Press n or j to go to the next uncovered block, b, p or k for the previous block.
| 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 | 1x 1x 1x 1x 1x 1x 1x 1x 1280x 1280x 1280x 1280x 1280x 1280x 1x 1x 1279x 1279x 1279x 1279x 1279x 1279x 1279x 1279x 1279x 756x 540x 540x 1280x 537x 537x 537x 537x 537x 537x 537x 537x 537x 731x 731x 731x 731x 731x 1253x 490x 490x 490x 490x 490x 490x 490x 490x 731x 731x 731x 731x 731x 1261x 1280x 46x 46x 46x 1280x 1280x 1280x 1280x 1280x 730x 730x 89x 89x 89x 89x 89x 12x 12x 12x 89x 730x 731x 731x 731x 731x 731x 731x 731x 731x 731x 1280x 1280x 731x 731x 731x 731x 731x 731x 731x 731x 731x 731x 731x 731x 731x 731x 1280x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 730x 730x 730x 730x 730x 730x 730x 730x 730x 730x 730x 730x 730x 730x 727x 727x 1280x 439x 439x 439x 727x 727x 727x 727x 727x 727x 727x 727x 727x 727x 727x 727x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 4x 4x 4x 4x 4x 4x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 4x 4x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 3x 2x 4x 4x 4x 4x 4x 4x 4x 4x 4x 3x 3x 1468x 1468x 1468x 1468x 1468x 1468x 1468x 1468x 13926x 13926x 13926x 2440x 198x 198x 198x 198x 198x 198x 198x 2440x 2440x 13926x 13926x 13926x 107x 107x 107x 107x 107x 107x 107x 107x 107x 13926x 6902x 347x 347x 347x 347x 347x 347x 347x 6902x 6902x 13926x 577x 577x 577x 577x 577x 577x 577x 577x 577x 13926x 1468x 10909x 29x 29x 29x 29x 29x 29x 29x 1468x 1468x 10909x 4234x 4226x 4226x 1468x 1468x 1468x 1468x 1468x 1468x 1x 1x 12x 12x 1x 12x 12x 12x 12x 12x 12x 1x 2566x 2566x 2566x 2566x 2566x 2566x 2566x 2566x 2566x 1283x 1283x 1283x 1283x 2566x 1x 1x 1x 1x 2566x 2566x 2566x 2566x 1283x 2x 2x 2x 2x 2x 2x 2x 1283x 10x 9x 9x 9x 9x 9x 9x 9x 10x 1283x 1283x 1283x 1271x 1271x 1271x 1271x 1271x 1271x 1271x 1268x 1268x 1268x 1268x 1080x 1080x 1080x 11420x 11420x 11420x 5x 5x 1080x 5x 5x 5x 5x 5x 5x 5x 2202x 2202x 2202x 19743x 19743x 6522x 6522x 2202x 2202x 1468x 1468x 1468x 1468x 1468x 1468x 1468x 1468x 577x 577x 524x 516x 577x 1468x 1468x 1468x 1468x 1468x 670x 34x 1468x 1468x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 1264x 734x 734x 734x 4875x 4875x 4875x 291x 291x 291x 733x 733x 733x 102x 102x 102x 102x 102x 733x 733x 178x 178x 178x 150x 28x 28x 178x 178x 493x 493x 493x 493x 167x 167x 734x 734x 734x 734x 734x 734x 734x 734x 734x 734x 102x 102x 23x 23x 102x 23x 102x 10x 13x 102x 23x 102x 102x 102x 102x 10x 10x 10x 102x 2x 2x 2x 2x 2x 2x 2x 2x 2x 23x 4x 4x 4x 102x 102x 102x 102x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 14x 102x 102x 102x 102x 102x 734x 734x 46x 46x 46x 46x 46x 1460x 1460x 1460x 1460x 1460x 1460x 1460x 519x 519x 8x 8x 8x 8x 8x 8x 511x 513x 513x 509x 509x 509x 2x 519x 1460x 1460x 384x 384x 384x 1113x 1113x 1113x 3x 3x 1113x 384x 384x 145x 145x 145x 145x 145x 145x 145x 383x 383x 383x 383x 383x 383x | import { decodeText, encodeText, randomId, sha256 } from './hash.js';
import { History } from './history.js';
import { MAX_TEXT_MERGE, diff3 } from './diff3.js';
import { makeRow, missingRows } from './log.js';
import { HELD_AT, mergeEntries } from './merge.js';
import type {
ConflictCopyPolicy,
ConflictNameInfo,
ConflictReport,
MergeOptions,
} from './merge.js';
import { CURRENT_VERSION, extensionOf, readable } from './vfs-file.js';
import { holds, materialised } from './vfs-node.js';
import type { ContentHandle, ContentSource, VFSNode } from './vfs-node.js';
import { stateDigest } from './vfs-file.js';
import type { EntryKind, Hash, LogRow, VFSEntry, VFSFile } from './types.js';
export interface TextConflictInfo {
path: string;
/** Common ancestor, when one was found. */
base: string | null;
a: string;
b: string;
peerA: string;
peerB: string;
}
export interface SyncOptions {
conflictCopies?: ConflictCopyPolicy;
conflictName?: (info: ConflictNameInfo) => string;
now?: () => number;
/** Size from which a conflict copy stays on the peer that made it (§4). */
heldAt?: number;
/** Turn off the automatic three-way merge of text files. */
autoMerge?: boolean;
/**
* Whether to open a closed log segment when a conflict's loser is older than
* the active one. Defaults to `true`. Turning it off trades a cold read for
* the occasional conflict copy that need not have been kept — never for a
* different outcome.
