//! Artifact Format — draft-fossil-artifact-format-00.
//!
//! An artifact is an immutable byte sequence named by the lowercase-hex hash
//! of its content (Section 3). Structural artifacts are line-oriented "cards".
//! This module owns the canonical vocabulary for the whole suite: naming, the
//! card grammar, the structural types, and the two checksums (Z, R).
//!
//! Scope note: zit targets the version-control substrate — the check-in
//! (manifest), cluster, and tag/control artifact types. The collaboration
//! types (wiki, ticket, forum, attachment, technote) are carried opaquely.
const std = @import("std");
/// Card-type letters. Only the version-control cards are interpreted; others
/// are recognised syntactically and otherwise carried opaquely.
pub const CardType = enum(u8) {
B = 'B', // baseline (delta manifest)
C = 'C', // comment
D = 'D', // date (mtime)
F = 'F', // file
M = 'M', // cluster member
N = 'N', // mimetype
P = 'P', // parent(s)
Q = 'Q', // cherrypick
R = 'R', // aggregate content checksum (MD5)
T = 'T', // tag
U = 'U', // user
Z = 'Z', // artifact self-checksum (MD5), always last
};
/// Structural artifact types in scope for the version-control substrate.
pub const ArtifactType = enum {
manifest, // a check-in
cluster, // a set of artifact hashes (sync housekeeping)
control, // a tag/branch mutation
other, // recognised but carried opaquely (wiki/ticket/forum/...)
};
/// Hash algorithm that names an artifact (Section 3, Section 7 hash policy).
/// The canonical (writer) form is always lowercase hex.
pub const HashAlgo = enum { sha1, sha3_256 };
/// Compute the artifact name: the lowercase-hex hash of `content`.
/// Writes into `out` and returns the populated slice (40 hex for SHA-1,
/// 64 hex for SHA3-256).
pub fn artifactName(algo: HashAlgo, content: []const u8, out: *[64]u8) []u8 {
switch (algo) {
.sha1 => {
var digest: [20]u8 = undefined;
std.crypto.hash.Sha1.hash(content, &digest, .{});
const hex = std.fmt.bytesToHex(digest, .lower);
@memcpy(out[0..hex.len], &hex);
return out[0..hex.len];
},
.sha3_256 => {
var digest: [32]u8 = undefined;
std.crypto.hash.sha3.Sha3_256.hash(content, &digest, .{});
const hex = std.fmt.bytesToHex(digest, .lower);
@memcpy(out[0..hex.len], &hex);
return out[0..hex.len];
},
}
}
/// Length-to-algorithm dispatch (§3.2-m1). The two name lengths are disjoint,
/// so length alone fully determines the algorithm: 40 hex digits is SHA-1,
/// 64 is SHA3-256, any other length names nothing.
pub fn nameAlgo(len: usize) ?HashAlgo {
return switch (len) {
40 => .sha1,
64 => .sha3_256,
else => null,
};
}
/// ABNF `lhexdig` (§3.5): a lowercase hex digit, 0-9 or a-f. Governs the
/// canonical (writer/parse) form only.
fn isHexLower(c: u8) bool {
return (c >= '0' and c <= '9') or (c >= 'a' and c <= 'f');
}
/// ABNF `hexdig-ci` (§3.5): 0-9, A-F, or a-f. Used only on the prefix-lookup
/// path (§3.6), which canonicalizes case; the parse path never uses this.
fn isHexCi(c: u8) bool {
return (c >= '0' and c <= '9') or (c >= 'a' and c <= 'f') or (c >= 'A' and c <= 'F');
}
/// Canonical-form validation (§3.2, §3.3, §3.5 `artifact-name`). A token is
/// valid iff it is exactly 40 or 64 lowercase hex digits. Length is checked
/// before character class (§3.2-r1: length gate precedes char-class check).
/// Uppercase digits are rejected outright, never case-normalized (§3.3-m2).
pub fn validateName(token: []const u8) bool {
if (nameAlgo(token.len) == null) return false;
std.debug.assert(token.len == 40 or token.len == 64);
var i: usize = 0;
while (i < token.len) : (i += 1) {
if (!isHexLower(token[i])) return false;
}
return true;
}
/// Verify `name` against `content` (§3.4). Dispatch on length picks the
/// digest to recompute; the compare is exact byte-for-byte lowercase hex.
/// Wrong length, non-hex characters, and a digest mismatch collapse into the
/// same `false` (§3.4-m1) — nothing here distinguishes them for the caller.
/// An uppercase name never verifies either, since the compare never
/// case-folds (§3.4-r1).
pub fn verify(name: []const u8, content: []const u8) bool {
const algo = nameAlgo(name.len) orelse return false;
var buf: [64]u8 = undefined;
const computed = artifactName(algo, content, &buf);
std.debug.assert(computed.len == name.len);
return std.mem.eql(u8, name, computed);
}
/// Prefix-resolution outcome (§3.6, informative — not part of the wire
/// format). `one` carries the matching index into the caller's `names`.
pub const Resolution = union(enum) { none, one: usize, ambiguous };
/// Resolve a hex name or prefix against a candidate universe (§3.6).
