1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484
|
// Copyright 2014, Joe Tsai. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE.md file.
// Package bufpipe implements a buffered pipe.
package bufpipe
import "io"
import "sync"
// There are a number of modes of operation that BufferPipe can operate in.
//
// As such, there are 4 different (and mostly orthogonal) flags that can be
// bitwise ORed together to create the mode of operation. Thus, with 4 flags,
// there are technically 16 different possible combinations (although, some of
// them are illogical). All 16 combinations are allowed, even if no sensible
// programmer should be using them.
//
// The first flag determines the buffer's structure (linear vs. ring). In linear
// mode, a writer can only write up to the internal buffer length's worth of
// data. On the other hand, in ring mode, the buffer is treated like a circular
// buffer and allow indefinite reading and writing.
//
// The second flag determines concurrent access to the pipe (mono vs. dual).
// In mono access mode, the writer has sole access to the pipe. Only after the
// Close method is called can a reader start reading data. In dual access
// mode, readers can read written data the moment it is ready.
//
// The third and fourth flag determines waiting protocol for reading and writing
// (polling vs. blocking). In blocking mode, a writer or reader will block until
// there is available buffer space or valid data to consume. In polling mode,
// read and write operations return immediately, possibly with an ErrShortWrite
// or ErrNoProgress error.
//
// Combining Ring with Mono is an illogical combination. Mono access dictates
// that no reader can drain the pipe until it is closed. However, the benefit
// of a Ring buffer is that it can circularly write data as a reader consumes
// the input. Ring with Mono is effectively Line with Mono.
//
// Combining Line with BlockI is an illogical combination. In Line mode, the
// amount that can be written is fixed and independent of how much is read.
// Thus, using BlockI in this case will cause the writer to block forever when
// the buffer is full.
//
// With all illogical combinations removed, there are only 8 logical
// combinations that programmers should use.
const (
Ring = 1 << iota // Ring buffer vs. linear buffer
Dual // Dual access IO vs. mono access IO
BlockI // Blocking input vs. polling input
BlockO // Blocking output vs. polling output
// The below flags are the inverse of the ones above. They exist to make it
// obvious what the inverse is.
Line = 0 // Inverse of Ring
Mono = 0 // Inverse of Dual
PollI = 0 // Inverse of BlockI
PollO = 0 // Inverse of BlockO
)
// The most common combination of flags are predefined with convenient aliases.
const (
LineMono = Line | Mono | PollI | BlockO
LineDual = Line | Dual | PollI | BlockO
RingPoll = Ring | Dual | PollI | PollO
RingBlock = Ring | Dual | BlockI | BlockO
)
type BufferPipe struct {
buf []byte
mode int
rdPtr int64
wrPtr int64
closed bool
err error
mutex sync.Mutex
rdCond sync.Cond
wrCond sync.Cond
}
// BufferPipe is similar in operation to io.Pipe and is intended to be the
// communication channel between producer and consumer routines. There are some
// specific use cases for BufferPipes over io.Pipe.
//
// First, in cases where a writer may need to go back a modify a portion of the
// past "written" data before allowing the reader to consume it.
// Second, for performance applications, where the cost of copying of data is
// noticeable. Thus, it would be more efficient to read/write data from/to the
// internal buffer directly.
//
// See the defined constants for more on the buffer mode of operation.
func NewBufferPipe(buf []byte, mode int) *BufferPipe {
b := new(BufferPipe)
b.buf = buf
b.mode = mode
b.rdCond.L = &b.mutex
b.wrCond.L = &b.mutex
return b
}
// The entire internal buffer. Be careful when touching the raw buffer.
// Line buffers are always guaranteed to be aligned to be front of the slice.
// Ring buffers use wrap around logic and could be physically split apart.
func (b *BufferPipe) Buffer() []byte {
return b.buf
}
// The BufferPipe mode of operation.
func (b *BufferPipe) Mode() int {
return b.mode
}
// The internal pointer values.
func (b *BufferPipe) Pointers() (rdPtr, wrPtr int64) {
return b.rdPtr, b.wrPtr
}
// The total number of bytes the buffer can store.
func (b *BufferPipe) Capacity() int {
return len(b.buf)
}
// The number of valid bytes that can be read.
func (b *BufferPipe) Length() int {
b.mutex.Lock()
defer b.mutex.Unlock()
return int(b.wrPtr - b.rdPtr)
}
func (b *BufferPipe) writeWait() int {
var rdZero int64 // Zero value
isLine := b.mode&Ring == 0
isBlock := b.mode&BlockI > 0
rdPtr := &b.rdPtr
if isLine {
rdPtr = &rdZero // Amount read has no effect on amount available
}
if isBlock {
for !b.closed && len(b.buf) == int(b.wrPtr-(*rdPtr)) {
b.wrCond.Wait()
}
}
if b.closed {
return 0 // Closed buffer is never available
}
return len(b.buf) - int(b.wrPtr-(*rdPtr))
}
// Slices of available buffer that can be written to. This does not advance the
// internal write pointer. All of the available write space is the logical
// concatenation of the two slices.
