Commit cc1a9792 authored by Russ Cox's avatar Russ Cox

package patch

R=r
http://go/go-review/1018043
parent 31053d4e
......@@ -5,8 +5,8 @@ bignum.install: fmt.install
bufio.install: io.install os.install strconv.install utf8.install
bytes.install: os.install unicode.install utf8.install
compress/flate.install: bufio.install bytes.install io.install math.install os.install sort.install strconv.install
compress/gzip.install: bufio.install compress/flate.install hash/crc32.install hash.install io.install os.install
compress/zlib.install: bufio.install compress/flate.install hash/adler32.install hash.install io.install os.install
compress/gzip.install: bufio.install compress/flate.install hash.install hash/crc32.install io.install os.install
compress/zlib.install: bufio.install compress/flate.install hash.install hash/adler32.install io.install os.install
container/heap.install: sort.install
container/list.install:
container/ring.install:
......@@ -49,7 +49,7 @@ hash/adler32.install: hash.install os.install
hash/crc32.install: hash.install os.install
http.install: bufio.install bytes.install container/vector.install fmt.install io.install log.install net.install os.install path.install strconv.install strings.install utf8.install
image.install:
image/png.install: bufio.install compress/zlib.install hash/crc32.install hash.install image.install io.install os.install strconv.install
image/png.install: bufio.install compress/zlib.install hash.install hash/crc32.install image.install io.install os.install strconv.install
io.install: bytes.install os.install sort.install strings.install sync.install
json.install: bytes.install container/vector.install fmt.install math.install reflect.install strconv.install strings.install utf8.install
log.install: fmt.install io.install os.install runtime.install time.install
......@@ -58,6 +58,7 @@ math.install:
net.install: fmt.install io.install once.install os.install reflect.install sync.install syscall.install
once.install: sync.install
os.install: once.install syscall.install
patch.install: bytes.install compress/zlib.install crypto/sha1.install encoding/git85.install fmt.install io.install os.install path.install strings.install
path.install: io.install os.install strings.install
rand.install: math.install
reflect.install: runtime.install strconv.install
......
......@@ -72,6 +72,7 @@ DIRS=\
net\
once\
os\
patch\
path\
rand\
reflect\
......
......@@ -44,7 +44,7 @@ var decode = [256]uint8{
// bytes of dst. As a convenience, it returns the number
// of bytes written to dst, but this value is always EncodedLen(len(src)).
// Encode implements the radix 85 encoding used in the
// Git version control tool.
// GIT version control tool.
//
// The encoding splits src into chunks of at most 52 bytes
// and encodes each chunk on its own line.
......@@ -146,9 +146,9 @@ func MaxDecodedLen(n int) int {
return n/5*4;
}
// NewEncoder returns a new Git base85 stream encoder. Data written to
// NewEncoder returns a new GIT base85 stream encoder. Data written to
// the returned writer will be encoded and then written to w.
// The Git encoding operates on 52-byte blocks; when finished
// The GIT encoding operates on 52-byte blocks; when finished
// writing, the caller must Close the returned encoder to flush any
// partially written blocks.
func NewEncoder(w io.Writer) io.WriteCloser {
......@@ -223,7 +223,7 @@ func (e *encoder) Close() os.Error {
return e.err;
}
// NewDecoder returns a new Git base85 stream decoder.
// NewDecoder returns a new GIT base85 stream decoder.
func NewDecoder(r io.Reader) io.Reader {
return &decoder{r: r};
}
......
# Copyright 2009 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.
include $(GOROOT)/src/Make.$(GOARCH)
TARG=patch
GOFILES=\
apply.go\
git.go\
patch.go\
textdiff.go\
include $(GOROOT)/src/Make.pkg
// Copyright 2009 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 patch
import "os"
// An Op is a single operation to execute to apply a patch.
type Op struct {
Verb Verb; // action
Src string; // source file
Dst string; // destination file
Mode int; // mode for destination (if non-zero)
Data []byte; // data for destination (if non-nil)
}
// Apply applies the patch set to the files named in the patch set,
// constructing an in-memory copy of the new file state.
// It is the client's job to write the changes to the file system
// if desired.
//
// The function readFile should return the contents of the named file.
// Typically this function will be io.ReadFile.
