Add opt-in Unicode NFC path identity for CouchDB

This commit is contained in:
vorotamoroz committed 2026-10-17 13:31:19 +00:00
1 parent 72dc88c7ef
commit ddb4d52931
39 files changed
+2245 -97

No files matched your search

@@ -1,5 +1,12 @@
import { afterEach, describe, expect, it, vi } from "vitest";
import { fsPromises as fs, os, path } from "@vrtmrz/livesync-commonlib/node";
import { fsPromises as fs, NodeStorageAdapter, os, path } from "@vrtmrz/livesync-commonlib/node";
import { DEFAULT_SETTINGS, type FilePath } from "@vrtmrz/livesync-commonlib/compat/common/types";
import {
ServiceRebuilder,
type ServiceRebuilderDependencies,
} from "@vrtmrz/livesync-commonlib/compat/serviceModules/Rebuilder";
import { PathServiceCompat } from "@vrtmrz/livesync-commonlib/compat/services/implements/injectable/InjectablePathService";
import { ServiceContext } from "@vrtmrz/livesync-commonlib/compat/services/base/ServiceBase";
import { NodeFileSystemAdapter } from "./NodeFileSystemAdapter";
describe("NodeFileSystemAdapter file enumeration", () => {
@@ -116,3 +123,382 @@ describe("NodeFileSystemAdapter file enumeration", () => {
await expect(adapter.getFiles()).resolves.toEqual([]);
});
});
describe("NodeFileSystemAdapter Unicode path resolution", () => {
const tempDirs: string[] = [];
async function createVault() {
const directory = await fs.mkdtemp(path.join(os.tmpdir(), "livesync-cli-unicode-"));
tempDirs.push(directory);
return directory;
}
function createNormalisingAdapter(directory: string, caseInsensitive = false) {
return new NodeFileSystemAdapter(
directory,
undefined,
() => 1,
() => caseInsensitive
);
}
afterEach(async () => {
await Promise.all(tempDirs.splice(0).map((directory) => fs.rm(directory, { recursive: true, force: true })));
});
it.each([0, 1])("keeps an NFD-only file in the storage listing in identity mode %s", async (version) => {
const directory = await createVault();
const originalPath = "café.md".normalize("NFD");
await fs.writeFile(path.join(directory, originalPath), "original");
const report = vi.fn();
const adapter = new NodeFileSystemAdapter(directory, report, () => version);
expect((await adapter.getFiles()).map((file) => file.path)).toEqual([originalPath]);
expect(report).not.toHaveBeenCalled();
await expect(fs.readFile(path.join(directory, originalPath), "utf8")).resolves.toBe("original");
expect(await fs.readdir(directory)).toEqual([originalPath]);
});
it.each([false, true])(
"enumerates nested mixed names once per directory with case-insensitive matching %s",
async (caseInsensitive) => {
const directory = await createVault();
const parent = "café".normalize("NFD");
const child = `${parent}/Étage`;
await fs.mkdir(path.join(directory, child), { recursive: true });
const names = Array.from(
{ length: 12 },
(_, index) =>
`${child}/${index}-résumé.md`
.normalize(index % 2 ? "NFC" : "NFD")
.split("/")
.pop()!
