Class FileHashService

Namespace

Archiver.Core.Services

Assembly

Archiver.Core.dll

T-F128: computes CRC-32/SHA-256 for the Explorer context menu's "Хеш-суми" submenu. A single folder gets NanaZip-compatible combined DataSum (all file contents) and NamesSum (all file names+paths+contents) values, via Archiver.Core.IO.HashDigestAccumulator — the algorithm of 7-Zip's HashCalc.cpp, checked live against the vendored 7za h by FolderHashParityTests (T-F225). NamesSum includes one item per directory, the selected folder itself included, hashed with an all-zero digest (7-Zip resets it before every item), so the result does not depend on enumeration order. Remaining difference: symbolic links and junctions are skipped here (T-F251), where 7-Zip follows them.

Files are hashed in parallel (ForEachAsync<TSource>(IEnumerable<TSource>, ParallelOptions, Func<TSource, CancellationToken, ValueTask>), up to ProcessorCount at once) — safe specifically because Add(ReadOnlySpan<byte>) is commutative, so combining DataSum/NamesSum in whatever order files finish hashing produces the exact same result as combining them sequentially (already relied on for the recursion-safety argument above).

T-F128 follow-up: a single large CRC-32 file is also hashed in parallel — the across-files parallelism above gives no benefit to a folder containing one huge file (or to a lone large file passed to ComputeAsync(IReadOnlyList<string>, HashAlgorithmKind, IProgress<ProgressReport>?, CancellationToken) directly). Files at or above Archiver.Core.Services.FileHashService.ParallelCrc32MinFileBytes are split into independently-hashed chunks and folded back together with Combine(uint, uint, long) — safe for the same reason cross-file combining is: CRC-32 combining is associative/order-preserving as long as chunks are folded back in their original byte order (unlike DataSum/NamesSum, chunk order here does matter, so chunks are combined sequentially by index after all finish, not as they complete).

public static class FileHashService

Inheritance

object

FileHashService

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Methods

ComputeAsync(IReadOnlyList<string>, HashAlgorithmKind, IProgress<ProgressReport>?, CancellationToken)

Hashes each of paths. A single-folder selection is routed to the recursive DataSum/NamesSum path instead (see Folder); a mixed or multi-item selection hashes each file independently in parallel.

public static Task<HashResult> ComputeAsync(IReadOnlyList<string> paths, HashAlgorithmKind algorithm, IProgress<ProgressReport>? progress, CancellationToken ct)

Parameters

paths IReadOnlyList<string>

algorithm HashAlgorithmKind

progress IProgress<ProgressReport>

ct CancellationToken

Returns

Task<HashResult>

ComputeStreamDigestAsync(Stream, HashAlgorithmKind, CancellationToken)

T-F09 follow-up (pakko h -si): hashes an arbitrary stream directly — a genuine single-pass pipeline, unlike x/t/l's -si (which must stage stdin to a real seekable temp file first, since ZIP central-directory reads and tar.exe's pre-scan can't operate on a raw pipe). CRC-32/SHA-256 need no seeking, so no staging is needed here either — bytes are read once, straight from source, with no intermediate file.

public static Task<string> ComputeStreamDigestAsync(Stream source, HashAlgorithmKind algorithm, CancellationToken ct)

Parameters

source Stream

algorithm HashAlgorithmKind

ct CancellationToken

Returns

Task<string>