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
|
/**
* @file llstreamqueue.h
* @author Nat Goodspeed
* @date 2012-01-04
* @brief Definition of LLStreamQueue
*
* $LicenseInfo:firstyear=2012&license=viewerlgpl$
* Copyright (c) 2012, Linden Research, Inc.
* $/LicenseInfo$
*/
#if ! defined(LL_LLSTREAMQUEUE_H)
#define LL_LLSTREAMQUEUE_H
#include <string>
#include <list>
#include <iosfwd> // std::streamsize
#include <boost/iostreams/categories.hpp>
/**
* This class is a growable buffer between a producer and consumer. It serves
* as a queue usable with Boost.Iostreams -- hence, a "stream queue."
*
* This is especially useful for buffering nonblocking I/O. For instance, we
* want application logic to be able to serialize LLSD to a std::ostream. We
* may write more data than the destination pipe can handle all at once, but
* it's imperative NOT to block the application-level serialization call. So
* we buffer it instead. Successive frames can try nonblocking writes to the
* destination pipe until all buffered data has been sent.
*
* Similarly, we want application logic be able to deserialize LLSD from a
* std::istream. Again, we must not block that deserialize call waiting for
* more data to arrive from the input pipe! Instead we build up a buffer over
* a number of frames, using successive nonblocking reads, until we have
* "enough" data to be able to present it through a std::istream.
*
* @note The use cases for this class overlap somewhat with those for the
* LLIOPipe/LLPumpIO hierarchies, and indeed we considered using those. This
* class has two virtues over the older machinery:
*
* # It's vastly simpler -- way fewer concepts. It's not clear to me whether
* there were ever LLIOPipe/etc. use cases that demanded all the fanciness
* rolled in, or whether they were simply overdesigned. In any case, no
* remaining Lindens will admit to familiarity with those classes -- and
* they're sufficiently obtuse that it would take considerable learning
* curve to figure out how to use them properly. The bottom line is that
* current management is not keen on any more engineers climbing that curve.
* # This class is designed around available components such as std::string,
* std::list, Boost.Iostreams. There's less proprietary code.
*/
template <typename Ch>
class LLGenericStreamQueue
{
public:
LLGenericStreamQueue():
mSize(0),
mClosed(false)
{}
/**
* Boost.Iostreams Source Device facade for use with other Boost.Iostreams
* functionality. LLGenericStreamQueue doesn't quite fit any of the Boost
* 1.48 Iostreams concepts; instead it behaves as both a Sink and a
* Source. This is its Source facade.
*/
struct Source
{
typedef Ch char_type;
typedef boost::iostreams::source_tag category;
/// Bind the underlying LLGenericStreamQueue
Source(LLGenericStreamQueue& sq):
mStreamQueue(sq)
{}
// Read up to n characters from the underlying data source into the
// buffer s, returning the number of characters read; return -1 to
// indicate EOF
std::streamsize read(Ch* s, std::streamsize n)
{
return mStreamQueue.read(s, n);
}
LLGenericStreamQueue& mStreamQueue;
};
/**
* Boost.Iostreams Sink Device facade for use with other Boost.Iostreams
* functionality. LLGenericStreamQueue doesn't quite fit any of the Boost
* 1.48 Iostreams concepts; instead it behaves as both a Sink and a
* Source. This is its Sink facade.
*/
struct Sink
{
typedef Ch char_type;
typedef boost::iostreams::sink_tag category;
/// Bind the underlying LLGenericStreamQueue
Sink(LLGenericStreamQueue& sq):
mStreamQueue(sq)
{}
/// Write up to n characters from the buffer s to the output sequence,
/// returning the number of characters written
std::streamsize write(const Ch* s, std::streamsize n)
{
return mStreamQueue.write(s, n);
}
/// Send EOF to consumer
void close()
{
mStreamQueue.close();
}
LLGenericStreamQueue& mStreamQueue;
};
/// Present Boost.Iostreams Source facade
Source asSource() { return Source(*this); }
/// Present Boost.Iostreams Sink facade
Sink asSink() { return Sink(*this); }
/// append data to buffer
std::streamsize write(const Ch* s, std::streamsize n)
{
// Unclear how often we might be asked to write 0 bytes -- perhaps a
// naive caller responding to an unready nonblocking read. But if we
// do get such a call, don't add a completely empty BufferList entry.
if (n == 0)
return n;
// We could implement this using a single std::string object, a la
// ostringstream. But the trouble with appending to a string is that
// you might have to recopy all previous contents to grow its size. If
// we want this to scale to large data volumes, better to allocate
// individual pieces.
mBuffer.push_back(string(s, n));
mSize += n;
return n;
}
/**
* Inform this LLGenericStreamQueue that no further data are forthcoming.
* For our purposes, close() is strictly a producer-side operation;
* there's little point in closing the consumer side.
*/
void close()
{
mClosed = true;
}
/// consume data from buffer
std::streamsize read(Ch* s, std::streamsize n)
{
// read() is actually a convenience method for peek() followed by
// skip().
std::streamsize got(peek(s, n));
// We can only skip() as many characters as we can peek(); ignore
// skip() return here.
skip(n);
return got;
}
/// Retrieve data from buffer without consuming. Like read(), return -1 on
/// EOF.
std::streamsize peek(Ch* s, std::streamsize n) const;
/// Consume data from buffer without retrieving. Unlike read() and peek(),
/// at EOF we simply skip 0 characters.
std::streamsize skip(std::streamsize n);
/// How many characters do we currently have buffered?
std::streamsize size() const
{
return mSize;
}
private:
typedef std::basic_string<Ch> string;
typedef std::list<string> BufferList;
BufferList mBuffer;
std::streamsize mSize;
bool mClosed;
};
template <typename Ch>
std::streamsize LLGenericStreamQueue<Ch>::peek(Ch* s, std::streamsize n) const
{
// Here we may have to build up 'n' characters from an arbitrary
// number of individual BufferList entries.
typename BufferList::const_iterator bli(mBuffer.begin()), blend(mBuffer.end());
// Indicate EOF if producer has closed the pipe AND we've exhausted
// all previously-buffered data.
if (mClosed && bli == blend)
{
return -1;
}
// Here either producer hasn't yet closed, or we haven't yet exhausted
// remaining data.
std::streamsize needed(n), got(0);
// Loop until either we run out of BufferList entries or we've
// completely satisfied the request.
for ( ; bli != blend && needed; ++bli)
{
std::streamsize chunk(std::min(needed, std::streamsize(bli->length())));
std::copy(bli->begin(), bli->begin() + chunk, s);
needed -= chunk;
s += chunk;
got += chunk;
}
return got;
}
template <typename Ch>
std::streamsize LLGenericStreamQueue<Ch>::skip(std::streamsize n)
{
typename BufferList::iterator bli(mBuffer.begin()), blend(mBuffer.end());
std::streamsize toskip(n), skipped(0);
while (bli != blend && toskip >= bli->length())
{
std::streamsize chunk(bli->length());
typename BufferList::iterator zap(bli++);
mBuffer.erase(zap);
mSize -= chunk;
toskip -= chunk;
skipped += chunk;
}
if (bli != blend && toskip)
{
bli->erase(bli->begin(), bli->begin() + toskip);
mSize -= toskip;
skipped += toskip;
}
return skipped;
}
typedef LLGenericStreamQueue<char> LLStreamQueue;
typedef LLGenericStreamQueue<wchar_t> LLWStreamQueue;
#endif /* ! defined(LL_LLSTREAMQUEUE_H) */
|