//
func (set *Set) Apply(readFile func(string) ([]byte, os.Error)) ([]Op, os.Error) {
op := make([]Op, len(set.File));
for i, f := range set.File {
o := &op[i];
o.Verb = f.Verb;
o.Src = f.Src;
o.Dst = f.Dst;
o.Mode = f.NewMode;
if f.Diff != NoDiff || o.Verb != Edit {
// Clients assume o.Data == nil means no data diff.
// Start with a non-nil data.
var old []byte = make([]byte, 0); // not nil
var err os.Error;
if f.Src != "" {
old, err = readFile(f.Src);
if err != nil {
return nil, &os.PathError{string(f.Verb), f.Src, err};
}
}
o.Data, err = f.Diff.Apply(old);
if err != nil {
return nil, &os.PathError{string(f.Verb), f.Src, err};
}
}
}
return op, nil;
}
// Copyright 2009 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 patch
import (
"bytes";
"compress/zlib";
"crypto/sha1";
"encoding/git85";
"fmt";
"io";
"os";
)
func gitSHA1(data []byte) []byte {
if len(data) == 0 {
// special case: 0 length is all zeros sum
return make([]byte, 20);
}
h := sha1.New();
fmt.Fprintf(h, "blob %d\x00", len(data));
h.Write(data);
return h.Sum();
}
// BUG(rsc): The GIT binary delta format is not implemented, only GIT binary literals.
// GITBinaryLiteral represents a GIT binary literal diff.
type GITBinaryLiteral struct {
OldSHA1 []byte; // if non-empty, the SHA1 hash of the original
New []byte; // the new contents
}
// Apply implements the Diff interface's Apply method.
func (d *GITBinaryLiteral) Apply(old []byte) ([]byte, os.Error) {
if sum := gitSHA1(old); !bytes.HasPrefix(sum, d.OldSHA1) {
return nil, ErrPatchFailure;
}
return d.New, nil;
}
func unhex(c byte) uint8 {
switch {
case '0' <= c && c <= '9':
return c-'0';
case 'a' <= c && c <= 'f':
return c-'a'+10;
case 'A' <= c && c <= 'F':
return c-'A'+10;
}
return 255;
}
func getHex(s []byte) (data []byte, rest []byte) {
n := 0;
for n < len(s) && unhex(s[n]) != 255 {
n++;
}
n &^= 1; // Only take an even number of hex digits.
data = make([]byte, n/2);
for i := range data {
data[i] = unhex(s[2*i])<<4 | unhex(s[2*i + 1]);
}
rest = s[n:len(s)];
return;
}
// ParseGITBinary parses raw as a GIT binary patch.
func ParseGITBinary(raw []byte) (Diff, os.Error) {
var oldSHA1, newSHA1 []byte;
var sawBinary bool;
for {
var first []byte;
first, raw, _ = getLine(raw, 1);
first = bytes.TrimSpace(first);
if s, ok := skip(first, "index "); ok {
oldSHA1, s = getHex(s);
if s, ok = skip(s, ".."); !ok {
continue;
}
newSHA1, s = getHex(s);
continue;
}
if _, ok := skip(first, "GIT binary patch"); ok {
sawBinary = true;
continue;
}
if n, _, ok := atoi(first, "literal ", 10); ok && sawBinary {
data := make([]byte, n);
d := git85.NewDecoder(bytes.NewBuffer(raw));
z, err := zlib.NewInflater(d);
if err != nil {
return nil, err;
}
defer z.Close();
if _, err = io.ReadFull(z, data); err != nil {
if err == os.EOF {
err = io.ErrUnexpectedEOF;
}
return nil, err;
}
var buf [1]byte;
m, err := z.Read(&buf);
if m != 0 || err != os.EOF {
return nil, os.NewError("GIT binary literal longer than expected");
}
if sum := gitSHA1(data); !bytes.HasPrefix(sum, newSHA1) {
return nil, os.NewError("GIT binary literal SHA1 mismatch");
}
return &GITBinaryLiteral{oldSHA1, data}, nil;
}
if !sawBinary {
return nil, os.NewError("unexpected GIT patch header: " + string(first));
}
}
panic("unreachable");
}
// Copyright 2009 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 patch implements parsing and execution of the textual and
// binary patch descriptions used by version control tools such as
// CVS, GIT, Mercurial, and Subversion.
package patch
import (
"bytes";
"os";
"path";
"strings";
)
// A Set represents a set of patches to be applied as a single atomic unit.