*/
archives?: boolean;
/**
* Interactive resolution for a text conflict. Returning content settles it;
* returning `null` — or leaving the hook out — takes the headless path of
* last-writer-wins, a copy, and a pending decision.
*/
resolveText?: (info: TextConflictInfo) => Promise<string | null>;
/**
* Called after planning and before any cross-peer writes. Return `false` to
* abort this sync pass.
*/
approveMerge?: (preview: SyncDryRunResult) => Promise<boolean> | boolean;
/**
* Authorises merging two folders the pairing guard stopped, by naming one of
* the two `syncId`s the {@link PairingError} reported.
*
* Specific on purpose: a blanket `true` would disarm the guard at this call
* site forever, including a different collision months later. A `peerId`
* collision is **not** authorisable this way — merging two nodes with one
* identity is not a decision anyone can make well, and the remedy lies
* outside sync.
*/
adopt?: { syncId: string };
}
export interface SyncResult {
/** False when both folders were already identical. */
changed: boolean;
/** False when an approval hook vetoed the merge before writes. */
approved?: boolean;
conflicts: ConflictReport[];
/** Content copies performed in each direction. */
transferred: { toA: number; toB: number };
/** Text conflicts settled by a three-way merge instead of a copy. */
merged: number;
/** The digest both peers end on, or `null` when the pair is empty. */
state: Hash | null;
}
export type SyncDryRunActionType = 'write' | 'delete' | 'rename' | 'mkdir';
export interface SyncDryRunAction {
type: SyncDryRunActionType;
uuid: string;
kind: EntryKind;
path: string;
/** Present on `write` actions that create a new file on that side. */
created?: boolean;
from?: string;
to?: string;
}
export interface SyncDryRunResult {
/** False when the final `sync()` call would be a no-op. */
changed: boolean;
/** Whether `sync()` would first converge the `text` config. */
configChanged: boolean;
conflicts: ConflictReport[];
/** Predicted content copies performed in each direction. */
transferred: { toA: number; toB: number };
/** Text conflicts that would settle via three-way merge. */
merged: number;
/** Paths that would be settled by text auto-merge. */
mergedPaths: string[];
/** Predicted file-system actions `sync()` would perform. */
actions: { toA: SyncDryRunAction[]; toB: SyncDryRunAction[] };
/** The digest both peers would end on after `sync()`. */
state: Hash | null;
}
/**
* Syncs one edge of the mesh (§5). No ancestor negotiation: both sides are
* reconciled, one `state` digest is compared, and nothing else happens unless it
* differs.
*
* Nothing is assumed about who else either peer talks to, which is what lets
* changes travel down a chain (A <-> B <-> C) one edge at a time — and the log
* is deliberately not per-peer, so A also carries B's operations when it meets
* C, and both sides decide conflicts from the same information.
*/
export async function sync(a: VFSNode, b: VFSNode, options: SyncOptions = {}): Promise<SyncResult> {
const now = options.now ?? (() => Date.now());
const nothing: SyncResult['transferred'] = { toA: 0, toB: 0 };
// The mirror is only as good as its last reconciliation, and a scan is
// obligatory here anyway: it is what turns disk state into entries.
await a.commit();
await b.commit();
if (a === b) {
return { changed: false, conflicts: [], transferred: nothing, merged: 0, state: await a.state() };
}
const fileA = await a.file();
const fileB = await b.file();
// ---- 2. may these two folders merge at all? Nothing of the merge has been
// written yet, which is the property that makes throwing safe here.
const syncId = pair(fileA, fileB, options.adopt);
// ---- 3. config converges: `text` by union
const config = convergeConfig(fileA, fileB);
const configChanged = config.changed;
applyConfig(fileA, fileB, config);
// ---- 4. one comparison decides whether there is anything to do at all
const at = now();
if (
fileA.state === fileB.state &&
fileA.log.digest === fileB.log.digest &&
!refillable(a, fileA, fileB) &&
!refillable(b, fileB, fileA)
) {
await close(a, b, fileA, fileB, [], [], at, syncId);
return {
changed: configChanged,
conflicts: [],
transferred: nothing,
merged: 0,
state: fileA.state,
};
}
// ---- 5/6. entries, and the log only where the entries cannot answer alone
const sources: Array<Iterable<LogRow> | Iterable<VFSEntry>> = [fileA.entries, fileB.entries];
let rowsA: LogRow[] = [];
let rowsB: LogRow[] = [];
// Each side's *own* knowledge, which is what the path fallback turns on: the
// shared history below is the union of both and would vouch for everything.
