| Abhay Kumar | a2ae599 | 2025-11-10 14:02:24 +0000 | [diff] [blame^] | 1 | // Copyright 2010 The Go Authors. All rights reserved. |
| 2 | // Use of this source code is governed by a BSD-style |
| 3 | // license that can be found in the LICENSE file. |
| 4 | |
| 5 | package gzip |
| 6 | |
| 7 | import ( |
| 8 | "errors" |
| 9 | "fmt" |
| 10 | "hash/crc32" |
| 11 | "io" |
| 12 | |
| 13 | "github.com/klauspost/compress/flate" |
| 14 | ) |
| 15 | |
| 16 | // These constants are copied from the flate package, so that code that imports |
| 17 | // "compress/gzip" does not also have to import "compress/flate". |
| 18 | const ( |
| 19 | NoCompression = flate.NoCompression |
| 20 | BestSpeed = flate.BestSpeed |
| 21 | BestCompression = flate.BestCompression |
| 22 | DefaultCompression = flate.DefaultCompression |
| 23 | ConstantCompression = flate.ConstantCompression |
| 24 | HuffmanOnly = flate.HuffmanOnly |
| 25 | |
| 26 | // StatelessCompression will do compression but without maintaining any state |
| 27 | // between Write calls. |
| 28 | // There will be no memory kept between Write calls, |
| 29 | // but compression and speed will be suboptimal. |
| 30 | // Because of this, the size of actual Write calls will affect output size. |
| 31 | StatelessCompression = -3 |
| 32 | ) |
| 33 | |
| 34 | // A Writer is an io.WriteCloser. |
| 35 | // Writes to a Writer are compressed and written to w. |
| 36 | type Writer struct { |
| 37 | Header // written at first call to Write, Flush, or Close |
| 38 | w io.Writer |
| 39 | level int |
| 40 | err error |
| 41 | compressor *flate.Writer |
| 42 | digest uint32 // CRC-32, IEEE polynomial (section 8) |
| 43 | size uint32 // Uncompressed size (section 2.3.1) |
| 44 | wroteHeader bool |
| 45 | closed bool |
| 46 | buf [10]byte |
| 47 | } |
| 48 | |
| 49 | // NewWriter returns a new Writer. |
| 50 | // Writes to the returned writer are compressed and written to w. |
| 51 | // |
| 52 | // It is the caller's responsibility to call Close on the WriteCloser when done. |
| 53 | // Writes may be buffered and not flushed until Close. |
| 54 | // |
| 55 | // Callers that wish to set the fields in Writer.Header must do so before |
| 56 | // the first call to Write, Flush, or Close. |
| 57 | func NewWriter(w io.Writer) *Writer { |
| 58 | z, _ := NewWriterLevel(w, DefaultCompression) |
| 59 | return z |
| 60 | } |
| 61 | |
| 62 | // NewWriterLevel is like NewWriter but specifies the compression level instead |
| 63 | // of assuming DefaultCompression. |
| 64 | // |
| 65 | // The compression level can be DefaultCompression, NoCompression, or any |
| 66 | // integer value between BestSpeed and BestCompression inclusive. The error |
| 67 | // returned will be nil if the level is valid. |
| 68 | func NewWriterLevel(w io.Writer, level int) (*Writer, error) { |
| 69 | if level < StatelessCompression || level > BestCompression { |
| 70 | return nil, fmt.Errorf("gzip: invalid compression level: %d", level) |
| 71 | } |
| 72 | z := new(Writer) |
| 73 | z.init(w, level) |
| 74 | return z, nil |
| 75 | } |
| 76 | |
| 77 | // MinCustomWindowSize is the minimum window size that can be sent to NewWriterWindow. |
| 78 | const MinCustomWindowSize = flate.MinCustomWindowSize |
| 79 | |
| 80 | // MaxCustomWindowSize is the maximum custom window that can be sent to NewWriterWindow. |
| 81 | const MaxCustomWindowSize = flate.MaxCustomWindowSize |
