1GITFORMAT-COMMIT-G(5) Git Manual GITFORMAT-COMMIT-G(5)
2
6 gitformat-commit-graph - Git commit-graph format
7
9 $GIT_DIR/objects/info/commit-graph
10 $GIT_DIR/objects/info/commit-graphs/*
11
13 The Git commit-graph stores a list of commit OIDs and some associated
14 metadata, including:
15 • The generation number of the commit.
17 • The root tree OID.
19 • The commit date.
21 • The parents of the commit, stored using positional references
23 within the graph file.
24 • The Bloom filter of the commit carrying the paths that were changed
26 between the commit and its first parent, if requested.
27 These positional references are stored as unsigned 32-bit integers
29 corresponding to the array position within the list of commit OIDs. Due
30 to some special constants we use to track parents, we can store at most
31 (1 << 30) + (1 << 29) + (1 << 28) - 1 (around 1.8 billion) commits.
32
34 In order to allow extensions that add extra data to the graph, we
35 organize the body into "chunks" and provide a binary lookup table at
36 the beginning of the body. The header includes certain values, such as
37 number of chunks and hash type.
38 All multi-byte numbers are in network byte order.
40 HEADER:
42 4-byte signature:
43 The signature is: {'C', 'G', 'P', 'H'}
44 1-byte version number:
46 Currently, the only valid version is 1.
47 1-byte Hash Version
49 We infer the hash length (H) from this value:
50 1 => SHA-1
51 2 => SHA-256
52 If the hash type does not match the repository's hash algorithm, the
53 commit-graph file should be ignored with a warning presented to the
54 user.
55 1-byte number (C) of "chunks"
57 1-byte number (B) of base commit-graphs
59 We infer the length (H*B) of the Base Graphs chunk
60 from this value.
61 CHUNK LOOKUP:
63 (C + 1) * 12 bytes listing the table of contents for the chunks:
64 First 4 bytes describe the chunk id. Value 0 is a terminating label.
65 Other 8 bytes provide the byte-offset in current file for chunk to
66 start. (Chunks are ordered contiguously in the file, so you can infer
67 the length using the next chunk position if necessary.) Each chunk
68 ID appears at most once.
69 The CHUNK LOOKUP matches the table of contents from
71 the chunk-based file format, see linkgit:gitformat-chunk[5]
72 The remaining data in the body is described one chunk at a time, and
74 these chunks may be given in any order. Chunks are required unless
75 otherwise specified.
76 CHUNK DATA:
78 OID Fanout (ID: {O, I, D, F}) (256 * 4 bytes)
79 The ith entry, F[i], stores the number of OIDs with first
80 byte at most i. Thus F[255] stores the total
81 number of commits (N).
82 OID Lookup (ID: {O, I, D, L}) (N * H bytes)
84 The OIDs for all commits in the graph, sorted in ascending order.
85 Commit Data (ID: {C, D, A, T }) (N * (H + 16) bytes)
87 • The first H bytes are for the OID of the root tree.
88 • The next 8 bytes are for the positions of the first two parents
90 of the ith commit. Stores value 0x70000000 if no parent in that
91 position. If there are more than two parents, the second value
92 has its most-significant bit on and the other bits store an
93 array position into the Extra Edge List chunk.
94 • The next 8 bytes store the topological level (generation number
96 v1) of the commit and the commit time in seconds since EPOCH.
97 The generation number uses the higher 30 bits of the first 4
98 bytes, while the commit time uses the 32 bits of the second 4
99 bytes, along with the lowest 2 bits of the lowest byte, storing
100 the 33rd and 34th bit of the commit time.
101 Generation Data (ID: {G, D, A, 2 }) (N * 4 bytes) [Optional]
103 • This list of 4-byte values store corrected commit date offsets
104 for the commits, arranged in the same order as commit data
105 chunk.
106 • If the corrected commit date offset cannot be stored within 31
108 bits, the value has its most-significant bit on and the other
109 bits store the position of corrected commit date into the
110 Generation Data Overflow chunk.
111 • Generation Data chunk is present only when commit-graph file is
113 written by compatible versions of Git and in case of split
114 commit-graph chains, the topmost layer also has Generation Data
115 chunk.
116 Generation Data Overflow (ID: {G, D, O, 2 }) [Optional]
118 • This list of 8-byte values stores the corrected commit date
119 offsets for commits with corrected commit date offsets that
120 cannot be stored within 31 bits.
121 • Generation Data Overflow chunk is present only when Generation
123 Data chunk is present and atleast one corrected commit date
124 offset cannot be stored within 31 bits.
125 Extra Edge List (ID: {E, D, G, E}) [Optional]
127 This list of 4-byte values store the second through nth parents for
128 all octopus merges. The second parent value in the commit data stores
129 an array position within this list along with the most-significant bit
130 on. Starting at that array position, iterate through this list of commit
131 positions for the parents until reaching a value with the most-significant
132 bit on. The other bits correspond to the position of the last parent.
133 Bloom Filter Index (ID: {B, I, D, X}) (N * 4 bytes) [Optional]
135 • The ith entry, BIDX[i], stores the number of bytes in all Bloom
136 filters from commit 0 to commit i (inclusive) in lexicographic
137 order. The Bloom filter for the i-th commit spans from
138 BIDX[i-1] to BIDX[i] (plus header length), where BIDX[-1] is 0.
139 • The BIDX chunk is ignored if the BDAT chunk is not present.
141 Bloom Filter Data (ID: {B, D, A, T}) [Optional]
143 • It starts with header consisting of three unsigned 32-bit
144 integers:
145 • Version of the hash algorithm being used. We currently only
147 support value 1 which corresponds to the 32-bit version of
148 the murmur3 hash implemented exactly as described in
149 https://en.wikipedia.org/wiki/MurmurHash#Algorithm and the
150 double hashing technique using seed values 0x293ae76f and
151 0x7e646e2 as described in
152 https://doi.org/10.1007/978-3-540-30494-4_26 "Bloom Filters
153 in Probabilistic Verification"
154 • The number of times a path is hashed and hence the number
156 of bit positions that cumulatively determine whether a file
157 is present in the commit.
158 • The minimum number of bits b per entry in the Bloom filter.
160 If the filter contains n entries, then the filter size is
161 the minimum number of 64-bit words that contain n*b bits.
162 • The rest of the chunk is the concatenation of all the computed
164 Bloom filters for the commits in lexicographic order.
165 • Note: Commits with no changes or more than 512 changes have
167 Bloom filters of length one, with either all bits set to zero
168 or one respectively.
169 • The BDAT chunk is present if and only if BIDX is present.
171 Base Graphs List (ID: {B, A, S, E}) [Optional]
173 This list of H-byte hashes describe a set of B commit-graph files that
174 form a commit-graph chain. The graph position for the ith commit in this
175 file's OID Lookup chunk is equal to i plus the number of commits in all
176 base graphs. If B is non-zero, this chunk must exist.
177 TRAILER:
179 H-byte HASH-checksum of all of the above.
180
182 The Generation Data (GDA2) and Generation Data Overflow (GDO2) chunks
183 have the number 2 in their chunk IDs because a previous version of Git
184 wrote possibly erroneous data in these chunks with the IDs "GDAT" and
185 "GDOV". By changing the IDs, newer versions of Git will silently ignore
186 those older chunks and write the new information without trusting the
187 incorrect data.
188
190 Part of the git(1) suite
191Git 2.39.1 2023-01-13 GITFORMAT-COMMIT-G(5)