const ownA = History.from([fileA.entries]);
const ownB = History.from([fileB.entries]);
if (needsLog(fileA.entries, fileB.entries)) {
rowsA = await a.store.logRows();
rowsB = await readPeerLog(b, fileA.peers[b.peerId]);
const snapA = await a.store.readSnapshot(fileA);
const snapB = await b.store.readSnapshot(fileB);
ownA.add(rowsA).add(snapA);
ownB.add(rowsB).add(snapB);
sources.push(rowsA, rowsB, snapA, snapB);
}
const history = History.from(sources);
const sides = { peerId: a.peerId, entries: fileA.entries, knows: (uuid: string) => ownA.knows(uuid) };
const other = { peerId: b.peerId, entries: fileB.entries, knows: (uuid: string) => ownB.knows(uuid) };
// ---- merge
const mergeOptions: MergeOptions = {
history,
heldAt: options.heldAt ?? HELD_AT,
text: (path) => {
const extension = extensionOf(path);
return extension !== '' && config.text.includes(extension);
},
...(options.conflictCopies !== undefined ? { conflictCopies: options.conflictCopies } : {}),
...(options.conflictName ? { conflictName: options.conflictName } : {}),
};
let merge = mergeEntries(sides, other, mergeOptions);
// A conflict whose loser predates the active segment may not be one at all:
// the link that would prove ancestry is in a closed archive. Reading it is the
// one cold read in the protocol, and it is optional by design — it can only
// ever *avoid* a conflict copy, never change what the state ends up being.
if (options.archives !== false) {
const oldest = coldest(merge.conflicts);
if (oldest !== null) {
const rows = [
...(await readArchives(a, fileA, oldest)),
...(await readArchives(b, fileB, oldest)),
];
if (rows.length > 0) {
history.add(rows);
merge = mergeEntries(sides, other, mergeOptions);
}
}
}
// ---- text: try to settle a content conflict rather than park a copy
const overlay = new Map<Hash, Uint8Array>();
const extra: Array<Omit<LogRow, 'op'>> = [];
const batch = randomId();
const merged = await autoMergeText(a, b, merge, history, {
overlay,
rows: extra,
batch,
at,
enabled: options.autoMerge !== false,
...(options.resolveText ? { resolveText: options.resolveText } : {}),
});
const mergedCount = merged.count;
const mergedPaths = merged.paths;
const target = merge.entries;
const planA = planChanges(fileA.entries, target, a);
const planB = planChanges(fileB.entries, target, b);
const overlayHashes = new Set(overlay.keys());
const remoteFromB = new Set(fileHashes(fileB.entries, b));
const remoteForB = new Set([...fileHashes(target, a), ...fileHashes(fileA.entries, a)]);
const previewTransferred = {
toA: countTransfers(planA.writes, planA.localHashes, overlayHashes, remoteFromB),
toB: countTransfers(planB.writes, planB.localHashes, overlayHashes, remoteForB),
};
const previewState = await stateDigest(target);
const approve = options.approveMerge;
if (approve) {
const approved = await approve({
changed:
configChanged ||
merge.conflicts.length > 0 ||
mergedCount > 0 ||
planA.actions.length > 0 ||
planB.actions.length > 0,
configChanged,
conflicts: merge.conflicts,
transferred: previewTransferred,
merged: mergedCount,
mergedPaths,
actions: { toA: planA.actions, toB: planB.actions },
state: previewState,
});
if (!approved) {
return {
changed: false,
approved: false,
conflicts: merge.conflicts,
transferred: { toA: 0, toB: 0 },
merged: mergedCount,
state: null,
};
}
}
// ---- 6. content moves, verified on arrival
const transferred = { toA: 0, toB: 0 };
const beforeB = fileB.entries;
await a.apply(target, chain(overlay, [{ node: b, entries: beforeB }], () => transferred.toA++));
await b.apply(
target,
chain(
overlay,
[
{ node: a, entries: target },
{ node: a, entries: fileA.entries },
],
() => transferred.toB++,
),
);
// ---- 7. close: content is on disk, then the logs, then the two headers
const rowsForA = [...(await Promise.all(extra.map(makeRow)))];
const rowsForB = [...rowsForA];
if (rowsA.length > 0 || rowsB.length > 0) {
rowsForA.push(...missingRows(rowsA, rowsB));
rowsForB.push(...missingRows(rowsB, rowsA));
}
await a.adopt(target, fileA);
await b.adopt(target, fileB);
await close(a, b, fileA, fileB, rowsForA, rowsForB, at, syncId);
return {
changed: true,
approved: true,
conflicts: merge.conflicts,
transferred,
merged: mergedCount,
state: previewState,
};
}
/**
* Computes what `sync(a, b)` would do, without writing either folder.
