1289177Speter/* reps.h --- FSX representation container
2289177Speter *
3289177Speter * ====================================================================
4289177Speter *    Licensed to the Apache Software Foundation (ASF) under one
5289177Speter *    or more contributor license agreements.  See the NOTICE file
6289177Speter *    distributed with this work for additional information
7289177Speter *    regarding copyright ownership.  The ASF licenses this file
8289177Speter *    to you under the Apache License, Version 2.0 (the
9289177Speter *    "License"); you may not use this file except in compliance
10289177Speter *    with the License.  You may obtain a copy of the License at
11289177Speter *
12289177Speter *      http://www.apache.org/licenses/LICENSE-2.0
13289177Speter *
14289177Speter *    Unless required by applicable law or agreed to in writing,
15289177Speter *    software distributed under the License is distributed on an
16289177Speter *    "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
17289177Speter *    KIND, either express or implied.  See the License for the
18289177Speter *    specific language governing permissions and limitations
19289177Speter *    under the License.
20289177Speter * ====================================================================
21289177Speter */
22289177Speter
23289177Speter#ifndef SVN_LIBSVN_FS__REPS_H
24289177Speter#define SVN_LIBSVN_FS__REPS_H
25289177Speter
26289177Speter#include "svn_io.h"
27289177Speter#include "fs.h"
28289177Speter
29289177Speter/* This container type implements the start-delta (aka pick lists) data
30289177Speter * structure plus functions to create it and read data from it.  The key
31289177Speter * point is to identify common sub-strings within a whole set of fulltexts
32289177Speter * instead of only two as in the classic txdelta code.
33289177Speter *
34289177Speter * Because it is relatively expensive to optimize the final in-memory
35289177Speter * layout, representation containers cannot be updated.  A builder object
36289177Speter * will do most of the space saving when adding fulltexts but the final
37289177Speter * data will only be created immediately before serializing everything to
38289177Speter * disk.  So, builders are write only and representation containers are
39289177Speter * read-only.
40289177Speter *
41289177Speter * Extracting data from a representation container is O(length) but it
42289177Speter * may require multiple iterations if base representations outside the
43289177Speter * container were used.  Therefore, you will first create an extractor
44289177Speter * object (this may happen while holding a cache lock) and the you need
45289177Speter * to "drive" the extractor outside any cache context.
46289177Speter */
47289177Speter
48289177Speter/* A write-only constructor object for representation containers.
49289177Speter */
50289177Spetertypedef struct svn_fs_x__reps_builder_t svn_fs_x__reps_builder_t;
51289177Speter
52289177Speter/* A read-only representation container -
53289177Speter * an opaque collection of fulltexts, i.e. byte strings.
54289177Speter */
55289177Spetertypedef struct svn_fs_x__reps_t svn_fs_x__reps_t;
56289177Speter
57289177Speter/* The fulltext extractor utility object.
58289177Speter */
59289177Spetertypedef struct svn_fs_x__rep_extractor_t svn_fs_x__rep_extractor_t;
60289177Speter
61289177Speter/* Baton type to be passed to svn_fs_x__reps_get_func.
62289177Speter */
63289177Spetertypedef struct svn_fs_x__reps_baton_t
64289177Speter{
65289177Speter  /* filesystem the resulting extractor shall operate on */
66289177Speter  svn_fs_t *fs;
67289177Speter
68289177Speter  /* element index of the item to extract from the container */
69289177Speter  apr_size_t idx;
70289177Speter} svn_fs_x__reps_baton_t;
71289177Speter
72289177Speter/* Create and populate noderev containers. */
73289177Speter
74289177Speter/* Create and return a new builder object, allocated in RESULT_POOL.
75289177Speter */
76289177Spetersvn_fs_x__reps_builder_t *
77289177Spetersvn_fs_x__reps_builder_create(svn_fs_t *fs,
78289177Speter                              apr_pool_t *result_pool);
79289177Speter
80289177Speter/* To BUILDER, add reference to the fulltext currently stored in
81289177Speter * representation REP.  Substrings matching with any of the base reps
82289177Speter * in BUILDER can be removed from the text base and be replaced by
83289177Speter * references to those base representations.
84289177Speter *
85289177Speter * The PRIORITY is a mere hint on which base representations should
86289177Speter * preferred in case we could re-use the same contents from multiple bases.
87289177Speter * Higher numerical value means higher priority / likelihood of being
88289177Speter * selected over others.
89289177Speter *
90289177Speter * Use SCRATCH_POOL for temporary allocations.
91289177Speter */
92289177Spetersvn_error_t *
93289177Spetersvn_fs_x__reps_add_base(svn_fs_x__reps_builder_t *builder,
94289177Speter                        svn_fs_x__representation_t *rep,
95289177Speter                        int priority,
96289177Speter                        apr_pool_t *scratch_pool);
97289177Speter
98289177Speter/* Add the byte string CONTENTS to BUILDER.  Return the item index under
99289177Speter * which the fulltext can be retrieved from the final container in *REP_IDX.
