fix(server): use stat instead of exifinfo for file date metadata (#17311)

* use stat instead of filecreatedate

* update tests

* unused import
This commit is contained in:
Mert 2025-04-01 18:24:07 -04:00 committed by GitHub
parent 502854cee1
commit d911b76c08
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 92 additions and 75 deletions

View File

@ -1,5 +1,6 @@
import { BinaryField, ExifDateTime } from 'exiftool-vendored';
import { randomBytes } from 'node:crypto';
import { Stats } from 'node:fs';
import { constants } from 'node:fs/promises';
import { defaults } from 'src/config';
import { AssetEntity } from 'src/entities/asset.entity';
@ -21,14 +22,8 @@ describe(MetadataService.name, () => {
let mocks: ServiceMocks;
const mockReadTags = (exifData?: Partial<ImmichTags>, sidecarData?: Partial<ImmichTags>) => {
exifData = {
FileSize: '123456',
FileCreateDate: '2024-01-01T00:00:00.000Z',
FileModifyDate: '2024-01-01T00:00:00.000Z',
...exifData,
};
mocks.metadata.readTags.mockReset();
mocks.metadata.readTags.mockResolvedValueOnce(exifData);
mocks.metadata.readTags.mockResolvedValueOnce(exifData ?? {});
mocks.metadata.readTags.mockResolvedValueOnce(sidecarData ?? {});
};
@ -114,6 +109,17 @@ describe(MetadataService.name, () => {
});
describe('handleMetadataExtraction', () => {
beforeEach(() => {
const time = new Date('2022-01-01T00:00:00.000Z');
const timeMs = time.valueOf();
mocks.storage.stat.mockResolvedValue({
size: 123_456,
mtime: time,
mtimeMs: timeMs,
birthtimeMs: timeMs,
} as Stats);
});
it('should handle an asset that could not be found', async () => {
await expect(sut.handleMetadataExtraction({ id: assetStub.image.id })).resolves.toBe(JobStatus.FAILED);
@ -145,10 +151,13 @@ describe(MetadataService.name, () => {
const fileCreatedAt = new Date('2022-01-01T00:00:00.000Z');
const fileModifiedAt = new Date('2021-01-01T00:00:00.000Z');
mocks.asset.getByIds.mockResolvedValue([assetStub.image]);
mockReadTags({
FileCreateDate: fileCreatedAt.toISOString(),
FileModifyDate: fileModifiedAt.toISOString(),
});
mocks.storage.stat.mockResolvedValue({
size: 123_456,
mtime: fileModifiedAt,
mtimeMs: fileModifiedAt.valueOf(),
birthtimeMs: fileCreatedAt.valueOf(),
} as Stats);
mockReadTags();
await sut.handleMetadataExtraction({ id: assetStub.image.id });
expect(mocks.asset.getByIds).toHaveBeenCalledWith([assetStub.image.id], { faces: { person: false } });
@ -168,10 +177,13 @@ describe(MetadataService.name, () => {
const fileCreatedAt = new Date('2021-01-01T00:00:00.000Z');
const fileModifiedAt = new Date('2022-01-01T00:00:00.000Z');
mocks.asset.getByIds.mockResolvedValue([assetStub.image]);
mockReadTags({
FileCreateDate: fileCreatedAt.toISOString(),
FileModifyDate: fileModifiedAt.toISOString(),
});
mocks.storage.stat.mockResolvedValue({
size: 123_456,
mtime: fileModifiedAt,
mtimeMs: fileModifiedAt.valueOf(),
birthtimeMs: fileCreatedAt.valueOf(),
} as Stats);
mockReadTags();
await sut.handleMetadataExtraction({ id: assetStub.image.id });
expect(mocks.asset.getByIds).toHaveBeenCalledWith([assetStub.image.id], { faces: { person: false } });
@ -206,10 +218,14 @@ describe(MetadataService.name, () => {
it('should handle lists of numbers', async () => {
mocks.asset.getByIds.mockResolvedValue([assetStub.image]);
mocks.storage.stat.mockResolvedValue({
size: 123_456,
mtime: assetStub.image.fileModifiedAt,
mtimeMs: assetStub.image.fileModifiedAt.valueOf(),
birthtimeMs: assetStub.image.fileCreatedAt.valueOf(),
} as Stats);
mockReadTags({
ISO: [160],
FileCreateDate: assetStub.image.fileCreatedAt.toISOString(),
FileModifyDate: assetStub.image.fileModifiedAt.toISOString(),
});
await sut.handleMetadataExtraction({ id: assetStub.image.id });
@ -228,11 +244,15 @@ describe(MetadataService.name, () => {
mocks.asset.getByIds.mockResolvedValue([assetStub.withLocation]);