*
* It still scans both sides (like `sync` does through `commit`) so the preview
* includes pending local edits, but it appends no log rows and performs no
* content writes.
*/
export async function syncDryRun(
a: VFSNode,
b: VFSNode,
options: SyncOptions = {},
): Promise<SyncDryRunResult> {
if (a === b) {
return {
changed: false,
configChanged: false,
conflicts: [],
transferred: { toA: 0, toB: 0 },
merged: 0,
mergedPaths: [],
actions: { toA: [], toB: [] },
state: await a.state(),
};
}
const now = options.now ?? (() => Date.now());
const at = now();
const [scanA, scanB] = await Promise.all([a.scan(), b.scan()]);
const fileA = await a.file();
const fileB = await b.file();
// The guard runs here too, and this is the healthy order: ask first, do not
// rescue afterwards. Nothing has been written, so a caller can discover a
// `foreign-mesh` without either folder having been touched.
pair(fileA, fileB, options.adopt);
const entriesA = scanA.entries;
const entriesB = scanB.entries;
// Computed, never applied: a dry run must leave both folders untouched.
const { text, changed: configChanged } = convergeConfig(fileA, fileB);
const sources: Array<Iterable<LogRow> | Iterable<VFSEntry>> = [entriesA, entriesB, scanA.rows, scanB.rows];
let rowsA: LogRow[] = [];
let rowsB: LogRow[] = [];
const ownA = History.from([entriesA, scanA.rows]);
const ownB = History.from([entriesB, scanB.rows]);
if (needsLog(entriesA, entriesB)) {
rowsA = await a.store.logRows();
rowsB = await readPeerLog(b, fileA.peers[b.peerId]);
const snapA = await a.store.readSnapshot(fileA);
const snapB = await b.store.readSnapshot(fileB);
ownA.add(rowsA).add(snapA);
ownB.add(rowsB).add(snapB);
sources.push(rowsA, rowsB, snapA, snapB);
}
const history = History.from(sources);
const sides = { peerId: a.peerId, entries: entriesA, knows: (uuid: string) => ownA.knows(uuid) };
const other = { peerId: b.peerId, entries: entriesB, knows: (uuid: string) => ownB.knows(uuid) };
const mergeOptions: MergeOptions = {
history,
heldAt: options.heldAt ?? HELD_AT,
text: (path) => {
const extension = extensionOf(path);
return extension !== '' && text.includes(extension);
},
...(options.conflictCopies !== undefined ? { conflictCopies: options.conflictCopies } : {}),
...(options.conflictName ? { conflictName: options.conflictName } : {}),
};
let merge = mergeEntries(sides, other, mergeOptions);
if (options.archives !== false) {
const oldest = coldest(merge.conflicts);
if (oldest !== null) {
const rows = [
...(await readArchives(a, fileA, oldest)),
...(await readArchives(b, fileB, oldest)),
];
if (rows.length > 0) {
history.add(rows);
merge = mergeEntries(sides, other, mergeOptions);
}
}
}
const overlay = new Map<Hash, Uint8Array>();
const merged = await autoMergeText(a, b, merge, history, {
overlay,
rows: [],
batch: randomId(),
at,
enabled: options.autoMerge !== false,
...(options.resolveText ? { resolveText: options.resolveText } : {}),
});
const mergedCount = merged.count;
const mergedPaths = merged.paths;
const target = merge.entries;
const planA = planChanges(entriesA, target, a);
const planB = planChanges(entriesB, target, b);
const overlayHashes = new Set(overlay.keys());
const remoteFromB = new Set(fileHashes(entriesB, b));
const remoteForB = new Set([...fileHashes(target, a), ...fileHashes(entriesA, a)]);
const transferred = {
toA: countTransfers(planA.writes, planA.localHashes, overlayHashes, remoteFromB),
toB: countTransfers(planB.writes, planB.localHashes, overlayHashes, remoteForB),
};
const changed =
configChanged ||
merge.conflicts.length > 0 ||
mergedCount > 0 ||
planA.actions.length > 0 ||
planB.actions.length > 0;
return {
changed,
configChanged,
conflicts: merge.conflicts,
transferred,
merged: mergedCount,
mergedPaths,
actions: { toA: planA.actions, toB: planB.actions },
state: await stateDigest(target),
};
}
interface PlannedChanges {
actions: SyncDryRunAction[];
writes: VFSEntry[];
localHashes: Set<Hash>;
}
function planChanges(currentEntries: VFSEntry[], target: VFSEntry[], node: VFSNode): PlannedChanges {
const nodeId = node.peerId;
const current = new Map(currentEntries.map((entry) => [entry.uuid, entry]));
const currentLive = currentEntries.filter((entry) => !entry.deleted);
const targetByUuid = new Set(target.map((entry) => entry.uuid));
const actions: SyncDryRunAction[] = [];
const writes: VFSEntry[] = [];
for (const entry of target) {
const before = current.get(entry.uuid);
const wasLive = !!before && !before.deleted;
if (entry.deleted) {
if (wasLive && before) {
actions.push({
type: 'delete',
uuid: before.uuid,
kind: before.kind,
path: before.path,
});
}
continue;
}
if (entry.held && entry.held !== nodeId) continue;
// The same rule `apply()` applies, and it has to be the same one or the
// dry run describes a sync that will not happen: the policy decides what
// arrives, never what is already on disk.