// Patch sets are often preceded by a descriptive header.
type Set struct {
Header string; // free-form text
File []*File;
}
// A File represents a collection of changes to be made to a single file.
type File struct {
Verb Verb;
Src string; // source for Verb == Copy, Verb == Rename
Dst string;
OldMode, NewMode int; // 0 indicates not used
Diff; // changes to data; == NoDiff if operation does not edit file
}
// A Verb is an action performed on a file.
type Verb string
const (
Add Verb = "add";
Copy Verb = "copy";
Delete Verb = "delete";
Edit Verb = "edit";
Rename Verb = "rename";
)
// A Diff is any object that describes changes to transform
// an old byte stream to a new one.
type Diff interface {
// Apply applies the changes listed in the diff
// to the string s, returning the new version of the string.
// Note that the string s need not be a text string.
Apply(old []byte) (new []byte, err os.Error);
}
// NoDiff is a no-op Diff implementation: it passes the
// old data through unchanged.
var NoDiff Diff = noDiffType(0)
type noDiffType int
func (noDiffType) Apply(old []byte) ([]byte, os.Error) {
return old, nil;
}
// A SyntaxError represents a syntax error encountered while parsing a patch.
type SyntaxError string
func (e SyntaxError) String() string {
return string(e);
}
var newline = []byte{'\n'}
// Parse patches the patch text to create a patch Set.
// The patch text typically comprises a textual header and a sequence
// of file patches, as would be generated by CVS, Subversion,
// Mercurial, or GIT.
func Parse(text []byte) (*Set, os.Error) {
// Split text into files.
// CVS and Subversion begin new files with
// Index: file name.
// ==================
// diff -u blah blah
//
// Mercurial and GIT use
// diff [--git] a/file/path b/file/path.
//
// First look for Index: lines. If none, fall back on diff lines.
text, files := sections(text, "Index: ");
if len(files) == 0 {
text, files = sections(text, "diff ");
}
set := &Set{string(text), make([]*File, len(files))};
// Parse file header and then
// parse files into patch chunks.
// Each chunk begins with @@.
for i, raw := range files {
p := new(File);
set.File[i] = p;
// First line of hdr is the Index: that
// begins the section. After that is the file name.
s, raw, _ := getLine(raw, 1);
if hasPrefix(s, "Index: ") {
p.Dst = string(bytes.TrimSpace(s[7:len(s)]));
goto HaveName;
} else if hasPrefix(s, "diff ") {
str := string(bytes.TrimSpace(s));
i := strings.LastIndex(str, " b/");
if i >= 0 {
p.Dst = str[i+3 : len(str)];
goto HaveName;
}
}
return nil, SyntaxError("unexpected patch header line: " + string(s));
HaveName:
p.Dst = path.Clean(p.Dst);
if strings.HasPrefix(p.Dst, "../") || strings.HasPrefix(p.Dst, "/") {
return nil, SyntaxError("invalid path: " + p.Dst);
}
// Parse header lines giving file information:
// new file mode %o - file created
// deleted file mode %o - file deleted
// old file mode %o - file mode changed
// new file mode %o - file mode changed
// rename from %s - file renamed from other file
// rename to %s
// copy from %s - file copied from other file
// copy to %s
p.Verb = Edit;
for len(raw) > 0 {
oldraw := raw;
var l []byte;
l, raw, _ = getLine(raw, 1);
l = bytes.TrimSpace(l);
if m, s, ok := atoi(l, "new file mode ", 8); ok && len(s) == 0 {
p.NewMode = m;
p.Verb = Add;
continue;
}
if m, s, ok := atoi(l, "deleted file mode ", 8); ok && len(s) == 0 {
p.OldMode = m;
p.Verb = Delete;
p.Src = p.Dst;
p.Dst = "";
continue;
}
if m, s, ok := atoi(l, "old file mode ", 8); ok && len(s) == 0 {
// usually implies p.Verb = "rename" or "copy"
// but we'll get that from the rename or copy line.