mocks.systemMetadata.get.mockResolvedValue({ reverseGeocoding: { enabled: true } });
mocks.map.reverseGeocode.mockResolvedValue({ city: 'City', state: 'State', country: 'Country' });
mocks.storage.stat.mockResolvedValue({
size: 123_456,
mtime: assetStub.withLocation.fileModifiedAt,
mtimeMs: assetStub.withLocation.fileModifiedAt.valueOf(),
birthtimeMs: assetStub.withLocation.fileCreatedAt.valueOf(),
} as Stats);
mockReadTags({
GPSLatitude: assetStub.withLocation.exifInfo!.latitude!,
GPSLongitude: assetStub.withLocation.exifInfo!.longitude!,
FileCreateDate: assetStub.withLocation.fileCreatedAt.toISOString(),
FileModifyDate: assetStub.withLocation.fileModifiedAt.toISOString(),
});
await sut.handleMetadataExtraction({ id: assetStub.image.id });
@ -475,6 +495,12 @@ describe(MetadataService.name, () => {
it('should extract the MotionPhotoVideo tag from Samsung HEIC motion photos', async () => {
mocks.asset.getByIds.mockResolvedValue([{ ...assetStub.livePhotoWithOriginalFileName, livePhotoVideoId: null }]);
mocks.storage.stat.mockResolvedValue({
size: 123_456,
mtime: assetStub.livePhotoWithOriginalFileName.fileModifiedAt,
mtimeMs: assetStub.livePhotoWithOriginalFileName.fileModifiedAt.valueOf(),
birthtimeMs: assetStub.livePhotoWithOriginalFileName.fileCreatedAt.valueOf(),
} as Stats);
mockReadTags({
Directory: 'foo/bar/',
MotionPhoto: 1,
@ -483,8 +509,6 @@ describe(MetadataService.name, () => {
// instead of the EmbeddedVideoFile, since HEIC MotionPhotos include both
EmbeddedVideoFile: new BinaryField(0, ''),
EmbeddedVideoType: 'MotionPhoto_Data',
FileCreateDate: assetStub.livePhotoWithOriginalFileName.fileCreatedAt.toISOString(),
FileModifyDate: assetStub.livePhotoWithOriginalFileName.fileModifiedAt.toISOString(),
});
mocks.crypto.hashSha1.mockReturnValue(randomBytes(512));
mocks.asset.create.mockResolvedValue(assetStub.livePhotoMotionAsset);
@ -525,14 +549,18 @@ describe(MetadataService.name, () => {
});
it('should extract the EmbeddedVideo tag from Samsung JPEG motion photos', async () => {
mocks.storage.stat.mockResolvedValue({
size: 123_456,
mtime: assetStub.livePhotoWithOriginalFileName.fileModifiedAt,
mtimeMs: assetStub.livePhotoWithOriginalFileName.fileModifiedAt.valueOf(),
birthtimeMs: assetStub.livePhotoWithOriginalFileName.fileCreatedAt.valueOf(),
} as Stats);
mocks.asset.getByIds.mockResolvedValue([{ ...assetStub.livePhotoWithOriginalFileName, livePhotoVideoId: null }]);
mockReadTags({
Directory: 'foo/bar/',
EmbeddedVideoFile: new BinaryField(0, ''),
EmbeddedVideoType: 'MotionPhoto_Data',
MotionPhoto: 1,
FileCreateDate: assetStub.livePhotoWithOriginalFileName.fileCreatedAt.toISOString(),
FileModifyDate: assetStub.livePhotoWithOriginalFileName.fileModifiedAt.toISOString(),
});
mocks.crypto.hashSha1.mockReturnValue(randomBytes(512));
mocks.asset.create.mockResolvedValue(assetStub.livePhotoMotionAsset);
@ -574,13 +602,17 @@ describe(MetadataService.name, () => {
it('should extract the motion photo video from the XMP directory entry ', async () => {
mocks.asset.getByIds.mockResolvedValue([{ ...assetStub.livePhotoWithOriginalFileName, livePhotoVideoId: null }]);
mocks.storage.stat.mockResolvedValue({
size: 123_456,
mtime: assetStub.livePhotoWithOriginalFileName.fileModifiedAt,
mtimeMs: assetStub.livePhotoWithOriginalFileName.fileModifiedAt.valueOf(),
birthtimeMs: assetStub.livePhotoWithOriginalFileName.fileCreatedAt.valueOf(),
} as Stats);
mockReadTags({
Directory: 'foo/bar/',
MotionPhoto: 1,
MicroVideo: 1,
MicroVideoOffset: 1,
FileCreateDate: assetStub.livePhotoWithOriginalFileName.fileCreatedAt.toISOString(),
FileModifyDate: assetStub.livePhotoWithOriginalFileName.fileModifiedAt.toISOString(),
});
mocks.crypto.hashSha1.mockReturnValue(randomBytes(512));
mocks.asset.create.mockResolvedValue(assetStub.livePhotoMotionAsset);

