go/build: use custom file readers to avoid I/O

When reading Go files, read through import block.
When reading non-Go files, read only leading comments.

R=nigeltao, adg, r
CC=golang-dev
https://golang.org/cl/6493068

src/pkg/go/build/build.go

@@ -512,7 +512,13 @@ Found:
 		if err != nil {
 			return p, err
 		}
-		data, err := ioutil.ReadAll(f)
+
+		var data []byte
+		if strings.HasSuffix(filename, ".go") {
+			data, err = readImports(f, false)
+		} else {
+			data, err = readComments(f)
+		}
 		f.Close()
 		if err != nil {
 			return p, fmt.Errorf("read %s: %v", filename, err)

src/pkg/go/build/read.go

@@ -0,0 +1,238 @@
+// Copyright 2012 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 build
+
+import (
+	"bufio"
+	"errors"
+	"io"
+)
+
+type importReader struct {
+	b    *bufio.Reader
+	buf  []byte
+	peek byte
+	err  error
+	eof  bool
+	nerr int
+}
+
+func isIdent(c byte) bool {
+	return 'A' <= c && c <= 'Z' || 'a' <= c && c <= 'z' || '0' <= c && c <= '9' || c == '_' || c >= 0x80
+}
+
+var (
+	errSyntax = errors.New("syntax error")
+	errNUL    = errors.New("unexpected NUL in input")
+)
+
+// syntaxError records a syntax error, but only if an I/O error has not already been recorded.
+func (r *importReader) syntaxError() {
+	if r.err == nil {
+		r.err = errSyntax
+	}
+}
+
+// readByte reads the next byte from the input, saves it in buf, and returns it.
+// If an error occurs, readByte records the error in r.err and returns 0.
+func (r *importReader) readByte() byte {
+	c, err := r.b.ReadByte()
+	if err == nil {
+		r.buf = append(r.buf, c)
+		if c == 0 {
+			err = errNUL
+		}
+	}
+	if err != nil {
+		if err == io.EOF {
+			r.eof = true
+		} else if r.err == nil {
+			r.err = err
+		}
+		c = 0
+	}
+	return c
+}
+
+// peekByte returns the next byte from the input reader but does not advance beyond it.
+// If skipSpace is set, peekByte skips leading spaces and comments.
+func (r *importReader) peekByte(skipSpace bool) byte {
+	if r.err != nil {
+		if r.nerr++; r.nerr > 10000 {
+			panic("go/build: import reader looping")
+		}
+		return 0
+	}
+
+	// Use r.peek as first input byte.
+	// Don't just return r.peek here: it might have been left by peekByte(false)
+	// and this might be peekByte(true).
+	c := r.peek
+	if c == 0 {
+		c = r.readByte()
+	}
+	for r.err == nil && !r.eof {
+		if skipSpace {
+			// For the purposes of this reader, semicolons are never necessary to
+			// understand the input and are treated as spaces.
+			switch c {
+			case ' ', '\f', '\t', '\r', '\n', ';':
+				c = r.readByte()
+				continue
+
+			case '/':
+				c = r.readByte()
+				if c == '/' {
+					for c != '\n' && r.err == nil && !r.eof {
+						c = r.readByte()
+					}
+				} else if c == '*' {
+					var c1 byte
+					for (c != '*' || c1 != '/') && r.err == nil {
+						if r.eof {
+							r.syntaxError()
+						}
+						c, c1 = c1, r.readByte()
+					}
+				} else {
+					r.syntaxError()
+				}
+				c = r.readByte()
+				continue
+			}
+		}
+		break
+	}
+	r.peek = c
+	return r.peek
+}
+
+// nextByte is like peekByte but advances beyond the returned byte.