);
for (const name of names) await fs.writeFile(path.join(directory, child, name), "content");
const adapter = createNormalisingAdapter(directory, caseInsensitive);
const list = vi.spyOn(NodeStorageAdapter.prototype, "list");
try {
const files = await adapter.getFiles();
expect(files.map((file) => file.path).sort()).toEqual(names.map((name) => `${child}/${name}`).sort());
expect(list).toHaveBeenCalledTimes(3);
} finally {
list.mockRestore();
}
}
);
it("uses the existing mixed parent during directory preparation and preserves the new leaf spelling", async () => {
const directory = await createVault();
const parent = "café".normalize("NFD");
const child = `${parent}/Étage`;
await fs.mkdir(path.join(directory, child), { recursive: true });
const adapter = createNormalisingAdapter(directory);
await adapter.storage.mkdir("café");
await adapter.storage.mkdir("café/Étage");
const newName = "résumé.md".normalize("NFD");
const file = await adapter.vault.create(`café/Étage/${newName}`, "created");
expect(file.path).toBe(`${child}/${newName}`);
expect(await fs.readdir(directory)).toEqual([parent]);
expect(await fs.readdir(path.join(directory, parent))).toEqual(["Étage"]);
expect(await fs.readdir(path.join(directory, child))).toEqual([newName]);
await expect(adapter.storage.stat(`café/Étage/${newName.normalize("NFC")}`)).resolves.toMatchObject({
size: 7,
});
});
it("rejects an NFC-equivalent file and directory before scanning or writing a child", async () => {
const directory = await createVault();
const composed = "café";
const decomposed = composed.normalize("NFD");
await fs.writeFile(path.join(directory, composed), "file");
await fs.mkdir(path.join(directory, decomposed));
const adapter = createNormalisingAdapter(directory);
await expect(adapter.getFiles()).rejects.toThrow(/ambiguous/i);
await expect(adapter.storage.write(`${composed}/new.md`, "new")).rejects.toThrow(/ambiguous/i);
await expect(fs.readFile(path.join(directory, composed), "utf8")).resolves.toBe("file");
expect(await fs.readdir(path.join(directory, decomposed))).toEqual([]);
});
it.each([false, true])(
"enumerates each directory once for a flat NFC scan with case-insensitive matching %s",
async (caseInsensitive) => {
const directory = await createVault();
for (let index = 0; index < 12; index++) {
await fs.writeFile(path.join(directory, `${index}.md`), "content");
}
const adapter = createNormalisingAdapter(directory, caseInsensitive);
const list = vi.spyOn(NodeStorageAdapter.prototype, "list");
try {
expect(await adapter.getFiles()).toHaveLength(12);
expect(list).toHaveBeenCalledTimes(1);
} finally {
list.mockRestore();
}
}
);
it("reads, updates, and deletes an NFD file through its NFC path", async () => {
const directory = await createVault();
const nfcFolder = "café";
const nfdFolder = nfcFolder.normalize("NFD");
const nfcName = "Ångström.md";
const nfdName = nfcName.normalize("NFD");
await fs.mkdir(path.join(directory, nfdFolder));
await fs.writeFile(path.join(directory, nfdFolder, nfdName), "before", "utf8");
const adapter = createNormalisingAdapter(directory);
const file = await adapter.getAbstractFileByPath(`${nfcFolder}/${nfcName}`);
expect(file?.path).toBe(`${nfdFolder}/${nfdName}`);
if (!file) throw new Error("Expected the NFC path to resolve to the NFD file");
await expect(adapter.vault.read(file)).resolves.toBe("before");
await adapter.vault.modify(file, "after");
await expect(fs.readFile(path.join(directory, nfdFolder, nfdName), "utf8")).resolves.toBe("after");
expect(await fs.readdir(directory)).toEqual([nfdFolder]);
expect(await fs.readdir(path.join(directory, nfdFolder))).toEqual([nfdName]);
await adapter.vault.delete(file);
await expect(fs.readdir(path.join(directory, nfdFolder))).resolves.toEqual([]);
