svn_packed_data.h revision 299742 
1/* packed_data.h : Interface to the packed binary stream data structure
2 *
3 * ====================================================================
4 *    Licensed to the Apache Software Foundation (ASF) under one
5 *    or more contributor license agreements.  See the NOTICE file
6 *    distributed with this work for additional information
7 *    regarding copyright ownership.  The ASF licenses this file
8 *    to you under the Apache License, Version 2.0 (the
9 *    "License"); you may not use this file except in compliance
10 *    with the License.  You may obtain a copy of the License at
11 *
12 *      http://www.apache.org/licenses/LICENSE-2.0
13 *
14 *    Unless required by applicable law or agreed to in writing,
15 *    software distributed under the License is distributed on an
16 *    "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
17 *    KIND, either express or implied.  See the License for the
18 *    specific language governing permissions and limitations
19 *    under the License.
20 * ====================================================================
21 */
22
23#ifndef SVN_PACKED_DATA_H
24#define SVN_PACKED_DATA_H
25
26#include "svn_string.h"
27#include "svn_io.h"
28
29#ifdef __cplusplus
30extern "C" {
31#endif /* __cplusplus */
32
33/* This API provides Yet Another Serialization Framework.
34 *
35 * It is geared towards efficiently encoding collections of structured
36 * binary data (e.g. an array of noderev objects).  The basic idea is to
37 * transform them into hierarchies of streams with each stream usually
38 * corresponding to a single attribute in the original data structures.
39 * The user is free model the mapping structure <-> streams mapping as she
40 * sees fit.
41 *
42 * With all data inside the same (sub-)stream carrying similar attribute
43 * values, the whole stream lends itself to data compression.  Strings /
44 * plain byte sequences will be stored as is.  Numbers use a 7b/8b encoding
45 * scheme to eliminate leading zeros.  Because values are often dependent
46 * (increasing offsets, roughly similar revision number, etc.), streams
47 * can be configured as storing (hopefully shorter) deltas instead of the
48 * original value.
49 *
50 * Two stream types are provided: integer and byte streams.  While the
51 * first store 64 bit integers only and can be configured to assume
52 * signed and / or deltifyable data, the second will store arbitrary
53 * byte sequences including their length.  At the root level, you may
54 * create an arbitrary number of integer and byte streams.  Any stream
55 * may have an arbitrary number of sub-streams of the same kind.  You
56 * should create the full stream hierarchy before writing any data to it.
57 *
58 * As a convenience, when an integer stream has sub-streams, you may write
59 * to the parent stream instead of all sub-streams individually and the
60 * values will be passed down automatically in a round-robin fashion.
61 * Reading from the parent stream is similarly supported.
62 *
63 * When all data has been added to the stream, it can be written to an
64 * ordinary svn_stream_t.  First, we write a description of the stream
65 * structure (types, sub-streams, sizes and configurations) followed by
66 * zlib compressed stream content.  For each top-level stream, all sub-
67 * stream data will be concatenated and then compressed as a single block.
68 * To maximize the effect of this, make sure all data in that stream
69 * hierarchy has a similar value distribution.
70 *
71 * Reading data starts with an svn_stream_t and automatically recreates
72 * the stream hierarchies.  You only need to extract data from it in the
73 * same order as you wrote it.
74 *
75 * Although not enforced programmatically, you may either only write to a
76 * stream hierarchy or only read from it but you cannot do both on the
77 * same data structure.
78 */
79
80
81
82/* We pack / unpack integers en block to minimize calling and setup overhead.
83 * This is the number of integers we put into a buffer before writing them
84 * them to / after reading them from the 7b/8b stream.  Under 64 bits, this
85 * value creates a 128 byte data structure (14 + 2 integers, 8 bytes each).
86 */
87#define SVN__PACKED_DATA_BUFFER_SIZE 14
88
89
90/* Data types. */
91
92/* Opaque type for the root object.
93 */
94typedef struct svn_packed__data_root_t svn_packed__data_root_t;
95
96/* Opaque type for byte streams.
97 */
98typedef struct svn_packed__byte_stream_t svn_packed__byte_stream_t;
99
100/* Semi-opaque type for integer streams.  We expose the unpacked buffer
101 * to allow for replacing svn_packed__add_uint and friends by macros.
102 */
103typedef struct svn_packed__int_stream_t
104{
105  /* pointer to the remainder of the data structure */
106  void *private_data;
107
108  /* number of value entries in BUFFER */
109  apr_size_t buffer_used;
110
111  /* unpacked integers (either yet to be packed or pre-fetched from the
112   * packed buffers).  Only the first BUFFER_USED entries are valid. */
113  apr_uint64_t buffer[SVN__PACKED_DATA_BUFFER_SIZE];
114} svn_packed__int_stream_t;
115
116
117/* Writing data. */
118
119/* Return a new serialization root object, allocated in POOL.
