zesven 3.2.0

A pure Rust implementation of the 7z archive format
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245

//! Entries are compressed in batches so that independent folders can be built
//! at the same time. Batching is only sound if what comes out the far end is
//! indistinguishable from having compressed them one at a time: same entries,
//! same order, same bytes.

#![cfg(feature = "lzma2")]

use std::io::Cursor;

use zesven::ArchivePath;
use zesven::read::Archive;
use zesven::write::{EntryMeta, WriteOptions, Writer};

/// Data worth compressing, so entries take a real trip through the codec.
fn payload(seed: u64, len: usize) -> Vec<u8> {
    let mut data = Vec::with_capacity(len);
    let mut state = seed.wrapping_mul(0x9E37_79B9_7F4A_7C15) | 1;
    while data.len() < len {
        state ^= state << 13;
        state ^= state >> 7;
        state ^= state << 17;
        data.extend_from_slice(b"batched entry payload, repeated for the matcher ");
        data.extend_from_slice(&state.to_le_bytes());
    }
    data.truncate(len);
    data
}

/// Entries must appear in the order they were added, whatever kind they are.
///
/// Directories and anti-items are recorded the moment they are added, while
/// file entries wait in the batch, so adding one kind after the other is what
/// lets a batched file overtake the entry that was added before it.
#[test]
fn test_entry_order_survives_batching() {
    let data = payload(1, 4096);

    let mut writer = Writer::create(Cursor::new(Vec::new())).unwrap();
    writer
        .add_bytes(ArchivePath::new("01-file.bin").unwrap(), &data)
        .unwrap();
    writer
        .add_directory(ArchivePath::new("02-dir").unwrap(), EntryMeta::directory())
        .unwrap();
    writer
        .add_bytes(ArchivePath::new("03-file.bin").unwrap(), &data)
        .unwrap();
    writer
        .add_anti_item(ArchivePath::new("04-anti.bin").unwrap())
        .unwrap();
    writer
        .add_bytes(ArchivePath::new("05-file.bin").unwrap(), &data)
        .unwrap();
    writer
        .add_anti_directory(ArchivePath::new("06-anti-dir").unwrap())
        .unwrap();
    writer
        .add_bytes(ArchivePath::new("07-file.bin").unwrap(), &data)
        .unwrap();
    let (_result, cursor) = writer.finish_into_inner().unwrap();

    let archive = Archive::open(Cursor::new(cursor.into_inner())).unwrap();
    let paths: Vec<&str> = archive.entries().iter().map(|e| e.path.as_str()).collect();

    assert_eq!(
        paths,
        vec![
            "01-file.bin",
            "02-dir",
            "03-file.bin",
            "04-anti.bin",
            "05-file.bin",
            "06-anti-dir",
            "07-file.bin",
        ],
    );
}

/// Every entry must still decode to what was put in, across a batch boundary.
///
/// The batch is flushed once it holds enough data, so this writes more than one
/// batch worth and checks the entries either side of the boundary.
#[test]
fn test_entries_round_trip_across_batch_boundaries() {
    // The batch flushes at 64 MiB; 10 MiB apiece crosses that twice.
    const ENTRY: usize = 10 * 1024 * 1024;
    const COUNT: usize = 14;

    let mut writer = Writer::create(Cursor::new(Vec::new()))
        .unwrap()
        .options(WriteOptions::new().level(1).unwrap());
    for i in 0..COUNT {
        writer
            .add_bytes(
                ArchivePath::new(&format!("{i:02}.bin")).unwrap(),
                &payload(i as u64, ENTRY),
            )
            .unwrap();
    }
    let (_result, cursor) = writer.finish_into_inner().unwrap();

    let mut archive = Archive::open(Cursor::new(cursor.into_inner())).unwrap();
    assert_eq!(archive.entries().len(), COUNT);
    for i in 0..COUNT {
        let extracted = archive.extract_to_vec(&format!("{i:02}.bin")).unwrap();
        assert_eq!(
            extracted,
            payload(i as u64, ENTRY),
            "entry {i} did not round-trip",
        );
    }
}