/// `prefix` must be 4-64 hex characters of any case; it is lowercased before
/// matching (the lookup path canonicalizes case, unlike the parse path —
/// see `validateName`). One query walk covers both name lengths, so a
/// cross-algorithm collision (one 40-char match, one 64-char match) reports
/// `.ambiguous` exactly like a same-algorithm collision (§3.6-r1).
pub fn resolvePrefix(prefix: []const u8, names: []const []const u8) error{InvalidPrefix}!Resolution {
if (prefix.len < 4 or prefix.len > 64) return error.InvalidPrefix;
std.debug.assert(prefix.len >= 4 and prefix.len <= 64);
var lowered: [64]u8 = undefined;
var i: usize = 0;
while (i < prefix.len) : (i += 1) {
const c = prefix[i];
if (!isHexCi(c)) return error.InvalidPrefix;
lowered[i] = if (c >= 'A' and c <= 'F') c + ('a' - 'A') else c;
}
const query = lowered[0..prefix.len];
var first: ?usize = null;
var matches: usize = 0;
var n: usize = 0;
while (n < names.len) : (n += 1) {
std.debug.assert(matches <= names.len);
if (std.mem.startsWith(u8, names[n], query)) {
matches += 1;
if (first == null) first = n;
if (matches > 1) return .ambiguous;
}
}
if (first) |idx| {
std.debug.assert(idx < names.len);
return .{ .one = idx };
}
return .none;
}
/// The Z-card value: lowercase-hex MD5 of all stripped-content bytes that
/// precede the `Z ` line (Section 6, Z card).
pub fn selfChecksum(stripped_prefix: []const u8, out: *[32]u8) []u8 {
var digest: [16]u8 = undefined;
std.crypto.hash.Md5.hash(stripped_prefix, &digest, .{});
const hex = std.fmt.bytesToHex(digest, .lower);
@memcpy(out, &hex);
return out;
}
// ---------------------------------------------------------------------------
// Manifest parsing (Sections 4, 5) — TODO
// ---------------------------------------------------------------------------
pub const Manifest = struct {
// Populated by parse(): cards, type, name. Skeleton for now.
kind: ArtifactType = .other,
};
pub const ParseError = error{ SyntaxError, NotImplemented };
/// Parse a structural artifact from its bytes. See Section 4 (card syntax),
/// Section 5 (types + per-type validity). Not yet implemented.
pub fn parse(allocator: std.mem.Allocator, bytes: []const u8) ParseError!Manifest {
_ = allocator;
_ = bytes;
return ParseError.NotImplemented;
}
// ---------------------------------------------------------------------------
// Tests
// ---------------------------------------------------------------------------
test "artifact §3.1: name is SHA3-256 of exact raw bytes" {
var out: [64]u8 = undefined;
// §A.3 / Table 8 (RFC line 2584): alpha.txt's content is "hello world\n",
// named by the SHA3-256 given as its F-card hash.
const name = artifactName(.sha3_256, "hello world\n", &out);
try std.testing.expectEqualStrings(
"a8009a7a528d87778c356da3a55d964719e818666a04e4f960c9e2439e35f138",
name,
);
}
// The §A.1.1 SHA-1 worked-example manifest: 127 raw bytes (27+22+2+35+6+35
// per the §A.2 Z-card offset table), literal backslash-s escapes retained
// exactly as printed in the RFC. Its stated name is
// a28c83647dfa805f05f3204a7e146eb1f0d90505 (RFC line 2426).
const a1_1_manifest = @embedFile("rfc_vectors/artifact_a1_1_sha1_manifest.txt");
const a1_1_name = "a28c83647dfa805f05f3204a7e146eb1f0d90505";
test "artifact §3.2: 40-hex name is SHA-1 over raw bytes" {
try std.testing.expectEqual(@as(usize, 127), a1_1_manifest.len);
var out: [64]u8 = undefined;
const name = artifactName(.sha1, a1_1_manifest, &out);
try std.testing.expectEqualStrings(a1_1_name, name);
}
test "artifact §3.2: name length selects algorithm" {
try std.testing.expectEqual(HashAlgo.sha1, nameAlgo(40).?);
try std.testing.expectEqual(HashAlgo.sha3_256, nameAlgo(64).?);
}
test "artifact §3.2: name of any other length is rejected" {
const bad_lens = [_]usize{ 0, 4, 39, 41, 63, 65, 128 };
for (bad_lens) |len| {
try std.testing.expect(nameAlgo(len) == null);
// Fabricate an all-'a' token of that length: rejection here must be
// on length alone, before any character-class check (§3.2-r1).