120 */
121svn_packed__data_root_t *
122svn_packed__data_create_root(apr_pool_t *pool);
123
124/* Create and return a new top-level integer stream in ROOT.  If signed,
125 * negative numbers will be put into that stream, SIGNED_INTS should be
126 * TRUE as a more efficient encoding will be used in that case.  Set
127 * DIFF to TRUE if you expect the difference between consecutive numbers
128 * to be much smaller (~100 times) than the actual numbers.
129 */
130svn_packed__int_stream_t *
131svn_packed__create_int_stream(svn_packed__data_root_t *root,
132                              svn_boolean_t diff,
133                              svn_boolean_t signed_ints);
134
135/* Create and return a sub-stream to the existing integer stream PARENT.
136 * If signed, negative numbers will be put into that stream, SIGNED_INTS
137 * should be TRUE as a more efficient encoding will be used in that case.
138 * Set DIFF to TRUE if you expect the difference between consecutive numbers
139 * to be much smaller (~100 times) than the actual numbers.
140 */
141svn_packed__int_stream_t *
142svn_packed__create_int_substream(svn_packed__int_stream_t *parent,
143                                 svn_boolean_t diff,
144                                 svn_boolean_t signed_ints);
145
146/* Create and return a new top-level byte sequence stream in ROOT.
147 */
148svn_packed__byte_stream_t *
149svn_packed__create_bytes_stream(svn_packed__data_root_t *root);
150
151/* Write the unsigned integer VALUE to STEAM.
152 */
153void
154svn_packed__add_uint(svn_packed__int_stream_t *stream,
155                     apr_uint64_t value);
156
157/* Write the signed integer VALUE to STEAM.
158 */
159void
160svn_packed__add_int(svn_packed__int_stream_t *stream,
161                    apr_int64_t value);
162
163/* Write the sequence stating at DATA containing LEN bytes to STEAM.
164 */
165void
166svn_packed__add_bytes(svn_packed__byte_stream_t *stream,
167                      const char *data,
168                      apr_size_t len);
169
170/* Write all contents of ROOT (including all sub-streams) to STREAM.
171 * Use SCRATCH_POOL for temporary allocations.
172 */
173svn_error_t *
174svn_packed__data_write(svn_stream_t *stream,
175                       svn_packed__data_root_t *root,
176                       apr_pool_t *scratch_pool);
177
178
179/* Reading data. */
180
181/* Return the first integer stream in ROOT.  Returns NULL in case there
182 * aren't any.
183 */
184svn_packed__int_stream_t *
185svn_packed__first_int_stream(svn_packed__data_root_t *root);
186
187/* Return the first byte sequence stream in ROOT.  Returns NULL in case
188 * there aren't any.
189 */
190svn_packed__byte_stream_t *
191svn_packed__first_byte_stream(svn_packed__data_root_t *root);
192
193/* Return the next (sibling) integer stream to STREAM.  Returns NULL in
194 * case there isn't any.
195 */
196svn_packed__int_stream_t *
197svn_packed__next_int_stream(svn_packed__int_stream_t *stream);
198
199/* Return the next (sibling) byte sequence stream to STREAM.  Returns NULL
200 * in case there isn't any.
201 */
202svn_packed__byte_stream_t *
203svn_packed__next_byte_stream(svn_packed__byte_stream_t *stream);
204
205/* Return the first sub-stream of STREAM.  Returns NULL in case there
206 * isn't any.
207 */
208svn_packed__int_stream_t *
209svn_packed__first_int_substream(svn_packed__int_stream_t *stream);
210
211/* Return the number of integers left to read from STREAM.
212 */
213apr_size_t
214svn_packed__int_count(svn_packed__int_stream_t *stream);
215
216/* Return the number of bytes left to read from STREAM.
217 */
218apr_size_t
219svn_packed__byte_count(svn_packed__byte_stream_t *stream);
220
221/* Return the next number from STREAM as unsigned integer.  Returns 0 when
222 * reading beyond the end of the stream.
223 */
224apr_uint64_t
225svn_packed__get_uint(svn_packed__int_stream_t *stream);
226
227/* Return the next number from STREAM as signed integer.  Returns 0 when
228 * reading beyond the end of the stream.
229 */
230apr_int64_t
231svn_packed__get_int(svn_packed__int_stream_t *stream);
232
233/* Return the next byte sequence from STREAM and set *LEN to the length
234 * of that sequence.  Sets *LEN to 0 when reading beyond the end of the
235 * stream.
236 */
237const char *
238svn_packed__get_bytes(svn_packed__byte_stream_t *stream,
239                      apr_size_t *len);
240
241/* Allocate a new packed data root in RESULT_POOL, read its structure and
242 * stream contents from STREAM and return it in *ROOT_P.  Use SCRATCH_POOL
243 * for temporary allocations.
244 */
245svn_error_t *
246svn_packed__data_read(svn_packed__data_root_t **root_p,
247                      svn_stream_t *stream,
248                      apr_pool_t *result_pool,
249                      apr_pool_t *scratch_pool);
250
251#ifdef __cplusplus
252}
253#endif /* __cplusplus */
254
255#endif /* SVN_PACKED_DATA_H */
256