if (!node.wants(entry) && !materialised(before)) continue;
if (wasLive && before && before.path !== entry.path) {
actions.push({
type: 'rename',
uuid: entry.uuid,
kind: entry.kind,
path: entry.path,
from: before.path,
to: entry.path,
});
}
if (entry.kind === 'directory') {
if (!wasLive) {
actions.push({
type: 'mkdir',
uuid: entry.uuid,
kind: 'directory',
path: entry.path,
});
}
continue;
}
if (!wasLive || before?.hash !== entry.hash || !materialised(before)) {
writes.push(entry);
actions.push({
type: 'write',
uuid: entry.uuid,
kind: 'file',
path: entry.path,
...(wasLive ? {} : { created: true }),
});
}
}
for (const entry of currentLive) {
if (targetByUuid.has(entry.uuid)) continue;
actions.push({
type: 'delete',
uuid: entry.uuid,
kind: entry.kind,
path: entry.path,
});
}
const localHashes = new Set<Hash>();
for (const entry of currentLive) {
if (entry.kind !== 'file' || !entry.hash || entry.held) continue;
if (!materialised(entry)) continue; // the entry is here, the bytes are not
localHashes.add(entry.hash);
}
return {
actions: sortActions(actions),
writes,
localHashes,
};
}
export type PairingCode = 'version-unreconcilable' | 'peer-collision' | 'foreign-mesh';
/** Enough of one side to frame the decision the library refuses to make. */
export interface PairingSide {
peerId: string;
syncId: string | null;
version: number;
/** Live entries, so a caller can say "1,240 files against 890". */
entries: number;
log: { segment: number; digest: Hash };
}
/**
* Two folders that must not be merged without someone saying so.
*
* The engine's other errors are strings for humans. A caller cannot build a
* decision out of one of those: it needs the case, and the data to frame it.
*/
export class PairingError extends Error {
readonly code: PairingCode;
readonly a: PairingSide;
readonly b: PairingSide;
constructor(code: PairingCode, message: string, a: PairingSide, b: PairingSide) {
super(message);
this.name = 'PairingError';
this.code = code;
this.a = a;
this.b = b;
}
}
function sideOf(file: VFSFile): PairingSide {
return {
peerId: file.peerId,
syncId: file.syncId,
version: file.version,
entries: file.entries.filter((entry) => !entry.deleted).length,
log: { segment: file.log.segment, digest: file.log.digest },
};
}
/**
* Whether these two folders may merge at all, and which group id survives if
* they may.
*
* Two checks, in this order and not the other:
*
* 1. **version** — can I interpret this file? Reading `peerId` and `syncId`
* means nothing until you know which version wrote them, so this comes
* first. `decodeVFSFile` has already migrated everything it can, so a file
* still short of the current version is one no `migrate` covers.
* 2. **identity** — should I merge with this group?
*
* The benign `syncId` combinations are not errors and are what ordinary use
* looks like: equal (every sync after the first), both `null` (two virgin
* folders, mint one), and one set against one `null` (a folder joining a
* group — it has no affiliation to lose, so no tiebreak is needed).
*
* The library detects, describes and stops. It does not ask, and it does not
* decide: that is coherent with not interpreting content, and the question is
* not "which folder wins" — there is no such mode. Resolution stays per file
* and per version, with ancestry above the clock. The only thing decided here
* is **whether to merge at all**.
*/
function pair(fileA: VFSFile, fileB: VFSFile, adopt?: { syncId: string }): string {
const a = sideOf(fileA);
const b = sideOf(fileB);
for (const side of [a, b]) {
if (side.version === CURRENT_VERSION) continue;
throw new PairingError(
'version-unreconcilable',
readable(side.version)
? `a version ${side.version} folder did not migrate to ${CURRENT_VERSION}`
: `version ${side.version} cannot be read by an engine that writes ${CURRENT_VERSION}`,
a,
b,
);
}
if (a.peerId === b.peerId) {
// Not a bet against chance — `randomId()` is a UUIDv4 and a random
// collision is negligible. The collisions that happen are certainties: a
// copied `.vfs`, or a caller deriving `options.id` from something that is
// not unique. Deliberately one code for both, because a clone predating
// the first sync and an imposed id are identical from here; the log digests
// travel in the error so a caller can tell them apart without the library
// committing to a conclusion it would be guessing at.