});
it("creates a new file inside an existing NFD directory using that directory spelling", async () => {
const directory = await createVault();
const nfdFolder = "café".normalize("NFD");
await fs.mkdir(path.join(directory, nfdFolder));
const adapter = createNormalisingAdapter(directory);
const file = await adapter.vault.create("café/new.md", "content");
expect(file.path).toBe(`${nfdFolder}/new.md`);
await expect(fs.readFile(path.join(directory, nfdFolder, "new.md"), "utf8")).resolves.toBe("content");
expect(await fs.readdir(directory)).toEqual([nfdFolder]);
});
it("renames a file by case inside an NFD directory while preserving that directory spelling", async () => {
const directory = await createVault();
const nfcFolder = "Caf\u00e9";
const nfdFolder = nfcFolder.normalize("NFD");
await fs.mkdir(path.join(directory, nfdFolder));
await fs.writeFile(path.join(directory, nfdFolder, "Note.md"), "content", "utf8");
const adapter = createNormalisingAdapter(directory, true);
const file = await adapter.getAbstractFileByPathInsensitive(`${nfcFolder}/Note.md`);
if (!file) throw new Error("Expected the NFC path to resolve to the NFD file");
const renamedFile = await adapter.renameFile(file, `${nfcFolder}/note.md`);
expect(renamedFile.path).toBe(`${nfdFolder}/note.md`);
expect(await fs.readdir(directory)).toEqual([nfdFolder]);
await expect(fs.readdir(path.join(directory, nfdFolder))).resolves.toEqual(["note.md"]);
await expect(fs.readFile(path.join(directory, nfdFolder, "note.md"), "utf8")).resolves.toBe("content");
});
it("rejects an NFC/NFD file collision for lookup and vault mutations", async () => {
const directory = await createVault();
const nfcName = "café.md";
const nfdName = nfcName.normalize("NFD");
await fs.writeFile(path.join(directory, nfcName), "NFC", "utf8");
await fs.writeFile(path.join(directory, nfdName), "NFD", "utf8");
const adapter = createNormalisingAdapter(directory);
const exactFile = {
path: nfcName as FilePath,
stat: { ctime: 1, mtime: 2, size: 3, type: "file" as const },
};
await expect(adapter.getAbstractFileByPath(nfcName)).rejects.toThrow(/ambiguous/i);
await expect(adapter.vault.modify(exactFile, "changed")).rejects.toThrow(/ambiguous/i);
await expect(adapter.vault.delete(exactFile)).rejects.toThrow(/ambiguous/i);
await expect(fs.readFile(path.join(directory, nfcName), "utf8")).resolves.toBe("NFC");
await expect(fs.readFile(path.join(directory, nfdName), "utf8")).resolves.toBe("NFD");
expect((await fs.readdir(directory)).sort()).toEqual([nfcName, nfdName].sort());
});
it("rejects an NFC/NFD directory collision before creating a child", async () => {
const directory = await createVault();
const nfcFolder = "café";
const nfdFolder = nfcFolder.normalize("NFD");
await fs.mkdir(path.join(directory, nfcFolder));
await fs.mkdir(path.join(directory, nfdFolder));
await fs.writeFile(path.join(directory, nfcFolder, "inside.md"), "NFC", "utf8");
await fs.writeFile(path.join(directory, nfdFolder, "inside.md"), "NFD", "utf8");
const adapter = createNormalisingAdapter(directory);
await expect(adapter.getAbstractFileByPath(`${nfcFolder}/inside.md`)).rejects.toThrow(/ambiguous/i);
await expect(adapter.storage.write(`${nfcFolder}/created.md`, "created")).rejects.toThrow(/ambiguous/i);
await expect(fs.readdir(path.join(directory, nfcFolder))).resolves.toEqual(["inside.md"]);
await expect(fs.readdir(path.join(directory, nfdFolder))).resolves.toEqual(["inside.md"]);
await expect(fs.readFile(path.join(directory, nfcFolder, "inside.md"), "utf8")).resolves.toBe("NFC");
await expect(fs.readFile(path.join(directory, nfdFolder, "inside.md"), "utf8")).resolves.toBe("NFD");
});
it("rejects enumeration when NFC-equivalent directories contain different files", async () => {