View File

@ -1,9 +1,10 @@
import { Injectable } from '@nestjs/common';
import { ContainerDirectoryItem, ExifDateTime, Maybe, Tags } from 'exiftool-vendored';
import { ContainerDirectoryItem, Maybe, Tags } from 'exiftool-vendored';
import { firstDateTime } from 'exiftool-vendored/dist/FirstDateTime';
import { Insertable } from 'kysely';
import _ from 'lodash';
import { Duration } from 'luxon';
import { Stats } from 'node:fs';
import { constants } from 'node:fs/promises';
import path from 'node:path';
import { JOBS_ASSET_PAGINATION_SIZE } from 'src/constants';
@ -77,6 +78,11 @@ const validateRange = (value: number | undefined, min: number, max: number): Non
type ImmichTagsWithFaces = ImmichTags & { RegionInfo: NonNullable<ImmichTags['RegionInfo']> };
type Dates = {
dateTimeOriginal: Date;
localDateTime: Date;
};
@Injectable()
export class MetadataService extends BaseService {
@OnEvent({ name: 'app.bootstrap', workers: [ImmichWorker.MICROSERVICES] })
@ -171,18 +177,13 @@ export class MetadataService extends BaseService {
return JobStatus.FAILED;
}
const exifTags = await this.getExifTags(asset);
if (!exifTags.FileCreateDate || !exifTags.FileModifyDate || exifTags.FileSize === undefined) {
this.logger.warn(`Missing file creation or modification date for asset ${asset.id}: ${asset.originalPath}`);
const stat = await this.storageRepository.stat(asset.originalPath);
exifTags.FileCreateDate = stat.ctime.toISOString();
exifTags.FileModifyDate = stat.mtime.toISOString();
exifTags.FileSize = stat.size.toString();
}
const [exifTags, stats] = await Promise.all([
this.getExifTags(asset),
this.storageRepository.stat(asset.originalPath),
]);
this.logger.verbose('Exif Tags', exifTags);
const { dateTimeOriginal, localDateTime, timeZone, modifyDate } = this.getDates(asset, exifTags);
const dates = this.getDates(asset, exifTags, stats);
const { width, height } = this.getImageDimensions(exifTags);
let geo: ReverseGeocodeResult, latitude: number | null, longitude: number | null;
@ -200,9 +201,9 @@ export class MetadataService extends BaseService {
assetId: asset.id,
// dates
dateTimeOriginal,
modifyDate,
timeZone,
dateTimeOriginal: dates.dateTimeOriginal,
modifyDate: stats.mtime,
timeZone: dates.timeZone,
// gps
latitude,
@ -212,7 +213,7 @@ export class MetadataService extends BaseService {
city: geo.city,
// image/file
fileSizeInByte: Number.parseInt(exifTags.FileSize!),
fileSizeInByte: stats.size,
exifImageHeight: validate(height),
exifImageWidth: validate(width),
orientation: validate(exifTags.Orientation)?.toString() ?? null,
@ -245,15 +246,15 @@ export class MetadataService extends BaseService {
this.assetRepository.update({
id: asset.id,
duration: exifTags.Duration?.toString() ?? null,
localDateTime,
fileCreatedAt: exifData.dateTimeOriginal ?? undefined,
fileModifiedAt: exifData.modifyDate ?? undefined,
localDateTime: dates.localDateTime,
fileCreatedAt: dates.dateTimeOriginal ?? undefined,
fileModifiedAt: stats.mtime,
}),
this.applyTagList(asset, exifTags),
];
if (this.isMotionPhoto(asset, exifTags)) {
promises.push(this.applyMotionPhotos(asset, exifTags, exifData.fileSizeInByte!));
promises.push(this.applyMotionPhotos(asset, exifTags, dates, stats));
}
if (isFaceImportEnabled(metadata) && this.hasTaggedFaces(exifTags)) {
@ -432,7 +433,7 @@ export class MetadataService extends BaseService {