| 82 | |
| 83 | // NewWriterWindow returns a new Writer compressing data with a custom window size. |
| 84 | // windowSize must be from MinCustomWindowSize to MaxCustomWindowSize. |
| 85 | func NewWriterWindow(w io.Writer, windowSize int) (*Writer, error) { |
| 86 | if windowSize < MinCustomWindowSize { |
| 87 | return nil, errors.New("gzip: requested window size less than MinWindowSize") |
| 88 | } |
| 89 | if windowSize > MaxCustomWindowSize { |
| 90 | return nil, errors.New("gzip: requested window size bigger than MaxCustomWindowSize") |
| 91 | } |
| 92 | |
| 93 | z := new(Writer) |
| 94 | z.init(w, -windowSize) |
| 95 | return z, nil |
| 96 | } |
| 97 | |
| 98 | func (z *Writer) init(w io.Writer, level int) { |
| 99 | compressor := z.compressor |
| 100 | if level != StatelessCompression { |
| 101 | if compressor != nil { |
| 102 | compressor.Reset(w) |
| 103 | } |
| 104 | } |
| 105 | |
| 106 | *z = Writer{ |
| 107 | Header: Header{ |
| 108 | OS: 255, // unknown |
| 109 | }, |
| 110 | w: w, |
| 111 | level: level, |
| 112 | compressor: compressor, |
| 113 | } |
| 114 | } |
| 115 | |
| 116 | // Reset discards the Writer z's state and makes it equivalent to the |
| 117 | // result of its original state from NewWriter or NewWriterLevel, but |
| 118 | // writing to w instead. This permits reusing a Writer rather than |
| 119 | // allocating a new one. |
| 120 | func (z *Writer) Reset(w io.Writer) { |
| 121 | z.init(w, z.level) |
| 122 | } |
| 123 | |
| 124 | // writeBytes writes a length-prefixed byte slice to z.w. |
| 125 | func (z *Writer) writeBytes(b []byte) error { |
| 126 | if len(b) > 0xffff { |
| 127 | return errors.New("gzip.Write: Extra data is too large") |
| 128 | } |
| 129 | le.PutUint16(z.buf[:2], uint16(len(b))) |
| 130 | _, err := z.w.Write(z.buf[:2]) |
| 131 | if err != nil { |
| 132 | return err |
| 133 | } |
| 134 | _, err = z.w.Write(b) |
| 135 | return err |
| 136 | } |
| 137 | |
| 138 | // writeString writes a UTF-8 string s in GZIP's format to z.w. |
| 139 | // GZIP (RFC 1952) specifies that strings are NUL-terminated ISO 8859-1 (Latin-1). |
| 140 | func (z *Writer) writeString(s string) (err error) { |
| 141 | // GZIP stores Latin-1 strings; error if non-Latin-1; convert if non-ASCII. |
| 142 | needconv := false |
| 143 | for _, v := range s { |
| 144 | if v == 0 || v > 0xff { |
| 145 | return errors.New("gzip.Write: non-Latin-1 header string") |
| 146 | } |
| 147 | if v > 0x7f { |
| 148 | needconv = true |
| 149 | } |
| 150 | } |
| 151 | if needconv { |
| 152 | b := make([]byte, 0, len(s)) |
| 153 | for _, v := range s { |
| 154 | b = append(b, byte(v)) |
| 155 | } |
| 156 | _, err = z.w.Write(b) |
| 157 | } else { |
| 158 | _, err = io.WriteString(z.w, s) |
| 159 | } |
| 160 | if err != nil { |
| 161 | return err |
| 162 | } |
| 163 | // GZIP strings are NUL-terminated. |
| 164 | z.buf[0] = 0 |
| 165 | _, err = z.w.Write(z.buf[:1]) |
| 166 | return err |
| 167 | } |
| 168 | |
| 169 | // Write writes a compressed form of p to the underlying io.Writer. The |
| 170 | // compressed bytes are not necessarily flushed until the Writer is closed. |
| 171 | func (z *Writer) Write(p []byte) (int, error) { |
| 172 | if z.err != nil { |
| 173 | return 0, z.err |
| 174 | } |
| 175 | var n int |
| 176 | // Write the GZIP header lazily. |
| 177 | if !z.wroteHeader { |
| 178 | z.wroteHeader = true |