throw new PairingError(
'peer-collision',
`both folders identify as peer ${a.peerId}: one of them is a copy, or the id was imposed`,
a,
b,
);
}
if (a.syncId !== null && b.syncId !== null && a.syncId !== b.syncId) {
// Authorisation is specific: naming one of the two ids proves the caller
// saw *this* collision. `{ adopt: true }` would disarm the guard at this
// call site forever, including a different collision months later. It is
// the idiom `writeIf(path, data, tag)` already uses.
if (!adopt || (adopt.syncId !== a.syncId && adopt.syncId !== b.syncId)) {
throw new PairingError(
'foreign-mesh',
`these folders belong to different groups (${a.syncId} and ${b.syncId})`,
a,
b,
);
}
}
// The smaller survives — deterministic and transitive, so a whole mesh
// settles on one value without coordinating. Minting needs no agreement
// either: `sync()` has both files in front of it and hands one id to both.
if (a.syncId !== null && b.syncId !== null) return [a.syncId, b.syncId].sort()[0] as string;
return a.syncId ?? b.syncId ?? randomId();
}
interface ConvergedConfig {
text: string[];
changed: boolean;
}
/**
* The config both files have to agree on, and whether they already do.
*
* `text` converges by **union**, so no peer loses a classification another one
* added. Group identity is deliberately not here: `syncId` is decided by the
* pairing guard and written in `close()`, because affiliation records a sync
* that happened rather than one that was attempted.
*
* Computed without writing, because `syncDryRun()` has to answer the same
* question without touching either folder. One implementation for both: a
* second copy is a second answer.
*/
function convergeConfig(fileA: VFSFile, fileB: VFSFile): ConvergedConfig {
const text = [...new Set([...fileA.text, ...fileB.text])].sort();
return {
text,
changed: fileA.text.join() !== text.join() || fileB.text.join() !== text.join(),
};
}
/** Writes a converged config onto both files. `sync()` only — never the dry run. */
function applyConfig(fileA: VFSFile, fileB: VFSFile, config: ConvergedConfig): void {
fileA.text = [...config.text];
fileB.text = [...config.text];
}
/**
* Content `node` wants, does not have, and the other side looks able to hand
* over.
*
* The `state` comparison cannot see this. Which bytes a node stores is local
* policy and is deliberately outside the digest — it has to be, or two peers
* with different policies would never agree on `state` and would sync forever.
* So two peers can hold identical trees while one is still missing files it
* wants, and the fast path would skip the transfer that fixes it.
*
* It is checked against what the other side actually holds, not just against
* the wish: a mesh where nobody has the bytes stays on the fast path instead of
* re-planning a transfer nobody can serve on every pass. Entry inspection only,
* no I/O, and the common case — nothing missing — never looks at the far side.
*/
function refillable(node: VFSNode, own: VFSFile, other: VFSFile): boolean {
const missing: Hash[] = [];
for (const entry of own.entries) {
if (entry.deleted || entry.kind !== 'file' || !entry.hash) continue;
if (entry.held && entry.held !== node.peerId) continue;
if (materialised(entry) || !node.wants(entry)) continue;
missing.push(entry.hash);
}
if (missing.length === 0) return false;
const available = new Set<Hash>();
for (const entry of other.entries) {
if (entry.deleted || entry.kind !== 'file' || !entry.hash || entry.held) continue;
if (materialised(entry)) available.add(entry.hash);
}
return missing.some((hash) => available.has(hash));
}
/**
* Hashes `node` can serve: what it holds now, plus what its policy will make it
* hold. Only feeds the dry run's transfer estimate, so erring here misreports a
* number rather than losing anything.
*/
function fileHashes(entries: VFSEntry[], node: VFSNode): Hash[] {
const hashes: Hash[] = [];
for (const entry of entries) {
if (entry.deleted || entry.kind !== 'file' || !entry.hash || entry.held) continue;
if (!node.wants(entry) && !materialised(entry)) continue;
hashes.push(entry.hash);
}
return hashes;
}
function countTransfers(
writes: VFSEntry[],
localHashes: Set<Hash>,
overlayHashes: Set<Hash>,
remoteHashes: Set<Hash>,
): number {
let count = 0;
for (const entry of writes) {
if (!entry.hash) continue;
if (localHashes.has(entry.hash)) continue;
if (overlayHashes.has(entry.hash)) continue;
if (remoteHashes.has(entry.hash)) count++;
}
return count;
}
function sortActions(actions: SyncDryRunAction[]): SyncDryRunAction[] {
return [...actions].sort(
(x, y) =>
(x.path < y.path ? -1 : x.path > y.path ? 1 : 0) ||
(x.type < y.type ? -1 : x.type > y.type ? 1 : 0) ||
(x.uuid < y.uuid ? -1 : x.uuid > y.uuid ? 1 : 0),
);
}
/**
* Appends what each log is missing and writes both headers, in that order.