+func (r *importReader) nextByte(skipSpace bool) byte {
+	c := r.peekByte(skipSpace)
+	r.peek = 0
+	return c
+}
+
+// readKeyword reads the given keyword from the input.
+// If the keyword is not present, readKeyword records a syntax error.
+func (r *importReader) readKeyword(kw string) {
+	r.peekByte(true)
+	for i := 0; i < len(kw); i++ {
+		if r.nextByte(false) != kw[i] {
+			r.syntaxError()
+			return
+		}
+	}
+	if isIdent(r.peekByte(false)) {
+		r.syntaxError()
+	}
+}
+
+// readIdent reads an identifier from the input.
+// If an identifier is not present, readIdent records a syntax error.
+func (r *importReader) readIdent() {
+	c := r.peekByte(true)
+	if !isIdent(c) {
+		r.syntaxError()
+		return
+	}
+	for isIdent(r.peekByte(false)) {
+		r.peek = 0
+	}
+}
+
+// readString reads a quoted string literal from the input.
+// If an identifier is not present, readString records a syntax error.
+func (r *importReader) readString() {
+	switch r.nextByte(true) {
+	case '`':
+		for r.err == nil {
+			if r.nextByte(false) == '`' {
+				break
+			}
+			if r.eof {
+				r.syntaxError()
+			}
+		}
+	case '"':
+		for r.err == nil {
+			c := r.nextByte(false)
+			if c == '"' {
+				break
+			}
+			if r.eof || c == '\n' {
+				r.syntaxError()
+			}
+			if c == '\\' {
+				r.nextByte(false)
+			}
+		}
+	default:
+		r.syntaxError()
+	}
+}
+
+// readImport reads an import clause - optional identifier followed by quoted string -
+// from the input.
+func (r *importReader) readImport() {
+	c := r.peekByte(true)
+	if c == '.' {
+		r.peek = 0
+	} else if isIdent(c) {
+		r.readIdent()
+	}
+	r.readString()
+}
+
+// readComments is like ioutil.ReadAll, except that it only reads the leading
+// block of comments in the file.
+func readComments(f io.Reader) ([]byte, error) {
+	r := &importReader{b: bufio.NewReader(f)}
+	r.peekByte(true)
+	if r.err == nil && !r.eof {
+		// Didn't reach EOF, so must have found a non-space byte. Remove it.
+		r.buf = r.buf[:len(r.buf)-1]
+	}
+	return r.buf, r.err
+}
+
+// readImports is like ioutil.ReadAll, except that it expects a Go file as input
+// and stops reading the input once the imports have completed.
+func readImports(f io.Reader, reportSyntaxError bool) ([]byte, error) {
+	r := &importReader{b: bufio.NewReader(f)}
+
+	r.readKeyword("package")
+	r.readIdent()
+	for r.peekByte(true) == 'i' {
+		r.readKeyword("import")
+		if r.peekByte(true) == '(' {
+			r.nextByte(false)
+			for r.peekByte(true) != ')' && r.err == nil {
+				r.readImport()
+			}
+			r.nextByte(false)
+		} else {
+			r.readImport()
+		}
+	}
+
+	// If we stopped successfully before EOF, we read a byte that told us we were done.
+	// Return all but that last byte, which would cause a syntax error if we let it through.
+	if r.err == nil && !r.eof {
+		return r.buf[:len(r.buf)-1], nil
+	}
+
+	// If we stopped for a syntax error, consume the whole file so that
+	// we are sure we don't change the errors that go/parser returns.
+	if r.err == errSyntax && !reportSyntaxError {
+		r.err = nil
+		for r.err == nil && !r.eof {
+			r.readByte()
+		}
+	}
+
+	return r.buf, r.err
+}