p.OldMode = m;
continue;
}
if m, s, ok := atoi(l, "old mode ", 8); ok && len(s) == 0 {
p.OldMode = m;
continue;
}
if m, s, ok := atoi(l, "new mode ", 8); ok && len(s) == 0 {
p.NewMode = m;
continue;
}
if s, ok := skip(l, "rename from "); ok && len(s) > 0 {
p.Src = string(s);
p.Verb = Rename;
continue;
}
if s, ok := skip(l, "rename to "); ok && len(s) > 0 {
p.Verb = Rename;
continue;
}
if s, ok := skip(l, "copy from "); ok && len(s) > 0 {
p.Src = string(s);
p.Verb = Copy;
continue;
}
if s, ok := skip(l, "copy to "); ok && len(s) > 0 {
p.Verb = Copy;
continue;
}
if s, ok := skip(l, "Binary file "); ok && len(s) > 0 {
// Hg prints
// Binary file foo has changed
// when deleting a binary file.
continue;
}
if s, ok := skip(l, "RCS file: "); ok && len(s) > 0 {
// CVS prints
// RCS file: /cvs/plan9/bin/yesterday,v
// retrieving revision 1.1
// for each file.
continue;
}
if s, ok := skip(l, "retrieving revision "); ok && len(s) > 0 {
// CVS prints
// RCS file: /cvs/plan9/bin/yesterday,v
// retrieving revision 1.1
// for each file.
continue;
}
if hasPrefix(l, "===") || hasPrefix(l, "---") || hasPrefix(l, "+++") || hasPrefix(l, "diff ") {
continue;
}
if hasPrefix(l, "@@ -") {
diff, err := ParseTextDiff(oldraw);
if err != nil {
return nil, err;
}
p.Diff = diff;
break;
}
if hasPrefix(l, "index ") || hasPrefix(l, "GIT binary patch") {
diff, err := ParseGITBinary(oldraw);
if err != nil {
return nil, err;
}
p.Diff = diff;
break;
}
return nil, SyntaxError("unexpected patch header line: " + string(l));
}
if p.Diff == nil {
p.Diff = NoDiff;
}
if p.Verb == Edit {
p.Src = p.Dst;
}
}
return set, nil;
}
// getLine returns the first n lines of data and the remainder.
// If data has no newline, getLine returns data, nil, false
func getLine(data []byte, n int) (first []byte, rest []byte, ok bool) {
rest = data;
ok = true;
for ; n > 0; n-- {
nl := bytes.Index(rest, newline);
if nl < 0 {
rest = nil;
ok = false;
break;
}
rest = rest[nl+1 : len(rest)];
}
first = data[0 : len(data)-len(rest)];
return;
}
// sections returns a collection of file sections,
// each of which begins with a line satisfying prefix.
// text before the first instance of such a line is
// returned separately.
func sections(text []byte, prefix string) ([]byte, [][]byte) {
n := 0;
for b := text; ; {
if hasPrefix(b, prefix) {
n++;
}
nl := bytes.Index(b, newline);
if nl < 0 {
break;
}
b = b[nl+1 : len(b)];
}
sect := make([][]byte, n+1);
n = 0;
for b := text; ; {
if hasPrefix(b, prefix) {
sect[n] = text[0 : len(text)-len(b)];
n++;
text = b;
}
nl := bytes.Index(b, newline);
if nl < 0 {
sect[n] = text;
break;
}
b = b[nl+1 : len(b)];
}
return sect[0], sect[1:len(sect)];
}
// if s begins with the prefix t, skip returns
// s with that prefix removed and ok == true.
func skip(s []byte, t string) (ss []byte, ok bool) {
if len(s) < len(t) || string(s[0:len(t)]) != t {
return nil, false;
}
return s[len(t):len(s)], true;
}
// if s begins with the prefix t and then is a sequence
// of digits in the given base, atoi returns the number
// represented by the digits and s with the
// prefix and the digits removed.
func atoi(s []byte, t string, base int) (n int, ss []byte, ok bool) {
if s, ok = skip(s, t); !ok {
return;
}
var i int;
for i = 0; i < len(s) && '0' <= s[i] && s[i] <= byte('0'+base-1); i++ {
n = n*base + int(s[i]-'0');
}
if i == 0 {
return;
}
return n, s[i:len(s)], true;
}
// hasPrefix returns true if s begins with t.
func hasPrefix(s []byte, t string) bool {
_, ok := skip(s, t);
return ok;
}
// splitLines returns the result of splitting s into lines.
// The \n on each line is preserved.
func splitLines(s []byte) [][]byte {
return bytes.SplitAfter(s, newline, 0);
}
This diff is collapsed.
package patch
import (
"bytes";
"os";
)
type TextDiff []TextChunk
// A TextChunk specifies an edit to a section of a file:
// the text beginning at Line, which should be exactly Old,
// is to be replaced with New.
type TextChunk struct {
Line int;
Old []byte;
New []byte;
}
func ParseTextDiff(raw []byte) (TextDiff, os.Error) {
// Copy raw so it is safe to keep references to slices.