| 179 | z.buf[0] = gzipID1 |
| 180 | z.buf[1] = gzipID2 |
| 181 | z.buf[2] = gzipDeflate |
| 182 | z.buf[3] = 0 |
| 183 | if z.Extra != nil { |
| 184 | z.buf[3] |= 0x04 |
| 185 | } |
| 186 | if z.Name != "" { |
| 187 | z.buf[3] |= 0x08 |
| 188 | } |
| 189 | if z.Comment != "" { |
| 190 | z.buf[3] |= 0x10 |
| 191 | } |
| 192 | le.PutUint32(z.buf[4:8], uint32(z.ModTime.Unix())) |
| 193 | if z.level == BestCompression { |
| 194 | z.buf[8] = 2 |
| 195 | } else if z.level == BestSpeed { |
| 196 | z.buf[8] = 4 |
| 197 | } else { |
| 198 | z.buf[8] = 0 |
| 199 | } |
| 200 | z.buf[9] = z.OS |
| 201 | n, z.err = z.w.Write(z.buf[:10]) |
| 202 | if z.err != nil { |
| 203 | return n, z.err |
| 204 | } |
| 205 | if z.Extra != nil { |
| 206 | z.err = z.writeBytes(z.Extra) |
| 207 | if z.err != nil { |
| 208 | return n, z.err |
| 209 | } |
| 210 | } |
| 211 | if z.Name != "" { |
| 212 | z.err = z.writeString(z.Name) |
| 213 | if z.err != nil { |
| 214 | return n, z.err |
| 215 | } |
| 216 | } |
| 217 | if z.Comment != "" { |
| 218 | z.err = z.writeString(z.Comment) |
| 219 | if z.err != nil { |
| 220 | return n, z.err |
| 221 | } |
| 222 | } |
| 223 | |
| 224 | if z.compressor == nil && z.level != StatelessCompression { |
| 225 | z.compressor, _ = flate.NewWriter(z.w, z.level) |
| 226 | } |
| 227 | } |
| 228 | z.size += uint32(len(p)) |
| 229 | z.digest = crc32.Update(z.digest, crc32.IEEETable, p) |
| 230 | if z.level == StatelessCompression { |
| 231 | return len(p), flate.StatelessDeflate(z.w, p, false, nil) |
| 232 | } |
| 233 | n, z.err = z.compressor.Write(p) |
| 234 | return n, z.err |
| 235 | } |
| 236 | |
| 237 | // Flush flushes any pending compressed data to the underlying writer. |
| 238 | // |
| 239 | // It is useful mainly in compressed network protocols, to ensure that |
| 240 | // a remote reader has enough data to reconstruct a packet. Flush does |
| 241 | // not return until the data has been written. If the underlying |
| 242 | // writer returns an error, Flush returns that error. |
| 243 | // |
| 244 | // In the terminology of the zlib library, Flush is equivalent to Z_SYNC_FLUSH. |
| 245 | func (z *Writer) Flush() error { |
| 246 | if z.err != nil { |
| 247 | return z.err |
| 248 | } |
| 249 | if z.closed || z.level == StatelessCompression { |
| 250 | return nil |
| 251 | } |
| 252 | if !z.wroteHeader { |
| 253 | z.Write(nil) |
| 254 | if z.err != nil { |
| 255 | return z.err |
| 256 | } |
| 257 | } |
| 258 | z.err = z.compressor.Flush() |
| 259 | return z.err |
| 260 | } |
| 261 | |
| 262 | // Close closes the Writer, flushing any unwritten data to the underlying |
| 263 | // io.Writer, but does not close the underlying io.Writer. |
| 264 | func (z *Writer) Close() error { |
| 265 | if z.err != nil { |
| 266 | return z.err |
| 267 | } |
| 268 | if z.closed { |
| 269 | return nil |
| 270 | } |
| 271 | z.closed = true |
| 272 | if !z.wroteHeader { |
| 273 | z.Write(nil) |
| 274 | if z.err != nil { |
| 275 | return z.err |
| 276 | } |
| 277 | } |
| 278 | if z.level == StatelessCompression { |
| 279 | z.err = flate.StatelessDeflate(z.w, nil, true, nil) |
| 280 | } else { |
| 281 | z.err = z.compressor.Close() |
| 282 | } |
| 283 | if z.err != nil { |
| 284 | return z.err |
| 285 | } |
| 286 | le.PutUint32(z.buf[:4], z.digest) |
| 287 | le.PutUint32(z.buf[4:8], z.size) |
| 288 | _, z.err = z.w.Write(z.buf[:8]) |
| 289 | return z.err |
| 290 | } |