const directory = await createVault();
const nfcFolder = "café";
const nfdFolder = nfcFolder.normalize("NFD");
await fs.mkdir(path.join(directory, nfcFolder));
await fs.mkdir(path.join(directory, nfdFolder));
await fs.writeFile(path.join(directory, nfcFolder, "a.md"), "NFC", "utf8");
await fs.writeFile(path.join(directory, nfdFolder, "b.md"), "NFD", "utf8");
const adapter = createNormalisingAdapter(directory);
await expect(adapter.getFiles()).rejects.toThrow(/ambiguous/i);
await expect(adapter.scanDirectory()).rejects.toThrow(/ambiguous/i);
await expect(fs.readFile(path.join(directory, nfcFolder, "a.md"), "utf8")).resolves.toBe("NFC");
await expect(fs.readFile(path.join(directory, nfdFolder, "b.md"), "utf8")).resolves.toBe("NFD");
});
it("rechecks the directory after a unique file is cached", async () => {
const directory = await createVault();
const nfcName = "café.md";
const nfdName = nfcName.normalize("NFD");
await fs.writeFile(path.join(directory, nfdName), "NFD", "utf8");
const adapter = createNormalisingAdapter(directory);
await expect(adapter.getAbstractFileByPath(nfcName)).resolves.toMatchObject({ path: nfdName });
await fs.writeFile(path.join(directory, nfcName), "NFC", "utf8");
await expect(adapter.getAbstractFileByPath(nfcName)).rejects.toThrow(/ambiguous/i);
await expect(fs.readFile(path.join(directory, nfcName), "utf8")).resolves.toBe("NFC");
await expect(fs.readFile(path.join(directory, nfdName), "utf8")).resolves.toBe("NFD");
});
it.each([
["Foo", "foo"],
["Café", "café".normalize("NFD")],
])("rejects case-insensitive parent aliases %s and %s containing different files", async (first, second) => {
const directory = await createVault();
for (const [parent, name] of [
[first, "one.md"],
[second, "two.md"],
]) {
await fs.mkdir(path.join(directory, parent));
await fs.writeFile(path.join(directory, parent, name), name);
}
const adapter = createNormalisingAdapter(directory, true);
await expect(adapter.getFiles()).rejects.toThrow(/ambiguous/i);
await expect(adapter.scanDirectory()).rejects.toThrow(/ambiguous/i);
await expect(fs.readFile(path.join(directory, first, "one.md"), "utf8")).resolves.toBe("one.md");
await expect(fs.readFile(path.join(directory, second, "two.md"), "utf8")).resolves.toBe("two.md");
});
it.each([
{ version: 0, caseInsensitive: true },
{ version: 1, caseInsensitive: false },
])(
"compatibility: retains distinct case paths in mode $version with case-insensitive matching $caseInsensitive",
async ({ version, caseInsensitive }) => {
const directory = await createVault();
const names = ["Café/one.md", "café/two.md".normalize("NFD")];
for (const name of names) {
await fs.mkdir(path.dirname(path.join(directory, name)));
await fs.writeFile(path.join(directory, name), name);
}
const adapter = new NodeFileSystemAdapter(
directory,
undefined,
() => version,
() => caseInsensitive
);
expect((await adapter.getFiles()).map((file) => file.path).sort()).toEqual([...names].sort());
for (const name of names) await expect(adapter.storage.read(name)).resolves.toBe(name);
}
);
it("refuses reconstruction before starting remote activity when parent aliases are ambiguous", async () => {
const directory = await createVault();
for (const name of ["Café/one.md", "café/two.md".normalize("NFD")]) {
await fs.mkdir(path.dirname(path.join(directory, name)));
await fs.writeFile(path.join(directory, name), name);
}
const adapter = createNormalisingAdapter(directory, true);
const settings = {
...DEFAULT_SETTINGS,
pathNormalizationVersion: 1 as const,
handleFilenameCaseSensitive: false,
};
const setting = { currentSettings: () => settings, suspendAllSync: { addHandler: vi.fn() } };
const runBoundedRemoteActivity = vi.fn(async () => undefined);
const rebuilder = new ServiceRebuilder({