_, chunks := sections(raw, "@@ -");
delta := 0;
diff := make(TextDiff, len(chunks));
for i, raw := range chunks {
c := &diff[i];
// Parse start line: @@ -oldLine,oldCount +newLine,newCount @@ junk
chunk := splitLines(raw);
chunkHeader := chunk[0];
var ok bool;
var oldLine, oldCount, newLine, newCount int;
s := chunkHeader;
if oldLine, s, ok = atoi(s, "@@ -", 10); !ok {
ErrChunkHdr:
return nil, SyntaxError("unexpected chunk header line: " + string(chunkHeader));
}
if len(s) == 0 || s[0] != ',' {
oldCount = 1;
} else if oldCount, s, ok = atoi(s, ",", 10); !ok {
goto ErrChunkHdr;
}
if newLine, s, ok = atoi(s, " +", 10); !ok {
goto ErrChunkHdr;
}
if len(s) == 0 || s[0] != ',' {
newCount = 1;
} else if newCount, s, ok = atoi(s, ",", 10); !ok {
goto ErrChunkHdr;
}
if !hasPrefix(s, " @@") {
goto ErrChunkHdr;
}
// Special case: for created or deleted files, the empty half
// is given as starting at line 0. Translate to line 1.
if oldCount == 0 && oldLine == 0 {
oldLine = 1;
}
if newCount == 0 && newLine == 0 {
newLine = 1;
}
// Count lines in text
var dropOldNL, dropNewNL bool;
var nold, nnew int;
var lastch byte;
chunk = chunk[1:len(chunk)];
for _, l := range chunk {
if nold == oldCount && nnew == newCount && (len(l) == 0 || l[0] != '\\') {
if len(bytes.TrimSpace(l)) != 0 {
return nil, SyntaxError("too many chunk lines");
}
continue;
}
if len(l) == 0 {
return nil, SyntaxError("empty chunk line");
}
switch l[0] {
case '+':
nnew++;
case '-':
nold++;
case ' ':
nnew++;
nold++;
case '\\':
if _, ok := skip(l, "\\ No newline at end of file"); ok {
switch lastch {
case '-':
dropOldNL = true;
case '+':
dropNewNL = true;
case ' ':
dropOldNL = true;
dropNewNL = true;
default:
return nil, SyntaxError("message `\\ No newline at end of file' out of context");
}
break;
}
fallthrough;
default:
return nil, SyntaxError("unexpected chunk line: " + string(l));
}
lastch = l[0];
}
// Does it match the header?
if nold != oldCount || nnew != newCount {
return nil, SyntaxError("chunk header does not match line count: " + string(chunkHeader));
}
if oldLine+delta != newLine {
return nil, SyntaxError("chunk delta is out of sync with previous chunks");
}
delta += nnew-nold;
c.Line = oldLine;
var old, new bytes.Buffer;
nold = 0;
nnew = 0;
for _, l := range chunk {
if nold == oldCount && nnew == newCount {
break;
}
ch, l := l[0], l[1:len(l)];
if ch == '\\' {
continue;
}
if ch != '+' {
old.Write(l);
nold++;
}
if ch != '-' {
new.Write(l);
nnew++;
}
}
c.Old = old.Bytes();
c.New = new.Bytes();
if dropOldNL {
c.Old = c.Old[0 : len(c.Old)-1];
}
if dropNewNL {
c.New = c.New[0 : len(c.New)-1];
}
}
return diff, nil;
}
var ErrPatchFailure = os.NewError("patch did not apply cleanly")
// Apply applies the changes listed in the diff
// to the data, returning the new version.
func (d TextDiff) Apply(data []byte) ([]byte, os.Error) {
var buf bytes.Buffer;
line := 1;
for _, c := range d {
var ok bool;
var prefix []byte;
prefix, data, ok = getLine(data, c.Line - line);
if !ok || !bytes.HasPrefix(data, c.Old) {
return nil, ErrPatchFailure;
}
buf.Write(prefix);
data = data[len(c.Old):len(data)];
buf.Write(c.New);
line = c.Line + bytes.Count(c.Old, newline);
}
buf.Write(data);
return buf.Bytes(), nil;
}
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment