| 1 |
// Copyright ©2012 The bíogo 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 bgzf |
| 6 |
|
| 7 |
import ( |
| 8 |
"bytes" |
| 9 |
"compress/gzip" |
| 10 |
"io" |
| 11 |
) |
| 12 |
|
| 13 |
// Cache is a Block caching type. Basic cache implementations are provided |
| 14 |
// in the cache package. A Cache must be safe for concurrent use. |
| 15 |
// |
| 16 |
// If a Cache is a Wrapper, its Wrap method is called on newly created blocks. |
| 17 |
type Cache interface { |
| 18 |
// Get returns the Block in the Cache with the specified |
| 19 |
// base or a nil Block if it does not exist. The returned |
| 20 |
// Block must be removed from the Cache. |
| 21 |
Get(base int64) Block |
| 22 |
|
| 23 |
// Put inserts a Block into the Cache, returning the Block |
| 24 |
// that was evicted or nil if no eviction was necessary and |
| 25 |
// a boolean indicating whether the put Block was retained |
| 26 |
// by the Cache. |
| 27 |
Put(Block) (evicted Block, retained bool) |
| 28 |
|
| 29 |
// Peek returns whether a Block exists in the cache for the |
| 30 |
// given base. If a Block satisfies the request, then exists |
| 31 |
// is returned as true with the offset for the next Block in |
| 32 |
// the stream, otherwise false and -1. |
| 33 |
Peek(base int64) (exists bool, next int64) |
| 34 |
} |
| 35 |
|
| 36 |
// Wrapper defines Cache types that need to modify a Block at its creation. |
| 37 |
type Wrapper interface { |
| 38 |
Wrap(Block) Block |
| 39 |
} |
| 40 |
|
| 41 |
// Block wraps interaction with decompressed BGZF data blocks. |
| 42 |
type Block interface { |
| 43 |
// Base returns the file offset of the start of |
| 44 |
// the gzip member from which the Block data was |
| 45 |
// decompressed. |
| 46 |
Base() int64 |
| 47 |
|
| 48 |
io.Reader |
| 49 |
io.ByteReader |
| 50 |
|
| 51 |
// Used returns whether one or more bytes have |
| 52 |
// been read from the Block. |
| 53 |
Used() bool |
| 54 |
|
| 55 |
// header returns the gzip.Header of the gzip member |
| 56 |
// from which the Block data was decompressed. |
| 57 |
header() gzip.Header |
| 58 |
|
| 59 |
// isMagicBlock returns whether the Block is a BGZF |
| 60 |
// magic EOF marker block. |
| 61 |
isMagicBlock() bool |
| 62 |
|
| 63 |
// ownedBy returns whether the Block is owned by |
| 64 |
// the given Reader. |
| 65 |
ownedBy(*Reader) bool |
| 66 |
|
| 67 |
// setOwner changes the owner to the given Reader, |
| 68 |
// reseting other data to its zero state. |
| 69 |
setOwner(*Reader) |
| 70 |
|
| 71 |
// hasData returns whether the Block has read data. |
| 72 |
hasData() bool |
| 73 |
|
| 74 |
// The following are unexported equivalents |
| 75 |
// of the io interfaces. seek is limited to |
| 76 |
// the file origin offset case and does not |
| 77 |
// return the new offset. |
| 78 |
seek(offset int64) error |
| 79 |
readFrom(io.ReadCloser) error |
| 80 |
|
| 81 |
// len returns the number of remaining |
| 82 |
// bytes that can be read from the Block. |
| 83 |
len() int |
| 84 |
|
| 85 |
// setBase sets the file offset of the start |
| 86 |
// and of the gzip member that the Block data |
| 87 |
// was decompressed from. |
| 88 |
setBase(int64) |
| 89 |
|
| 90 |
// NextBase returns the expected position of the next |
| 91 |
// BGZF block. It returns -1 if the Block is not valid. |
| 92 |
NextBase() int64 |
| 93 |
|
| 94 |
// setHeader sets the file header of of the gzip |
| 95 |
// member that the Block data was decompressed from. |
| 96 |
setHeader(gzip.Header) |
| 97 |
|
| 98 |
// txOffset returns the current vitual offset. |
| 99 |
txOffset() Offset |
| 100 |
} |
| 101 |
|
| 102 |
type block struct { |
| 103 |
owner *Reader |
| 104 |
used bool |
| 105 |
|
| 106 |
base int64 |
| 107 |
h gzip.Header |
| 108 |
magic bool |
| 109 |
|
| 110 |
offset Offset |
| 111 |
|
| 112 |
buf *bytes.Reader |
| 113 |
data [MaxBlockSize]byte |
| 114 |
} |
| 115 |
|
| 116 |
func (b *block) Base() int64 { return b.base } |
| 117 |
|
| 118 |
func (b *block) Used() bool { return b.used } |
| 119 |
|
| 120 |
func (b *block) Read(p []byte) (int, error) { |
| 121 |
n, err := b.buf.Read(p) |
| 122 |
b.offset.Block += uint16(n) |
| 123 |
if n > 0 { |
| 124 |
b.used = true |
| 125 |
} |
| 126 |
return n, err |
| 127 |
} |
| 128 |
|
| 129 |
func (b *block) ReadByte() (byte, error) { |
| 130 |
c, err := b.buf.ReadByte() |
| 131 |
if err == nil { |
| 132 |
b.offset.Block++ |
| 133 |
b.used = true |
| 134 |
} |
| 135 |
return c, err |
| 136 |
} |
| 137 |
|
| 138 |
// readToEOF will exhaust r or fill buf. |
| 139 |
// If r does not EOF upon reading up to len(buf) bytes then readToEOF will return |
| 140 |
// io.ErrShortBuffer and an additional byte will be discarded from the reader. |
| 141 |
func readToEOF(r io.Reader, buf []byte) (n int, err error) { |
| 142 |
for err == nil && n < len(buf) { |
| 143 |
var nn int |
| 144 |
nn, err = r.Read(buf[n:]) |
| 145 |
n += nn |
| 146 |
} |
| 147 |
switch { |
| 148 |
case err == io.EOF: |
| 149 |
return n, nil |
| 150 |
case n == MaxBlockSize && err == nil: |
| 151 |
// This is paranoic, but some readers will return |
| 152 |
// quickly when passed a zero-length byte slice. |
| 153 |
var dummy [1]byte |
| 154 |
_, err = r.Read(dummy[:]) |
| 155 |
if err == nil { |
| 156 |
return n, io.ErrShortBuffer |
| 157 |
} |
| 158 |
if err == io.EOF { |
| 159 |
err = nil |
| 160 |
} |
| 161 |
} |
| 162 |
return n, err |
| 163 |
} |
| 164 |
|
| 165 |
func (b *block) readFrom(r io.ReadCloser) error { |
| 166 |
o := b.owner |
| 167 |
b.owner = nil |
| 168 |
n, err := readToEOF(r, b.data[:]) |
| 169 |
if err != nil { |
| 170 |
return err |
| 171 |
} |
| 172 |
b.buf = bytes.NewReader(b.data[:n]) |
| 173 |
b.owner = o |
| 174 |
b.magic = b.magic && b.len() == 0 |
| 175 |
return r.Close() |
| 176 |
} |
| 177 |
|
| 178 |
func (b *block) seek(offset int64) error { |
| 179 |
_, err := b.buf.Seek(offset, 0) |
| 180 |
if err == nil { |
| 181 |
b.offset.Block = uint16(offset) |
| 182 |
} |
| 183 |
return err |
| 184 |
} |
| 185 |
|
| 186 |
func (b *block) len() int { |
| 187 |
if b.buf == nil { |
| 188 |
return 0 |
| 189 |
} |
| 190 |
return b.buf.Len() |
| 191 |
} |
| 192 |
|
| 193 |
func (b *block) setBase(n int64) { |
| 194 |
b.base = n |
| 195 |
b.offset = Offset{File: n} |
| 196 |
} |
| 197 |
|
| 198 |
func (b *block) NextBase() int64 { |
| 199 |
size := int64(expectedMemberSize(b.h)) |
| 200 |
if size == -1 { |
| 201 |
return -1 |
| 202 |
} |
| 203 |
return b.base + size |
| 204 |
} |
| 205 |
|
| 206 |
func (b *block) setHeader(h gzip.Header) { |
| 207 |
b.h = h |
| 208 |
b.magic = h.OS == 0xff && |
| 209 |
// Test for zero time and old compress/gzip behaviour. |
| 210 |
(h.ModTime.IsZero() || h.ModTime.Equal(unixEpoch)) && |
| 211 |
h.Name == "" && |
| 212 |
h.Comment == "" && |
| 213 |
bytes.Equal(h.Extra, []byte("BC\x02\x00\x1b\x00")) |
| 214 |
} |
| 215 |
|
| 216 |
func (b *block) header() gzip.Header { return b.h } |
| 217 |
|
| 218 |
func (b *block) isMagicBlock() bool { return b.magic } |
| 219 |
|
| 220 |
func (b *block) setOwner(r *Reader) { |
| 221 |
b.owner = r |
| 222 |
b.used = false |
| 223 |
b.base = -1 |
| 224 |
b.h = gzip.Header{} |
| 225 |
b.offset = Offset{} |
| 226 |
b.buf = nil |
| 227 |
} |
| 228 |
|
| 229 |
func (b *block) ownedBy(r *Reader) bool { return b.owner == r } |
| 230 |
|
| 231 |
func (b *block) hasData() bool { return b.buf != nil } |
| 232 |
|
| 233 |
func (b *block) txOffset() Offset { return b.offset } |