1 gitformat-commit-graph(5)
2 =========================
6 gitformat-commit-graph - Git commit-graph format
11 $GIT_DIR/objects/info/commit-graph
12 $GIT_DIR/objects/info/commit-graphs/*
17 The Git commit-graph stores a list of commit OIDs and some associated
20 - The generation number of the commit.
26 - The parents of the commit, stored using positional references within
29 - The Bloom filter of the commit carrying the paths that were changed between
30 the commit and its first parent, if requested.
32 These positional references are stored as unsigned 32-bit integers
33 corresponding to the array position within the list of commit OIDs. Due
34 to some special constants we use to track parents, we can store at most
35 (1 << 30) + (1 << 29) + (1 << 28) - 1 (around 1.8 billion) commits.
37 == Commit-graph files have the following format:
39 In order to allow extensions that add extra data to the graph, we organize
40 the body into "chunks" and provide a binary lookup table at the beginning
41 of the body. The header includes certain values, such as number of chunks
44 All multi-byte numbers are in network byte order.
49 The signature is: {'C', 'G', 'P', 'H'}
51 1-byte version number:
52 Currently, the only valid version is 1.
55 We infer the hash length (H) from this value:
58 If the hash type does not match the repository's hash algorithm, the
59 commit-graph file should be ignored with a warning presented to the
62 1-byte number (C) of "chunks"
64 1-byte number (B) of base commit-graphs
65 We infer the length (H*B) of the Base Graphs chunk
70 (C + 1) * 12 bytes listing the table of contents for the chunks:
71 First 4 bytes describe the chunk id. Value 0 is a terminating label.
72 Other 8 bytes provide the byte-offset in current file for chunk to
73 start. (Chunks are ordered contiguously in the file, so you can infer
74 the length using the next chunk position if necessary.) Each chunk
75 ID appears at most once.
77 The CHUNK LOOKUP matches the table of contents from
78 the chunk-based file format, see linkgit:gitformat-chunk[5]
80 The remaining data in the body is described one chunk at a time, and
81 these chunks may be given in any order. Chunks are required unless
86 ==== OID Fanout (ID: {'O', 'I', 'D', 'F'}) (256 * 4 bytes)
87 The ith entry, F[i], stores the number of OIDs with first
88 byte at most i. Thus F[255] stores the total
89 number of commits (N).
91 ==== OID Lookup (ID: {'O', 'I', 'D', 'L'}) (N * H bytes)
92 The OIDs for all commits in the graph, sorted in ascending order.
94 ==== Commit Data (ID: {'C', 'D', 'A', 'T' }) (N * (H + 16) bytes)
95 * The first H bytes are for the OID of the root tree.
96 * The next 8 bytes are for the positions of the first two parents
97 of the ith commit. Stores value 0x70000000 if no parent in that
98 position. If there are more than two parents, the second value
99 has its most-significant bit on and the other bits store an array
100 position into the Extra Edge List chunk.
101 * The next 8 bytes store the topological level (generation number v1)
103 the commit time in seconds since EPOCH. The generation number
104 uses the higher 30 bits of the first 4 bytes, while the commit
105 time uses the 32 bits of the second 4 bytes, along with the lowest
106 2 bits of the lowest byte, storing the 33rd and 34th bit of the
109 ==== Generation Data (ID: {'G', 'D', 'A', '2' }) (N * 4 bytes) [Optional]
110 * This list of 4-byte values store corrected commit date offsets for the
111 commits, arranged in the same order as commit data chunk.
112 * If the corrected commit date offset cannot be stored within 31 bits,
113 the value has its most-significant bit on and the other bits store
114 the position of corrected commit date into the Generation Data Overflow
116 * Generation Data chunk is present only when commit-graph file is written
117 by compatible versions of Git and in case of split commit-graph chains,
118 the topmost layer also has Generation Data chunk.
120 ==== Generation Data Overflow (ID: {'G', 'D', 'O', '2' }) [Optional]
121 * This list of 8-byte values stores the corrected commit date offsets
122 for commits with corrected commit date offsets that cannot be
123 stored within 31 bits.
124 * Generation Data Overflow chunk is present only when Generation Data
125 chunk is present and atleast one corrected commit date offset cannot
126 be stored within 31 bits.
128 ==== Extra Edge List (ID: {'E', 'D', 'G', 'E'}) [Optional]
129 This list of 4-byte values store the second through nth parents for
130 all octopus merges. The second parent value in the commit data stores
131 an array position within this list along with the most-significant bit
132 on. Starting at that array position, iterate through this list of commit
133 positions for the parents until reaching a value with the most-significant
134 bit on. The other bits correspond to the position of the last parent.
136 ==== Bloom Filter Index (ID: {'B', 'I', 'D', 'X'}) (N * 4 bytes) [Optional]
137 * The ith entry, BIDX[i], stores the number of bytes in all Bloom filters
138 from commit 0 to commit i (inclusive) in lexicographic order. The Bloom
139 filter for the i-th commit spans from BIDX[i-1] to BIDX[i] (plus header
140 length), where BIDX[-1] is 0.
141 * The BIDX chunk is ignored if the BDAT chunk is not present.
143 ==== Bloom Filter Data (ID: {'B', 'D', 'A', 'T'}) [Optional]
144 * It starts with header consisting of three unsigned 32-bit integers:
145 - Version of the hash algorithm being used. We currently only support
146 value 1 which corresponds to the 32-bit version of the murmur3 hash
147 implemented exactly as described in
148 https://en.wikipedia.org/wiki/MurmurHash#Algorithm and the double
149 hashing technique using seed values 0x293ae76f and 0x7e646e2 as
150 described in https://doi.org/10.1007/978-3-540-30494-4_26 "Bloom Filters
151 in Probabilistic Verification"
152 - The number of times a path is hashed and hence the number of bit positions
153 that cumulatively determine whether a file is present in the commit.
154 - The minimum number of bits 'b' per entry in the Bloom filter. If the filter
155 contains 'n' entries, then the filter size is the minimum number of 64-bit
156 words that contain n*b bits.
157 * The rest of the chunk is the concatenation of all the computed Bloom
158 filters for the commits in lexicographic order.
159 * Note: Commits with no changes or more than 512 changes have Bloom filters
160 of length one, with either all bits set to zero or one respectively.
161 * The BDAT chunk is present if and only if BIDX is present.
163 ==== Base Graphs List (ID: {'B', 'A', 'S', 'E'}) [Optional]
164 This list of H-byte hashes describe a set of B commit-graph files that
165 form a commit-graph chain. The graph position for the ith commit in this
166 file's OID Lookup chunk is equal to i plus the number of commits in all
167 base graphs. If B is non-zero, this chunk must exist.
171 H-byte HASH-checksum of all of the above.
175 The Generation Data (GDA2) and Generation Data Overflow (GDO2) chunks have
176 the number '2' in their chunk IDs because a previous version of Git wrote
177 possibly erroneous data in these chunks with the IDs "GDAT" and "GDOV". By
178 changing the IDs, newer versions of Git will silently ignore those older
179 chunks and write the new information without trusting the incorrect data.
183 Part of the linkgit:git[1] suite