//
// In linear buffers, the first slice obtained is guaranteed to be the entire
// available writable buffer slice.
//
// In LineMono mode only, slices obtained may still be modified even after
// WriteMark has been called and before Close. This is valid because this mode
// blocks readers until the buffer has been closed.
//
// In ring buffers, the first slice obtained may not represent all of the
// available buffer space since slices always represent a contiguous pieces of
// memory. However, the first slice is guaranteed to be a non-empty slice if
// space is available.
//
// In the Block mode, this method blocks until there is available space in
// the buffer to write. Poll mode, on the contrary, will return empty slices if
// the buffer is full.
func (b *BufferPipe) WriteSlices() (bufLo, bufHi []byte, err error) {
if b == nil {
return nil, nil, nil
}
b.mutex.Lock()
defer b.mutex.Unlock()
return b.writeSlices()
}
func (b *BufferPipe) writeSlices() (bufLo, bufHi []byte, err error) {
availCnt := b.writeWait() // Block until there is available buffer
offLo := 0
if len(b.buf) > 0 { // Prevent division by zero
offLo = int(b.wrPtr) % len(b.buf)
}
offHi := offLo + availCnt
if modCnt := offHi - len(b.buf); modCnt > 0 {
offHi = len(b.buf)
bufHi = b.buf[:modCnt] // Upper half (possible for Ring)
}
bufLo = b.buf[offLo:offHi] // Bottom half (will contain all data for Line)
// Restrict the capacity to prevent users from accidentally going past end.
bufLo = bufLo[:len(bufLo):len(bufLo)]
bufHi = bufHi[:len(bufHi):len(bufHi)]
// Check error status
if len(bufLo) == 0 {
switch {
case b.err != nil:
err = b.err
case b.closed:
err = io.ErrClosedPipe
default:
err = io.ErrShortWrite
}
}
return
}
// Advances the write pointer.
//
// The amount that can be advanced must be non-negative and be less than the
// length of the slices returned by the previous WriteSlices. Calls to Write
// may not be done between these two calls. Also, another call to WriteMark is
// invalid until WriteSlices has been called again.
//
// If WriteMark is being used, only one writer routine is allowed.
func (b *BufferPipe) WriteMark(cnt int) {
if b == nil && cnt == 0 {
return
}
b.mutex.Lock()
defer b.mutex.Unlock()
b.writeMark(cnt)
}
func (b *BufferPipe) writeMark(cnt int) {
availCnt := b.writeWait()
if cnt < 0 || cnt > availCnt {
panic("invalid mark increment value")
}
b.wrPtr += int64(cnt)
b.rdCond.Signal()
}
// Write data to the buffer.
//
// In linear buffers, the length of the data slice may not exceed the capacity
// of the buffer. Otherwise, an ErrShortWrite error will be returned.
//
// In ring buffers, the length of the data slice may exceed the capacity.
//
// Under Block mode, this operation will block until all data has been written.
// If there is no consumer of the data, then this method may block forever.
func (b *BufferPipe) Write(data []byte) (cnt int, err error) {
b.mutex.Lock()
defer b.mutex.Unlock()
for cnt < len(data) {
buf, _, err := b.writeSlices()
if err != nil {
return cnt, err
}
copyCnt := copy(buf, data[cnt:])
b.writeMark(copyCnt)
cnt += copyCnt
}
return cnt, nil
}
// Continually read the contents of the reader and write them to the pipe.
func (b *BufferPipe) ReadFrom(rd io.Reader) (cnt int64, err error) {
for {
b.mutex.Lock()
buf, _, wrErr := b.writeSlices()
rdPtr, rdErr := rd.Read(buf)
b.writeMark(rdPtr)
b.mutex.Unlock()
cnt += int64(rdPtr)
switch {
case wrErr != nil:
return cnt, wrErr
case rdErr == io.EOF:
return cnt, nil
case rdErr != nil:
return cnt, rdErr
}
}
}
func (b *BufferPipe) readWait() int {
isBlock := b.mode&BlockO > 0
isMono := b.mode&Dual == 0
if isBlock {
for !b.closed && b.rdPtr == b.wrPtr {
b.rdCond.Wait()
}
for isMono && !b.closed {
b.rdCond.Wait()
}
}
if isMono && !b.closed {
return 0
}
return int(b.wrPtr - b.rdPtr)
}
// Slices of valid data that can be read. This does not advance the internal
// read pointer. All of the valid readable data is the logical concatenation of
// the two slices.
//
// In linear buffers, the first slice obtained is guaranteed to be the entire
// valid readable buffer slice.
//
// In ring buffers, the first slice obtained may not represent all of the
// valid buffer data since slices always represent a contiguous pieces of
// memory. However, the first slice is guaranteed to be a non-empty slice if
// space is available.