src/pkg/go/build/read_test.go

@@ -0,0 +1,226 @@
+// Copyright 2012 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 build
+
+import (
+	"io"
+	"strings"
+	"testing"
+)
+
+const quote = "`"
+
+type readTest struct {
+	// Test input contains ℙ where readImports should stop.
+	in  string
+	err string
+}
+
+var readImportsTests = []readTest{
+	{
+		`package p`,
+		"",
+	},
+	{
+		`package p; import "x"`,
+		"",
+	},
+	{
+		`package p; import . "x"`,
+		"",
+	},
+	{
+		`package p; import "x";ℙvar x = 1`,
+		"",
+	},
+	{
+		`package p
+		
+		// comment
+		
+		import "x"
+		import _ "x"
+		import a "x"
+		
+		/* comment */
+		
+		import (
+			"x" /* comment */
+			_ "x"
+			a "x" // comment
+			` + quote + `x` + quote + `
+			_ /*comment*/ ` + quote + `x` + quote + `
+			a ` + quote + `x` + quote + `
+		)
+		import (
+		)
+		import ()
+		import()import()import()
+		import();import();import()
+		
+		ℙvar x = 1
+		`,
+		"",
+	},
+}
+
+var readCommentsTests = []readTest{
+	{
+		`ℙpackage p`,
+		"",
+	},
+	{
+		`ℙpackage p; import "x"`,
+		"",
+	},
+	{
+		`ℙpackage p; import . "x"`,
+		"",
+	},
+	{
+		`// foo
+
+		/* bar */
+
+		/* quux */ // baz
+		
+		/*/ zot */
+
+		// asdf
+		ℙHello, world`,
+		"",
+	},
+}
+
+func testRead(t *testing.T, tests []readTest, read func(io.Reader) ([]byte, error)) {
+	for i, tt := range tests {
+		var in, testOut string
+		j := strings.Index(tt.in, "ℙ")
+		if j < 0 {
+			in = tt.in
+			testOut = tt.in
+		} else {
+			in = tt.in[:j] + tt.in[j+len("ℙ"):]
+			testOut = tt.in[:j]
+		}
+		r := strings.NewReader(in)
+		buf, err := read(r)
+		if err != nil {
+			if tt.err == "" {
+				t.Errorf("#%d: err=%q, expected success (%q)", i, err, string(buf))
+				continue
+			}
+			if !strings.Contains(err.Error(), tt.err) {
+				t.Errorf("#%d: err=%q, expected %q", i, err, tt.err)
+				continue
+			}
+			continue
+		}
+		if err == nil && tt.err != "" {
+			t.Errorf("#%d: success, expected %q", i, tt.err)
+			continue
+		}
+
+		out := string(buf)
+		if out != testOut {
+			t.Errorf("#%d: wrong output:\nhave %q\nwant %q\n", i, out, testOut)
+		}
+	}
+}
+
+func TestReadImports(t *testing.T) {
+	testRead(t, readImportsTests, func(r io.Reader) ([]byte, error) { return readImports(r, true) })
+}
+
+func TestReadComments(t *testing.T) {
+	testRead(t, readCommentsTests, readComments)
+}
+
+var readFailuresTests = []readTest{
+	{
+		`package`,
+		"syntax error",
+	},
+	{
+		"package p\n\x00\nimport `math`\n",
+		"unexpected NUL in input",
+	},
+	{
+		`package p; import`,
+		"syntax error",
+	},
+	{
+		`package p; import "`,
+		"syntax error",
+	},
+	{
+		"package p; import ` \n\n",
+		"syntax error",
+	},
+	{
+		`package p; import "x`,
+		"syntax error",
+	},
+	{
+		`package p; import _`,
+		"syntax error",
+	},
+	{
+		`package p; import _ "`,
+		"syntax error",
+	},
+	{
+		`package p; import _ "x`,
+		"syntax error",
+	},
+	{
+		`package p; import .`,
+		"syntax error",
+	},
+	{
+		`package p; import . "`,
+		"syntax error",
+	},
+	{
+		`package p; import . "x`,
+		"syntax error",
+	},
+	{
+		`package p; import (`,
+		"syntax error",
+	},
+	{
+		`package p; import ("`,
+		"syntax error",
+	},
+	{
+		`package p; import ("x`,
+		"syntax error",
+	},
+	{
+		`package p; import ("x"`,
+		"syntax error",
+	},
+}
+
+func TestReadFailures(t *testing.T) {
+	// Errors should be reported (true arg to readImports).
+	testRead(t, readFailuresTests, func(r io.Reader) ([]byte, error) { return readImports(r, true) })
+}
+
+func TestReadFailuresIgnored(t *testing.T) {
+	// Syntax errors should not be reported (false arg to readImports).
+	// Instead, entire file should be the output and no error.
+	// Convert tests not to return syntax errors.
+	tests := make([]readTest, len(readFailuresTests))
+	copy(tests, readFailuresTests)
+	for i := range tests {
+		tt := &tests[i]
+		if !strings.Contains(tt.err, "NUL") {
+			tt.err = ""
+		}
+	}
+	testRead(t, tests, func(r io.Reader) ([]byte, error) { return readImports(r, false) })
+}