return asset.type === AssetType.IMAGE && !!(tags.MotionPhoto || tags.MicroVideo);
}
private async applyMotionPhotos(asset: AssetEntity, tags: ImmichTags, fileSize: number) {
private async applyMotionPhotos(asset: AssetEntity, tags: ImmichTags, dates: Dates, stats: Stats) {
const isMotionPhoto = tags.MotionPhoto;
const isMicroVideo = tags.MicroVideo;
const videoOffset = tags.MicroVideoOffset;
@ -466,7 +467,7 @@ export class MetadataService extends BaseService {
this.logger.debug(`Starting motion photo video extraction for asset ${asset.id}: ${asset.originalPath}`);
try {
const position = fileSize - length - padding;
const position = stats.size - length - padding;
let video: Buffer;
// Samsung MotionPhoto video extraction
// HEIC-encoded
@ -505,13 +506,12 @@ export class MetadataService extends BaseService {
}
} else {
const motionAssetId = this.cryptoRepository.randomUUID();
const dates = this.getDates(asset, tags);
motionAsset = await this.assetRepository.create({
id: motionAssetId,
libraryId: asset.libraryId,
type: AssetType.VIDEO,
fileCreatedAt: dates.dateTimeOriginal,
fileModifiedAt: dates.modifyDate,
fileModifiedAt: stats.mtime,
localDateTime: dates.localDateTime,
checksum,
ownerId: asset.ownerId,
@ -634,7 +634,7 @@ export class MetadataService extends BaseService {
}
}
private getDates(asset: AssetEntity, exifTags: ImmichTags) {
private getDates(asset: AssetEntity, exifTags: ImmichTags, stats: Stats) {
const dateTime = firstDateTime(exifTags as Maybe<Tags>, EXIF_DATE_TAGS);
this.logger.verbose(`Date and time is ${dateTime} for asset ${asset.id}: ${asset.originalPath}`);
@ -654,17 +654,16 @@ export class MetadataService extends BaseService {
this.logger.debug(`No timezone information found for asset ${asset.id}: ${asset.originalPath}`);
}
const modifyDate = this.toDate(exifTags.FileModifyDate!);
let dateTimeOriginal = dateTime?.toDate();
let localDateTime = dateTime?.toDateTime().setZone('UTC', { keepLocalTime: true }).toJSDate();
if (!localDateTime || !dateTimeOriginal) {
const fileCreatedAt = this.toDate(exifTags.FileCreateDate!);
const earliestDate = this.earliestDate(fileCreatedAt, modifyDate);
// FileCreateDate is not available on linux, likely because exiftool hasn't integrated the statx syscall yet
// birthtime is not available in Docker on macOS, so it appears as 0
const earliestDate = stats.birthtimeMs ? new Date(Math.min(stats.mtimeMs, stats.birthtimeMs)) : stats.mtime;
this.logger.debug(
`No exif date time found, falling back on ${earliestDate.toISOString()}, earliest of file creation and modification for assset ${asset.id}: ${asset.originalPath}`,
`No exif date time found, falling back on ${earliestDate.toISOString()}, earliest of file creation and modification for asset ${asset.id}: ${asset.originalPath}`,
);
dateTimeOriginal = earliestDate;
localDateTime = earliestDate;
dateTimeOriginal = localDateTime = earliestDate;
}
this.logger.verbose(
@ -675,18 +674,9 @@ export class MetadataService extends BaseService {
dateTimeOriginal,
timeZone,
localDateTime,
modifyDate,
};
}
private toDate(date: string | ExifDateTime): Date {
return typeof date === 'string' ? new Date(date) : date.toDate();
}
private earliestDate(a: Date, b: Date) {
return new Date(Math.min(a.valueOf(), b.valueOf()));
}
private hasGeo(tags: ImmichTags): tags is ImmichTags & { GPSLatitude: number; GPSLongitude: number } {
return (
tags.GPSLatitude !== undefined &&