*
* The order is the recovery plan: if the process dies halfway, `vfs.json` comes
* up short — declaring less than what is on disk — and the next reconciliation
* catches up. The opposite, a `vfs.json` claiming content that never arrived,
* must not be reachable.
*/
async function close(
a: VFSNode,
b: VFSNode,
fileA: VFSFile,
fileB: VFSFile,
rowsForA: LogRow[],
rowsForB: LogRow[],
at: number,
syncId: string,
): Promise<void> {
if (rowsForA.length > 0) await a.store.append(rowsForA, fileA);
if (rowsForB.length > 0) await b.store.append(rowsForB, fileB);
fileA.peers[b.peerId] = {
lastSync: at,
segment: fileB.log.segment,
offset: fileB.log.size,
digest: fileB.log.digest,
};
fileB.peers[a.peerId] = {
lastSync: at,
segment: fileA.log.segment,
offset: fileA.log.size,
digest: fileA.log.digest,
};
// Affiliation lands here, with the peer marks, for two reasons: both files are
// written in the same pass, so there is no window where one is affiliated and
// the other is not; and a sync that fails before reaching this point leaves
// neither affiliated — the `syncId` records a sync that happened, not one that
// was attempted.
fileA.syncId = fileB.syncId = syncId;
await a.store.write(fileA);
await b.store.write(fileB);
}
/**
* Whether the log has to be opened at all.
*
* Two questions need it, and only two: an entry that exists on one side and not
* the other may be a delete whose tombstone was pruned, and a hash divergence
* has to be told apart from a propagation. Everything else `vfs.json` answers
* on its own.
*/
function needsLog(left: VFSEntry[], right: VFSEntry[]): boolean {
const byUuid = new Map(right.map((entry) => [entry.uuid, entry]));
for (const entry of left) {
const held = byUuid.get(entry.uuid);
if (!held || held.hash !== entry.hash) return true;
}
const mine = new Set(left.map((entry) => entry.uuid));
return right.some((entry) => !mine.has(entry.uuid));
}
/**
* The `updated` of the oldest losing version among the conflicts a copy would be
* kept for, or `null` when there is nothing an archive could help with.
*
* Location conflicts are excluded: they never keep a copy, so nothing is saved
* by proving ancestry for them.
*/
function coldest(conflicts: ConflictReport[]): number | null {
let oldest: number | null = null;
for (const report of conflicts) {
if (report.kind === 'location' || report.kind === 'kind') continue;
const loser = report.winner === 'a' ? report.b : report.a;
if (!loser) continue;
if (oldest === null || loser.updated < oldest) oldest = loser.updated;
}
return oldest;
}
/**
* Rows from the closed segments that could hold links at or after `since`.
*
* Segments are named for the moment they closed, so any archive stamped before
* `since` is entirely older than the version in question and cannot contain the
* link being looked for. They are immutable, so the store caches them for good —
* and a segment that has been deleted (§3 says they are deletable) simply
* contributes nothing, which degrades to the conflict copy that would have been
* kept anyway.
*/
async function readArchives(node: VFSNode, file: VFSFile, since: number): Promise<LogRow[]> {
const out: LogRow[] = [];
for (const segment of file.log.archives ?? []) {
if (segment < since) continue;
out.push(...(await node.store.readArchive(segment)));
}
return out;
}
/** The tail of a peer's active segment, or the whole of it when the peer rotated. */
async function readPeerLog(peer: VFSNode, mark?: { segment: number; offset: number; digest: Hash }) {
const file = await peer.file();
if (mark && mark.digest === file.log.digest) return peer.store.logRows();
if (!mark || mark.segment !== file.log.segment) return peer.store.logRows();
return peer.store.rowsSince(mark.offset);
}
interface AutoMergeContext {
overlay: Map<Hash, Uint8Array>;
rows: Array<Omit<LogRow, 'op'>>;
batch: string;
at: number;
enabled: boolean;
resolveText?: (info: TextConflictInfo) => Promise<string | null>;
}
/**
* Tries a three-way merge on the content conflicts the merge flagged as text.
*
* Computed on one side only — `sync()` has both nodes in front of it, so there
* is no need for two implementations to agree byte for byte forever. The result
* travels as an ordinary write whose row carries `prev` *and* `prev2`: two
* parents, which is what stops a third peer from reclassifying the merge as a
* fresh conflict on every pass.