100289177Speter */
101289177Spetersvn_error_t *
102289177Spetersvn_fs_x__reps_add(apr_size_t *rep_idx,
103289177Speter                   svn_fs_x__reps_builder_t *builder,
104289177Speter                   const svn_string_t *contents);
105289177Speter
106289177Speter/* Return a rough estimate in bytes for the serialized representation
107289177Speter * of BUILDER.
108289177Speter */
109289177Speterapr_size_t
110289177Spetersvn_fs_x__reps_estimate_size(const svn_fs_x__reps_builder_t *builder);
111289177Speter
112289177Speter/* Read from representation containers. */
113289177Speter
114289177Speter/* For fulltext IDX in CONTAINER in filesystem FS, create an extract object
115289177Speter * allocated in POOL and return it in *EXTRACTOR.
116289177Speter */
117289177Spetersvn_error_t *
118289177Spetersvn_fs_x__reps_get(svn_fs_x__rep_extractor_t **extractor,
119289177Speter                   svn_fs_t *fs,
120289177Speter                   const svn_fs_x__reps_t *container,
121289177Speter                   apr_size_t idx,
122289177Speter                   apr_pool_t *pool);
123289177Speter
124289177Speter/* Let the EXTRACTOR object fetch all parts of the desired fulltext and
125289177Speter * return the latter in *CONTENTS.  If SIZE is not 0, return SIZE bytes
126289177Speter * starting at offset START_OFFSET of the full contents.  If that range
127289177Speter * lies partly or completely outside the content, clip it accordingly.
128289177Speter * Allocate the result in RESULT_POOL and use SCRATCH_POOL for temporary
129289177Speter * allocations.
130289177Speter *
131289177Speter * Note, you may not run this inside a cache access function.
132289177Speter */
133289177Spetersvn_error_t *
134289177Spetersvn_fs_x__extractor_drive(svn_stringbuf_t** contents,
135289177Speter                          svn_fs_x__rep_extractor_t* extractor,
136289177Speter                          apr_size_t start_offset,
137289177Speter                          apr_size_t size,
138289177Speter                          apr_pool_t* result_pool,
139289177Speter                          apr_pool_t* scratch_pool);
140289177Speter
141289177Speter/* I/O interface. */
142289177Speter
143289177Speter/* Write a serialized representation of the final container described by
144289177Speter * BUILDER to STREAM.  Use SCRATCH_POOL for temporary allocations.
145289177Speter */
146289177Spetersvn_error_t *
147289177Spetersvn_fs_x__write_reps_container(svn_stream_t *stream,
148289177Speter                               const svn_fs_x__reps_builder_t *builder,
149289177Speter                               apr_pool_t *scratch_pool);
150289177Speter
151289177Speter/* Read a representations container from its serialized representation in
152289177Speter * STREAM.  Allocate the result in RESULT_POOL and return it in *CONTAINER.
153289177Speter * Use SCRATCH_POOL for temporary allocations.
154289177Speter */
155289177Spetersvn_error_t *
156289177Spetersvn_fs_x__read_reps_container(svn_fs_x__reps_t **container,
157289177Speter                              svn_stream_t *stream,
158289177Speter                              apr_pool_t *result_pool,
159289177Speter                              apr_pool_t *scratch_pool);
160289177Speter
161289177Speter/* Implements #svn_cache__serialize_func_t for svn_fs_x__reps_t objects.
162289177Speter */
163289177Spetersvn_error_t *
164289177Spetersvn_fs_x__serialize_reps_container(void **data,
165289177Speter                                   apr_size_t *data_len,
166289177Speter                                   void *in,
167289177Speter                                   apr_pool_t *pool);
168289177Speter
169289177Speter/* Implements #svn_cache__deserialize_func_t for svn_fs_x__reps_t objects.
170289177Speter */
171289177Spetersvn_error_t *
172289177Spetersvn_fs_x__deserialize_reps_container(void **out,
173289177Speter                                     void *data,
174289177Speter                                     apr_size_t data_len,
175289177Speter                                     apr_pool_t *pool);
176289177Speter
177289177Speter/* Implements svn_cache__partial_getter_func_t for svn_fs_x__reps_t,
178289177Speter * setting *OUT to an svn_fs_x__rep_extractor_t object defined by the
179289177Speter * svn_fs_x__reps_baton_t passed in as *BATON.  This function is similar
180289177Speter * to svn_fs_x__reps_get but operates on the cache serialized
181289177Speter * representation of the container.
182289177Speter */
183289177Spetersvn_error_t *
184289177Spetersvn_fs_x__reps_get_func(void **out,
185289177Speter                        const void *data,
186289177Speter                        apr_size_t data_len,
187289177Speter                        void *baton,
188289177Speter                        apr_pool_t *pool);
189289177Speter
190289177Speter#endif
191