var buf: [128]u8 = undefined;
std.debug.assert(len <= buf.len);
@memset(buf[0..len], 'a');
try std.testing.expect(!validateName(buf[0..len]));
}
}
test "artifact §3.3: computed name is lowercase hex" {
var out: [64]u8 = undefined;
const sha1_name = artifactName(.sha1, a1_1_manifest, &out);
for (sha1_name) |c| try std.testing.expect(isHexLower(c));
var out2: [64]u8 = undefined;
const sha3_name = artifactName(.sha3_256, "hello world\n", &out2);
for (sha3_name) |c| try std.testing.expect(isHexLower(c));
}
test "artifact §3.3: validation rejects uppercase, does not normalize" {
// Same fixture as the rejection test above, restated to make the point
// explicit: a normalize-then-accept parser would flip this to true.
var upper: [40]u8 = undefined;
for (a1_1_name, 0..) |c, i| upper[i] = std.ascii.toUpper(c);
try std.testing.expect(!validateName(&upper));
// And the lowercase original still validates, proving the reject is
// case-specific rather than a blanket failure.
try std.testing.expect(validateName(a1_1_name));
}
test "artifact §3.4: verify dispatches on length" {
try std.testing.expect(verify(a1_1_name, a1_1_manifest));
try std.testing.expect(verify(
"a8009a7a528d87778c356da3a55d964719e818666a04e4f960c9e2439e35f138",
"hello world\n",
));
}
test "artifact §3.4: uppercase name never verifies" {
var upper: [40]u8 = undefined;
for (a1_1_name, 0..) |c, i| upper[i] = std.ascii.toUpper(c);
try std.testing.expect(!verify(&upper, a1_1_manifest));
}
test "artifact §3.4: one undifferentiated failure signal" {
// Wrong length.
try std.testing.expect(!verify("deadbeef10", a1_1_manifest));
// Non-hex character at a valid (40) length.
var non_hex: [40]u8 = undefined;
@memcpy(&non_hex, a1_1_name);
non_hex[0] = 'g';
try std.testing.expect(!verify(&non_hex, a1_1_manifest));
// Valid-form 40-hex name, wrong digest (mismatched content).
try std.testing.expect(!verify(a1_1_name, "hello world\n"));
}
test "artifact §3.5: ABNF canonical grammar accepts lhexdig only" {
var boundary: [40]u8 = undefined;
@memset(&boundary, '0');
try std.testing.expect(validateName(&boundary));
@memset(&boundary, '9');
try std.testing.expect(validateName(&boundary));
@memset(&boundary, 'a');
try std.testing.expect(validateName(&boundary));
@memset(&boundary, 'f');
try std.testing.expect(validateName(&boundary));
var has_g: [40]u8 = undefined;
@memset(&has_g, 'a');
has_g[39] = 'g';
try std.testing.expect(!validateName(&has_g));
var has_upper: [40]u8 = undefined;
@memset(&has_upper, 'a');
has_upper[0] = 'A';
try std.testing.expect(!validateName(&has_upper));
var sha3_all_f: [64]u8 = undefined;
@memset(&sha3_all_f, 'f');
try std.testing.expect(validateName(&sha3_all_f));
}
test "artifact §3.6: prefix resolves and lowercases uppercase input" {
const names = [_][]const u8{a1_1_name};
switch (try resolvePrefix("a28c", &names)) {
.one => |idx| try std.testing.expectEqual(@as(usize, 0), idx),
else => return error.TestUnexpectedResult,
}
switch (try resolvePrefix("A28C", &names)) {
.one => |idx| try std.testing.expectEqual(@as(usize, 0), idx),
else => return error.TestUnexpectedResult,
}
}
test "artifact §3.6: prefix length outside 4..64 is invalid" {
const names = [_][]const u8{a1_1_name};
try std.testing.expectError(error.InvalidPrefix, resolvePrefix("a28", &names));
var too_long: [65]u8 = undefined;
@memset(&too_long, 'a');
try std.testing.expectError(error.InvalidPrefix, resolvePrefix(&too_long, &names));
try std.testing.expectError(error.InvalidPrefix, resolvePrefix("zzzz", &names));
}
test "artifact §3.6: ambiguous prefix, incl. cross-algorithm collision" {
// Cross-algorithm: one SHA-1-named, one SHA3-named, sharing "a28c".
const cross = [_][]const u8{
a1_1_name,
"a28c000000000000000000000000000000000000000000000000000000000000",
};
try std.testing.expectEqual(Resolution.ambiguous, try resolvePrefix("a28c", &cross));
// Same-algorithm: two 40-char names sharing a prefix, reported the same.
const same = [_][]const u8{
a1_1_name,
"a28c000000000000000000000000000000000000",
};
try std.testing.expectEqual(Resolution.ambiguous, try resolvePrefix("a28c", &same));
}
test "self-checksum = lowercase MD5 (empty input vector)" {
var out: [32]u8 = undefined;
const z = selfChecksum("", &out);
try std.testing.expectEqualStrings("d41d8cd98f00b204e9800998ecf8427e", z);
}