//
// Under the Block mode, this method blocks until there is at least some valid
// data to read. The Mono mode is special in that, none of the data is
// considered ready for reading until the writer closes the channel.
func (b *BufferPipe) ReadSlices() (bufLo, bufHi []byte, err error) {
if b == nil {
return nil, nil, nil
}
b.mutex.Lock()
defer b.mutex.Unlock()
return b.readSlices()
}
func (b *BufferPipe) readSlices() (bufLo, bufHi []byte, err error) {
validCnt := b.readWait() // Block until there is valid buffer
offLo := 0
if len(b.buf) > 0 { // Prevent division by zero
offLo = int(b.rdPtr) % len(b.buf)
}
offHi := offLo + validCnt
if modCnt := offHi - len(b.buf); modCnt > 0 {
offHi = len(b.buf)
bufHi = b.buf[:modCnt] // Upper half (possible for Ring)
}
bufLo = b.buf[offLo:offHi] // Bottom half (will contain all data for Line)
// Restrict the capacity to prevent users from accidentally going past end.
bufLo = bufLo[:len(bufLo):len(bufLo)]
bufHi = bufHi[:len(bufHi):len(bufHi)]
// Check error status
if len(bufLo) == 0 {
switch {
case b.err != nil:
err = b.err
case b.closed:
err = io.EOF
default:
err = io.ErrNoProgress
}
}
return
}
// Advances the read pointer.
//
// The amount that can be advanced must be non-negative and be less than the
// length of the slices returned by the previous ReadSlices. Calls to Read
// may not be done between these two calls. Also, another call to ReadMark is
// invalid until ReadSlices has been called again.
//
// If ReadMark is being used, only one reader routine is allowed.
func (b *BufferPipe) ReadMark(cnt int) {
if b == nil && cnt == 0 {
return
}
b.mutex.Lock()
defer b.mutex.Unlock()
b.readMark(cnt)
}
func (b *BufferPipe) readMark(cnt int) {
validCnt := b.readWait()
if cnt < 0 || cnt > validCnt {
panic("invalid mark increment value")
}
b.rdPtr += int64(cnt)
b.wrCond.Signal()
}
// Read data from the buffer.
//
// In all modes, the length of the data slice may exceed the capacity of
// the buffer. The operation will block until all data has been read or until
// the EOF is hit.
//
// Under Block mode, this method may block forever if there is no producer.
func (b *BufferPipe) Read(data []byte) (cnt int, err error) {
b.mutex.Lock()
defer b.mutex.Unlock()
for cnt < len(data) {
buf, _, err := b.readSlices()
if err != nil {
return cnt, err
}
copyCnt := copy(data[cnt:], buf)
b.readMark(copyCnt)
cnt += copyCnt
}
return cnt, nil
}
// Continually read the contents of the pipe and write them to the writer.
func (b *BufferPipe) WriteTo(wr io.Writer) (cnt int64, err error) {
for {
b.mutex.Lock()
data, _, rdErr := b.readSlices()
wrPtr, wrErr := wr.Write(data)
b.readMark(wrPtr)
b.mutex.Unlock()
cnt += int64(wrPtr)
switch {
case wrErr != nil:
return cnt, wrErr
case rdErr == io.EOF:
return cnt, nil
case rdErr != nil:
return cnt, rdErr
}
}
}
// Close the buffer down.
//
// All write operations have no effect after this, while all read operations
// will drain remaining data in the buffer. This operation is somewhat similar
// to how Go channels operate.
//
// Writers should close the buffer to indicate to readers to mark end-of-stream.
//
// Readers should only close the buffer in the event of unexpected termination.
// The mechanism allows readers to inform writers of consumer termination and
// prevents the producer from potentially being blocked forever.
func (b *BufferPipe) Close() error {
return b.CloseWithError(nil)
}
// Closes the pipe with the given error. This sets the error value for the pipe
// and returns the previous error value.
func (b *BufferPipe) CloseWithError(err error) (errPre error) {
b.mutex.Lock()
defer b.mutex.Unlock()
errPre, b.err = b.err, err
b.closed = true
b.rdCond.Broadcast()
b.wrCond.Broadcast()
return errPre
}
// Roll back the write pointer and return the number of bytes rolled back.
// If successful, this effectively makes the valid length zero. In order to
// prevent race conditions with the reader, this action is only valid in Mono
// access mode before the channel is closed.
func (b *BufferPipe) Rollback() int {
b.mutex.Lock()
defer b.mutex.Unlock()
if b.closed || b.mode&Dual > 0 {
return 0
}
cnt := b.wrPtr - b.rdPtr
b.wrPtr = b.rdPtr
return int(cnt)
}
// Makes the buffer ready for use again by opening the pipe for writing again.
// The read and write pointers will be reset to zero and errors will be cleared.
func (b *BufferPipe) Reset() {
b.mutex.Lock()
defer b.mutex.Unlock()
b.wrPtr, b.rdPtr = 0, 0
b.err, b.closed = nil, false
}
// TODO(jtsai): Allow BufferPipe to be grown. This is safe at Reset time and
// before the first call to Write. When else is this safe?
// TODO(jtsai): Double check why some methods allow the BufferPipe pointer to
// be nil.
// TODO(jtsai): Why are some methods not protected by a mutex?
|