*/
async function autoMergeText(
a: VFSNode,
b: VFSNode,
merge: { entries: VFSEntry[]; conflicts: ConflictReport[] },
history: History,
context: AutoMergeContext,
): Promise<{ count: number; paths: string[] }> {
let count = 0;
const paths: string[] = [];
for (const report of merge.conflicts) {
if (!report.text || !report.a || !report.b) continue;
if (!context.enabled && !context.resolveText) continue;
const left = report.a;
const right = report.b;
if (left.size > MAX_TEXT_MERGE || right.size > MAX_TEXT_MERGE) continue;
const ancestor = report.base ?? history.commonAncestor(report.uuid, left, right);
const baseBytes = ancestor
? ((await a.baseOf(ancestor)) ?? (await b.baseOf(ancestor)))
: null;
const mine = await readAt(a, left.path);
const theirs = await readAt(b, right.path);
if (!mine || !theirs) continue;
const baseText = baseBytes ? decodeText(baseBytes) : null;
let text: string | null = null;
if (context.enabled && baseText !== null) {
const attempt = diff3(baseText, decodeText(mine), decodeText(theirs));
if (attempt.ok) text = attempt.text;
}
if (text === null && context.resolveText) {
text = await context.resolveText({
path: report.path,
base: baseText,
a: decodeText(mine),
b: decodeText(theirs),
peerA: a.peerId,
peerB: b.peerId,
});
}
if (text === null) continue;
const data = encodeText(text);
const hash = await sha256(data);
context.overlay.set(hash, data);
const winner = report.winner === 'a' ? left : right;
const loser = report.winner === 'a' ? right : left;
const entry = merge.entries.find((item) => item.uuid === report.uuid);
if (!entry) continue;
entry.hash = hash;
entry.size = data.byteLength;
entry.updated = Math.max(context.at, left.updated, right.updated) + 1;
entry.peerId = a.peerId;
entry.prev = winner.hash;
if (loser.hash) entry.prev2 = loser.hash;
// The copy was only ever the pending decision; the merge settled it.
if (report.copy) {
const at = merge.entries.indexOf(report.copy);
if (at >= 0) merge.entries.splice(at, 1);
delete report.copy;
}
report.text = true;
context.rows.push({
batch: context.batch,
at: entry.updated,
peerId: a.peerId,
uuid: entry.uuid,
type: 'write',
kind: 'file',
path: entry.path,
hash,
size: data.byteLength,
prev: winner.hash ?? null,
...(loser.hash ? { prev2: loser.hash } : {}),
});
count++;
paths.push(report.path);
}
return { count, paths };
}
async function readAt(node: VFSNode, path: string): Promise<Uint8Array | null> {
try {
return await node.read(path);
} catch {
return null;
}
}
interface Candidate {
node: VFSNode;
entries: VFSEntry[];
}
/** Overlay first, then each candidate holder in turn. `onHit` tallies transfers. */
function chain(
overlay: Map<Hash, Uint8Array>,
candidates: Candidate[],
onHit: () => void,
): ContentSource {
return {
async open(hash: Hash): Promise<ContentHandle | null> {
const held = overlay.get(hash);
if (held) {
return {
size: held.byteLength,
read: async () => held,
stream: async () => new Response(held as BodyInit).body as ReadableStream<Uint8Array>,
};
}
for (const candidate of candidates) {
const handle = await holds(candidate.node, candidate.entries, hash);
if (!handle) continue;
onHit();
return handle;
}
return null;
},
};
}
export interface MeshEdge {
a: VFSNode;
b: VFSNode;
}
export interface MeshResult {
edge: MeshEdge;
/** Absent when this edge threw — see {@link MeshResult.error}. */
result?: SyncResult;
/** The edge failed. Every other edge in the pass still ran. */
error?: Error;
}
/**
* Runs every edge once, in order. Repeat until nothing changes to let updates
* propagate down a chain — each pass moves a change one hop further.
*/
export async function syncMesh(edges: MeshEdge[], options: SyncOptions = {}): Promise<MeshResult[]> {
const results: MeshResult[] = [];
for (const edge of edges) {
// Per edge, because one badly paired pair must not paralyse the mesh. A
// throw used to abort the loop and lose every result already collected: in
// a mesh of five folders with one foreign pairing, none of them synced.
try {
results.push({ edge, result: await sync(edge.a, edge.b, options) });
} catch (error) {
results.push({ edge, error: error instanceof Error ? error : new Error(String(error)) });
}
}
return results;
}
/** Repeats `syncMesh` until the mesh reaches a fixed point (or `maxRounds`). */
export async function syncUntilStable(
edges: MeshEdge[],
options: SyncOptions & { maxRounds?: number } = {},
): Promise<MeshResult[][]> {
const maxRounds = options.maxRounds ?? 10;
const rounds: MeshResult[][] = [];
for (let round = 0; round < maxRounds; round++) {
const results = await syncMesh(edges, options);
rounds.push(results);
// An edge that threw is not progress. Counting it as change would keep the
// loop going to `maxRounds` on every call, for a failure that repeats
// identically each time.
if (!results.some((item) => item.result?.changed)) break;
}
return rounds;
}
|