrUnexpectedEOF)
x.startCode = readCode()
parse(x)
} else {
x.code = readCode()
}
}
}
parse(&root)
buf = nil
for done := skipws(); !done; done = read() {
buf = append(buf, r)
}
fs := token.NewFileSet()
t, err := parser.ParseFile(fs, "", string(buf), parser.ImportsOnly)
if err != nil {
panic(err)
}
printer.Fprint(out, fs, t)
var file *token.File
fs.Iterate(func(f *token.File) bool {
file = f
return true
})
for m := file.LineCount(); m > 1; m-- {
i := 0
for '\n' != buf[i] {
i++
}
buf = buf[i+1:]
}
ExtraType := *yyExtraName
out.WriteString(`import ("bufio";"io";"strings")
type intstring struct {
i int
s string
}
type Lexer struct {
// The lexer runs in its own goroutine, and communicates via channel 'ch'.
ch chan intstring
// We record the level of nesting because the action could return, and a
// subsequent call expects to pick up where it left off. In other words,
// we're simulating a coroutine.
// TODO: Support a channel-based variant that compatible with Go's yacc.
stack []intstring
stale bool
// TODO: The following fields were added for
// https://github.com/wagerlabs/docker/blob/65694e801a7b80930961d70c69cba9f2465459be/buildfile.nex
// In general, there are times when it would be convenient to maintain
// state in the Lexer object itself. Rather than adding fields for every
// possible nex application, it should be configurable, like the 'yyextra'
// field in Flex.
l, c int // line number and character position
`)
if len(ExtraType) > 0 {
out.WriteString(`
yyextra ` + ExtraType + `
}
func NewLexer(in io.Reader, yyextra ` + ExtraType + `) *Lexer {
`)
} else {
out.WriteString(`
}
func NewLexer(in io.Reader) *Lexer {
`)
}
out.WriteString(`
type dfa struct {
acc []bool // Accepting states.
f []func(rune) int // Transitions.
startf, endf []int // Transitions at start and end of input.
nest []dfa
}
yylex := new(Lexer)
yylex.ch = make(chan intstring)`)
if len(ExtraType) > 0 {
out.WriteString(`
yylex.yyextra = yyextra
`)
}
out.WriteString(`
var scan func(in *bufio.Reader, ch chan intstring, family []dfa)
scan = func(in *bufio.Reader, ch chan intstring, family []dfa) {
// Index of DFA and length of highest-precedence match so far.
matchi, matchn := 0, -1
var buf []rune
n := 0
checkAccept := func(i int, st int) bool {
// Higher precedence match? DFAs are run in parallel, so matchn is at most len(buf), hence we may omit the length equality check.
if family[i].acc[st] && (matchn < n || matchi > i) {
matchi, matchn = i, n
return true
}
return false
}
var state [][2]int
for i := 0; i < len(family); i++ {
mark := make([]bool, len(family[i].startf))
// Every DFA starts at state 0.
st := 0
for {
state = append(state, [2]int{i, st})
mark[st] = true
// As we're at the start of input, follow all ^ transitions and append to our list of start states.
st = family[i].startf[st]
if -1 == st || mark[st] { break }
// We only check for a match after at least one transition.
checkAccept(i, st)
}
}
atEOF := false
for {
if n == len(buf) && !atEOF {
r,_,err := in.ReadRune()
switch err {
case io.EOF: atEOF = true
case nil: buf = append(buf, r)
default: panic(err)
}
}
if !atEOF {
r := buf[n]
n++
var nextState [][2]int
for _, x := range state {
x[1] = family[x[0]].f[x[1]](r)
if -1 == x[1] { continue }
nextState = append(nextState, x)
checkAccept(x[0], x[1])
}
state = nextState
} else {
dollar: // Handle $.
for _, x := range state {
mark := make([]bool, len(family[x[0]].endf))
for {
mark[x[1]] = true
x[1] = family[x[0]].endf[x[1]]
if -1 == x[1] || mark[x[1]] { break }
if checkAccept(x[0], x[1]) {
// Unlike before, we can break off the search. Now that we're at the end, there's no need to maintain the state of each DFA.
break dollar
}
}
}
state = nil
}
if state == nil {
// All DFAs stuck. Return last match if it exists, otherwise advance by one rune and restart all DFAs.
if matchn == -1 {
if len(buf) == 0 { // This can only happen at the end of input.
break
}
buf = buf[1:]
} else {
text := string(buf[:matchn])
buf = buf[matchn:]
matchn = -1
ch <- intstring{matchi, text}
if len(family[matchi].nest) > 0 {
scan(bufio.NewReader(strings.NewReader(text)), ch, family[matchi].nest)
}
if atEOF {
break
}
}
n = 0
for i := 0; i < len(family); i++ {
state = append(state, [2]int{i, 0})
}
}
}
ch <- intstring{-1, ""}
}
go scan(bufio.NewReader(in), yylex.ch, []dfa{`)
for _, kid := range root.kid {
gen(out, kid)
}
out.WriteString(`})
return yylex
}
func (yylex *Lexer) Text() string {
return yylex.stack[len(yylex.stack) - 1].s
}
func (yylex *Lexer) next(lvl int) int {
if lvl == len(yylex.stack) {
yylex.stack = append(yylex.stack, intstring{0, ""})
}
if lvl == len(yylex.stack) - 1 {
p := &yylex.stack[lvl]
*p = <-yylex.ch
yylex.stale = false
} else {
yylex.stale = true
}
return yylex.stack[lvl].i
}
func (yylex *Lexer) pop() {
yylex.stack = yylex.stack[:len(yylex.stack) - 1]
}
`)
if !*standalone {
writeLex(out, root)
out.WriteString(string(buf))
out.Flush()
return
}
m := 0
const funmac = "NN_FUN"
for m < len(buf) {
m++
if funmac[:m] != string(buf[:m]) {
out.WriteString(string(buf[:m]))
buf = buf[m:]
m = 0
} else if funmac == string(buf[:m]) {
writeNNFun(out, root)
buf = buf[m:]
m = 0
}
}
out.WriteString(string(buf))
out.Flush()
}