setting,
storageAccess: { getFileNames: async () => (await adapter.getFiles()).map((file) => file.path) },
database: {
onDatabaseReset: { addHandler: vi.fn() },
isDatabaseReady: () => true,
localDatabase: { findAllDocs: async function* () {} },
},
vault: { isTargetFile: async () => true },
path: new PathServiceCompat(new ServiceContext(), { settingService: setting as never }),
replicator: { runBoundedRemoteActivity },
UI: { showMarkdownDialog: vi.fn(async () => undefined) },
} as unknown as ServiceRebuilderDependencies);
await expect(rebuilder.$rebuildEverything()).rejects.toThrow(/ambiguous/i);
expect(runBoundedRemoteActivity).not.toHaveBeenCalled();
});
it("uses fresh listings after external additions, renames, deletions, and collisions", async () => {
const directory = await createVault();
const nfdName = "café.md".normalize("NFD");
await fs.writeFile(path.join(directory, nfdName), "original");
const adapter = createNormalisingAdapter(directory);
expect((await adapter.getFiles()).map((file) => file.path)).toEqual([nfdName]);
const renamed = "résumé.md".normalize("NFD");
await fs.rename(path.join(directory, nfdName), path.join(directory, renamed));
await fs.writeFile(path.join(directory, "new.md"), "new");
expect((await adapter.getFiles()).map((file) => file.path).sort()).toEqual(["new.md", renamed]);
await fs.rm(path.join(directory, "new.md"));
expect((await adapter.getFiles()).map((file) => file.path)).toEqual([renamed]);
await fs.writeFile(path.join(directory, renamed.normalize("NFC")), "collision");
await expect(adapter.getFiles()).rejects.toThrow(/ambiguous/i);
});
it("retains the storage boundary for exact metadata reads and directory scans", async () => {
const directory = await createVault();
const outside = await createVault();
await fs.writeFile(path.join(outside, "note.md"), "outside");
await fs.symlink(outside, path.join(directory, "linked"), "dir");
const adapter = createNormalisingAdapter(directory);
await expect(adapter.getFiles()).resolves.toEqual([]);
await expect(adapter.storage.statExact("linked/note.md")).rejects.toThrow(/symbolic links/i);
await expect(adapter.storage.write("linked/note.md", "changed")).rejects.toThrow(/symbolic links/i);
await expect(fs.readFile(path.join(outside, "note.md"), "utf8")).resolves.toBe("outside");
});
it("detects NFC and case collisions when lookup is case-insensitive", async () => {
const directory = await createVault();
const upperName = "Café.md";
const lowerNfdName = "café.md".normalize("NFD");
await fs.writeFile(path.join(directory, upperName), "upper", "utf8");
await fs.writeFile(path.join(directory, lowerNfdName), "lower", "utf8");
const adapter = createNormalisingAdapter(directory, true);
await expect(adapter.getAbstractFileByPathInsensitive("CAFÉ.md")).rejects.toThrow(/ambiguous/i);
await expect(adapter.getFiles()).rejects.toThrow(/ambiguous/i);
await expect(adapter.scanDirectory()).rejects.toThrow(/ambiguous/i);
await expect(fs.readFile(path.join(directory, upperName), "utf8")).resolves.toBe("upper");
await expect(fs.readFile(path.join(directory, lowerNfdName), "utf8")).resolves.toBe("lower");
});
it("retains exact-path behaviour when path normalisation is disabled", async () => {
const directory = await createVault();
const nfcName = "café.md";
const nfdName = nfcName.normalize("NFD");
await fs.writeFile(path.join(directory, nfdName), "NFD", "utf8");
const adapter = new NodeFileSystemAdapter(directory);
await expect(adapter.getAbstractFileByPath(nfcName)).resolves.toBeNull();
await adapter.storage.write(nfcName, "NFC");
await expect(fs.readFile(path.join(directory, nfcName), "utf8")).resolves.toBe("NFC");
await expect(fs.readFile(path.join(directory, nfdName), "utf8")).resolves.toBe("NFD");
});
});