1GITFORMAT-CHUNK(5) Git Manual GITFORMAT-CHUNK(5)
2
6 gitformat-chunk - Chunk-based file formats
7
9 Used by gitformat-commit-graph(5) and the "MIDX" format (see the pack
10 format documentation in gitformat-pack(5)).
11
13 Some file formats in Git use a common concept of "chunks" to describe
14 sections of the file. This allows structured access to a large file by
15 scanning a small "table of contents" for the remaining data. This
16 common format is used by the commit-graph and multi-pack-index files.
17 See the multi-pack-index format in gitformat-pack(5) and the
18 commit-graph format in gitformat-commit-graph(5) for how they use the
19 chunks to describe structured data.
20 A chunk-based file format begins with some header information custom to
22 that format. That header should include enough information to identify
23 the file type, format version, and number of chunks in the file. From
24 this information, that file can determine the start of the chunk-based
25 region.
26 The chunk-based region starts with a table of contents describing where
28 each chunk starts and ends. This consists of (C+1) rows of 12 bytes
29 each, where C is the number of chunks. Consider the following table:
30 | Chunk ID (4 bytes) | Chunk Offset (8 bytes) |
32 |--------------------|------------------------|
33 | ID[0] | OFFSET[0] |
34 | ... | ... |
35 | ID[C] | OFFSET[C] |
36 | 0x0000 | OFFSET[C+1] |
37 Each row consists of a 4-byte chunk identifier (ID) and an 8-byte
39 offset. Each integer is stored in network-byte order.
40 The chunk identifier ID[i] is a label for the data stored within this
42 file from OFFSET[i] (inclusive) to OFFSET[i+1] (exclusive). Thus, the
43 size of the i`th chunk is equal to the difference between `OFFSET[i+1]
44 and OFFSET[i]. This requires that the chunk data appears contiguously
45 in the same order as the table of contents.
46 The final entry in the table of contents must be four zero bytes. This
48 confirms that the table of contents is ending and provides the offset
49 for the end of the chunk-based data.
50 Note: The chunk-based format expects that the file contains at least a
52 trailing hash after OFFSET[C+1].
53 Functions for working with chunk-based file formats are declared in
55 chunk-format.h. Using these methods provide extra checks that assist
56 developers when creating new file formats.
57
59 To write a chunk-based file format, create a struct chunkfile by
60 calling init_chunkfile() and pass a struct hashfile pointer. The caller
61 is responsible for opening the hashfile and writing header information
62 so the file format is identifiable before the chunk-based format
63 begins.
64 Then, call add_chunk() for each chunk that is intended for writing.
66 This populates the chunkfile with information about the order and size
67 of each chunk to write. Provide a chunk_write_fn function pointer to
68 perform the write of the chunk data upon request.
69 Call write_chunkfile() to write the table of contents to the hashfile
71 followed by each of the chunks. This will verify that each chunk wrote
72 the expected amount of data so the table of contents is correct.
73 Finally, call free_chunkfile() to clear the struct chunkfile data. The
75 caller is responsible for finalizing the hashfile by writing the
76 trailing hash and closing the file.
77
79 To read a chunk-based file format, the file must be opened as a
80 memory-mapped region. The chunk-format API expects that the entire file
81 is mapped as a contiguous memory region.
82 Initialize a struct chunkfile pointer with init_chunkfile(NULL).
84 After reading the header information from the beginning of the file,
86 including the chunk count, call read_table_of_contents() to populate
87 the struct chunkfile with the list of chunks, their offsets, and their
88 sizes.
89 Extract the data information for each chunk using pair_chunk() or
91 read_chunk():
92 • pair_chunk() assigns a given pointer with the location inside the
94 memory-mapped file corresponding to that chunk’s offset. If the
95 chunk does not exist, then the pointer is not modified.
96 • read_chunk() takes a chunk_read_fn function pointer and calls it
98 with the appropriate initial pointer and size information. The
99 function is not called if the chunk does not exist. Use this method
100 to read chunks if you need to perform immediate parsing or if you
101 need to execute logic based on the size of the chunk.
102 After calling these methods, call free_chunkfile() to clear the struct
104 chunkfile data. This will not close the memory-mapped region. Callers
105 are expected to own that data for the timeframe the pointers into the
106 region are needed.
107
109 These file formats use the chunk-format API, and can be used as
110 examples for future formats:
111 • commit-graph: see write_commit_graph_file() and
113 parse_commit_graph() in commit-graph.c for how the chunk-format API
114 is used to write and parse the commit-graph file format documented
115 in the commit-graph file format in gitformat-commit-graph(5).
116 • multi-pack-index: see write_midx_internal() and
118 load_multi_pack_index() in midx.c for how the chunk-format API is
119 used to write and parse the multi-pack-index file format documented
120 in the multi-pack-index file format section of gitformat-pack(5).
121
123 Part of the git(1) suite
124Git 2.43.0 11/20/2023 GITFORMAT-CHUNK(5)