View File

@ -39,7 +39,12 @@ describe(MetadataService.name, () => {
beforeEach(() => {
({ sut, mocks } = newTestService(MetadataService, { metadata: metadataRepository }));
mocks.storage.stat.mockResolvedValue({ size: 123_456, ctime: new Date(), mtime: new Date() } as Stats);
mocks.storage.stat.mockResolvedValue({
size: 123_456,
mtime: new Date(654_321),
mtimeMs: 654_321,
birthtimeMs: 654_322,
} as Stats);
delete process.env.TZ;
});
@ -54,8 +59,6 @@ describe(MetadataService.name, () => {
description: 'should handle no time zone information',
exifData: {
DateTimeOriginal: '2022:01:01 00:00:00',
FileCreateDate: '2022:01:01 00:00:00',
FileModifyDate: '2022:01:01 00:00:00',
},
expected: {
localDateTime: '2022-01-01T00:00:00.000Z',
@ -68,8 +71,6 @@ describe(MetadataService.name, () => {
serverTimeZone: 'America/Los_Angeles',
exifData: {
DateTimeOriginal: '2022:01:01 00:00:00',
FileCreateDate: '2022:01:01 00:00:00',
FileModifyDate: '2022:01:01 00:00:00',
},
expected: {
localDateTime: '2022-01-01T00:00:00.000Z',
@ -82,8 +83,6 @@ describe(MetadataService.name, () => {
serverTimeZone: 'Europe/Brussels',
exifData: {
DateTimeOriginal: '2022:01:01 00:00:00',
FileCreateDate: '2022:01:01 00:00:00',
FileModifyDate: '2022:01:01 00:00:00',
},
expected: {
localDateTime: '2022-01-01T00:00:00.000Z',
@ -96,8 +95,6 @@ describe(MetadataService.name, () => {
serverTimeZone: 'Europe/Brussels',
exifData: {
DateTimeOriginal: '2022:06:01 00:00:00',
FileCreateDate: '2022:06:01 00:00:00',
FileModifyDate: '2022:06:01 00:00:00',
},
expected: {
localDateTime: '2022-06-01T00:00:00.000Z',
@ -109,8 +106,6 @@ describe(MetadataService.name, () => {
description: 'should handle a +13:00 time zone',
exifData: {
DateTimeOriginal: '2022:01:01 00:00:00+13:00',
FileCreateDate: '2022:01:01 00:00:00+13:00',
FileModifyDate: '2022:01:01 00:00:00+13:00',
},
expected: {
localDateTime: '2022-01-01T00:00:00.000Z',