btr support for para_client.
[paraslash.git] / buffer_tree.h
1 /**
2 * Buffer trees and buffer tree nodes.
3 *
4 * The buffer tree API offers a more powerful method than standard unix pipes
5 * for managing the data flow from the producer of the data (e.g. the network
6 * receiver) to its consumer(s) (e.g. a sound card).
7 *
8 * Each data buffer starts its way from the root of the buffer tree. At each
9 * child node the data is investigated and new data is fed to each child.
10 * Everything happens within one single-treaded process. There are no file
11 * descriptors, no calls to read() or write().
12 *
13 * A buffer tree consists of buffer tree nodes. Usually, there is exactly one
14 * node in the buffer tree, the root node, which has no parent. Every node
15 * different from the root node has exactly one parent.
16 *
17 * The root node represents a data source. Root nodes are thus used by the
18 * receivers of paraslash. Also, reading from stdin is realized as the root
19 * node of a buffer tree.
20 *
21 * Each node may have arbitrary many children, including none. Nodes with no
22 * children are called leaf nodes. They represent a data sink, like the alsa or
23 * the file writer.
24 *
25 * Hence there are three different types of buffer tree nodes: The root node
26 * and the leaf nodes and nodes which have both a parent and at least one
27 * child. Such a node is called an internal node.
28 *
29 * Internal nodes represent filters through which data buffers flow, possibly
30 * while being altered on their way to the children of the node. Examples of
31 * internal nodes are audio file decoders (mp3dec, oggdec, ...), but also the
32 * check for a wav header is implemented as an internal buffer tree node.
33 *
34 * Whenever a node in the buffer tree creates output, either by creating a new
35 * buffer or by pushing down buffers received from its parent, references to
36 * that buffer are created for all children of the node. The buffer tree code
37 * tries hard to avoid to copy buffer contents, but is forced to do so in case
38 * there are alignment constraints.
39 *
40 * Communication between nodes is possible via the btr_exec_up() function.
41 * For example, in para_audiod the alsa writer asks all parent nodes
42 * for for the number of channels and the sample rate of the current
43 * audio file.
44 */
45 struct btr_node;
46
47 /**
48 * Buffer pools - An alternative to malloc/free buffer management.
49 *
50 * Non-leaf nodes usually create output to be processed by their children. The
51 * data must be fed through the output channel(s) of the node in order to make
52 * that data available to each child.
53 *
54 * The easiest way to do so is to malloc() a buffer, fill it, and then call
55 * btr_add_output(). This adds references to that buffer to all children. The
56 * buffer is automatically freed if no buffer tree node is using it any more.
57 *
58 * This approach, while simple, has some drawbacks, especially affecting the
59 * root nodes of the buffer tree. Often the data source which is represented by
60 * a root node does not know in advance how much data will be available.
61 * Therefore the allocated buffer is either larger than what can currently be
62 * read, or is too small so that multiple buffers have to be used.
63 *
64 * While this could be worked around by using a large buffer and calling
65 * realloc() afterwards to shrink the buffer according to how much has been
66 * read, there is a second problem which comes from the alignment constraints
67 * of some filters, mainly the decoders. These need a minimal amount of data to
68 * proceed, and most of them even need this amount as one contiguous buffer,
69 * i.e. not spread out over two or more buffers.
70 *
71 * Although the buffer tree code handles this case just fine, it can be
72 * expensive because two or more buffers must be combined by copying buffer
73 * contents around in order to satisfy the constraint.
74 *
75 * This is where buffer pools come into play. Buffer pools try to satisfy
76 * alignment constraints without copying buffer content whenever possible. To
77 * avoid spreading out the input data over the address space like in the
78 * malloc/free approach, a fixed large contiguous buffer (the area) is used
79 * instead. A buffer pool consists basically of an area and two pointers, the
80 * read head and the write head.
81 *
82 * Once a buffer pool has been created, its node, e.g. a receiver, obtains the
83 * current value of the write head and writes new data to this location. Then
84 * it calls btr_add_output_pool() to tell much data it has written. This
85 * advances the write head accordingly, and it also creates references to the
86 * newly written part of the area for the children of the node to consume.
87 *
88 * Child nodes consume data by working through their input queue, which is a
89 * list of buffer references. Once the content of a buffer is no longer needed
90 * by a child node, the child calls btr_consume() to indicate the amount of
91 * data which can be dropped from the child's point of view. If no reference
92 * to some region of the buffer pool area remains, the read head of the buffer
93 * pool advances, making space available for the receiver node to fill.
94 *
95 * No matter if malloc() or a buffer pool is used, the buffer tree code takes
96 * care of alignment constraints imposed by the consumers. In the buffer pool
97 * case, automatic merging of references to contiguous buffers is performed.
98 * memcpy is only used if a constraint can not be satisfied by using the
99 * remaining part of the area only. This only happens when the end of the area
100 * is reached.
101 */
102
103 struct btr_pool;
104 typedef int (*btr_command_handler)(struct btr_node *btrn,
105 const char *command, char **result);
106
107 enum btr_node_type {
108 BTR_NT_ROOT,
109 BTR_NT_INTERNAL,
110 BTR_NT_LEAF,
111 };
112
113 struct btr_node_description {
114 const char *name;
115 struct btr_node *parent;
116 struct btr_node *child;
117 btr_command_handler handler;
118 void *context;
119 };
120
121 size_t btr_pool_size(struct btr_pool *btrp);
122 struct btr_pool *btr_pool_new(const char *name, size_t area_size);
123 void btr_pool_free(struct btr_pool *btrp);
124 size_t btr_pool_get_buffer(struct btr_pool *btrp, char **result);
125 void btr_pool_allocate(struct btr_pool *btrp, size_t size);
126 void btr_add_output_pool(struct btr_pool *btrp, size_t size,
127 struct btr_node *btrn);
128 size_t btr_pool_unused(struct btr_pool *btrp);
129 void btr_copy(const void *src, size_t n, struct btr_pool *btrp,
130 struct btr_node *btrn);
131
132 struct btr_node *btr_new_node(struct btr_node_description *bnd);
133 void btr_remove_node(struct btr_node *btrn);
134 void btr_free_node(struct btr_node *btrn);
135 void btr_add_output(char *buf, size_t size, struct btr_node *btrn);
136 bool btr_no_children(struct btr_node *btrn);
137 size_t btr_bytes_pending(struct btr_node *btrn);
138 size_t btr_get_input_queue_size(struct btr_node *btrn);
139 bool btr_no_parent(struct btr_node *btrn);
140 size_t btr_next_buffer(struct btr_node *btrn, char **bufp);
141 void btr_consume(struct btr_node *btrn, size_t numbytes);
142 int btr_exec(struct btr_node *btrn, const char *command, char **value_result);
143 int btr_exec_up(struct btr_node *btrn, const char *command, char **value_result);
144 void btr_splice_out_node(struct btr_node *btrn);
145 void btr_pushdown(struct btr_node *btrn);
146 void *btr_context(struct btr_node *btrn);
147 void btr_merge(struct btr_node *btrn, size_t dest_size);
148 bool btr_eof(struct btr_node *btrn);
149 void btr_log_tree(struct btr_node *btrn, int loglevel);
150 int btr_pushdown_one(struct btr_node *btrn);
151 bool btr_inplace_ok(struct btr_node *btrn);
152 int btr_node_status(struct btr_node *btrn, size_t min_iqs,
153 enum btr_node_type type);
154 void btr_get_node_start(struct btr_node *btrn, struct timeval *tv);
155 struct btr_node *btr_search_node(const char *name, struct btr_node *root);