/// Empty entries carry no stream, and must not shift the ones that do.
///
/// An empty entry is recorded in the file list but never becomes a folder, so
/// mixing them in is what exposes a folder index counted against the wrong list.
#[test]
fn test_empty_entries_keep_folders_aligned() {
    let first = payload(1, 8192);
    let second = payload(2, 8192);

    let mut writer = Writer::create(Cursor::new(Vec::new())).unwrap();
    writer
        .add_bytes(ArchivePath::new("empty-a.bin").unwrap(), b"")
        .unwrap();
    writer
        .add_bytes(ArchivePath::new("first.bin").unwrap(), &first)
        .unwrap();
    writer
        .add_bytes(ArchivePath::new("empty-b.bin").unwrap(), b"")
        .unwrap();
    writer
        .add_bytes(ArchivePath::new("second.bin").unwrap(), &second)
        .unwrap();
    writer
        .add_bytes(ArchivePath::new("empty-c.bin").unwrap(), b"")
        .unwrap();
    let (_result, cursor) = writer.finish_into_inner().unwrap();

    let mut archive = Archive::open(Cursor::new(cursor.into_inner())).unwrap();
    let paths: Vec<String> = archive
        .entries()
        .iter()
        .map(|e| e.path.as_str().to_string())
        .collect();
    assert_eq!(
        paths,
        vec![
            "empty-a.bin",
            "first.bin",
            "empty-b.bin",
            "second.bin",
            "empty-c.bin",
        ],
    );

    assert_eq!(archive.extract_to_vec("first.bin").unwrap(), first);
    assert_eq!(archive.extract_to_vec("second.bin").unwrap(), second);
    assert!(archive.extract_to_vec("empty-b.bin").unwrap().is_empty());
}

/// The same input must produce the same archive every time.
///
/// Entries are handed to whichever worker is free, so a writer that recorded
/// results in completion order rather than input order would still produce a
/// readable archive - a different one on each run.
#[test]
fn test_batched_output_is_deterministic() {
    let build = || {
        let mut writer = Writer::create(Cursor::new(Vec::new()))
            .unwrap()
            .options(WriteOptions::new().level(1).unwrap().deterministic(true));
        for i in 0..24u64 {
            writer
                .add_bytes(
                    ArchivePath::new(&format!("{i:02}.bin")).unwrap(),
                    &payload(i, 256 * 1024),
                )
                .unwrap();
        }
        writer.finish_into_inner().unwrap().1.into_inner()
    };

    assert_eq!(
        build(),
        build(),
        "the same entries produced two different archives",
    );
}

/// Neither the memory budget nor the thread count may change what is written.
///
/// Both decide how many entries a batch gathers and how many threads each of
/// them gets, and an entry long enough to be cut is cut into blocks by those
/// threads. If any of that reached the output, the same directory would give
/// one archive on a laptop and another on a build server, and a checksum of
/// the result would mean nothing.
///
/// Entries either side of the splitting threshold, so both the cut and the
/// uncut path are covered: at level 1 the dictionary is 1 MiB and a stream is
/// cut once it passes 1.25 MiB of it.
#[test]
fn test_the_machine_does_not_reach_the_archive() {
    use zesven::{MemoryLimit, Threads};

    let build = |limit: MemoryLimit, threads: Threads| {
        let mut writer = Writer::create(Cursor::new(Vec::new())).unwrap().options(
            WriteOptions::new()
                .level(1)
                .unwrap()
                .deterministic(true)
                .memory_limit(limit)
                .threads(threads),
        );
        for (i, len) in [4 << 20, 64 << 10, 3 << 20, 8 << 10, 5 << 20]
            .into_iter()
            .enumerate()
        {
            writer
                .add_bytes(
                    ArchivePath::new(&format!("{i:02}.bin")).unwrap(),
                    &payload(i as u64, len),
                )
                .unwrap();
        }
        writer.finish_into_inner().unwrap().1.into_inner()
    };

    // A budget that holds one encoder against one that holds a batch of them,
    // and two threads against eight: every combination has to agree.
    let cramped = build(
        MemoryLimit::bytes_or_auto(16 << 20),
        Threads::count_or_single(2),
    );
    let roomy = build(
        MemoryLimit::bytes_or_auto(4 << 30),
        Threads::count_or_single(8),
    );

    assert_eq!(
        cramped, roomy,
        "the budget or the thread count reached the bytes",
    );
}