Peach: Local-first video metadata scraper and media manager.
fix(scan): 目录改名又换了编码时,扫描按番号沿用原记录
longmeidaoc5a4dbf5d30e
diff --git a/CHANGELOG.md b/CHANGELOG.mdindex 400c69a28..16289df85 100644--- a/CHANGELOG.md+++ b/CHANGELOG.md@@ -37,6 +37,7 @@ Removed / Fixed / Security。- **采集**:升级后的自动修复遇到损坏视频不再整项失败,探不出时长或抽不出帧的文件留在问题清单里;因来源限流暂缓的影片仍会稍后重试。- **采集**:一次勾选上万部视频补资料或探测时长,不再因数据库参数超限报错。- **配置**:卸载被在跑的任务拦下时,提示写出是哪个任务,可以去任务中心等它完成或取消。+- **馆藏**:影片目录改名又换了编码后,扫描会沿用原来那条记录,标签、封面和观看记录都保留,不再多出一条没有番号的重复条目;同番号的旧记录有多条时仍交给资源同步处理。## [0.53.0] - 2026-11-02diff --git a/docs/OPERATIONS.md b/docs/OPERATIONS.mdindex ab1772bc3..e14f8743c 100644--- a/docs/OPERATIONS.md+++ b/docs/OPERATIONS.md@@ -276,7 +276,7 @@ Windows 测试包默认使用 `%LOCALAPPDATA%\Peach\peach-data`。首次设置没有保存过推送发现设置时,本地文件监控默认开启,CloudDrive 通知需另行配置;已明确关闭的设置保留。监控使用系统文件事件和去抖队列,空闲时等待事件,不定期遍历媒体目录;网络盘不递归监控。未收到通知或服务停机期间新增的文件可由「同步新增文件」补齐。-本机移动事件和已记录的系统文件身份可保留改名、同卷移动后的影片 ID、人工资料和个人记录。文件身份随本地扫描和登记积累;升级前没有身份记录、停机时已经发生的移动,以及跨卷复制,不能仅凭同名或同大小自动认作原片。旧路径仍存在的副本与硬链接分别登记。+本机移动事件和已记录的系统文件身份可保留改名、同卷移动后的影片 ID、人工资料和个人记录。文件身份随本地扫描和登记积累;升级前没有身份记录、停机时已经发生的移动,以及跨卷复制,不能仅凭同名或同大小自动认作原片。旧路径仍存在的副本与硬链接分别登记。改名时换了编码、文件身份对不上的影片,扫描按番号认回:同一媒体目录下这个番号的在库影片只有一部且旧文件已不在,新文件所在目录也只有这一部同番号视频,就沿用原记录与它的标签、封面和个人记录,时长与分辨率重新探测([ADR-0131](adr/0131-scan-relocates-missing-row-by-code.md))。一趟任务的结论、「7 天内不再问」、问题清单和完整记录收在同一块提示里。结束的任务可在完整记录旁「导出完整问题清单」,下载包含全部分页记录的 JSON。导出期间任务或记录变化会停止导出,取消或失败不会下载半份文件;缺少原始问题记录时提示导出诊断日志。「归档」收起这块提示,完整记录和任务历史仍保留;归档后留一条可「撤销」的提示,刷新页面后不再显示。diff --git a/docs/adr/0131-scan-relocates-missing-row-by-code.md b/docs/adr/0131-scan-relocates-missing-row-by-code.mdnew file mode 100644index 000000000..599f623a8--- /dev/null+++ b/docs/adr/0131-scan-relocates-missing-row-by-code.md@@ -0,0 +1,47 @@+# ADR-0131:扫描时按番号把文件已不在的在库行改指到新文件++- 状态:已采纳+- 日期:2026-11-03+- 关系:补充 [ADR-0087](0087-vanished-files-keep-personal-records.md) 决策二的触发时机;判据沿用它的番号键++## 背景++[#312](https://github.com/peach-mitao/peach/issues/312) 的报告者把影片目录改了名,同时换了编码。之后先在资源同步里「检查文件」,再跑增量扫描,结果旧行仍指向不存在的旧路径,扫描又登记了一条没有番号的新行,标签、封面与观看记录都留在旧行上。++已有两条接回路径都没有生效:++- 文件身份(`file_identity`)按卷号、inode、大小与修改时间认同一个文件,只覆盖同卷改名或移动。文件换了编码,指纹就对不上。+- ADR-0087 决策二按番号接回,但只在 `disposal='vanished'` 的行里找。旧行要等资源同步「清理」才会被标已消失,「检查」只报数不改。所以「检查 → 扫描」这个顺序下,扫描登记新文件时还没有可接的旧行。++即使之后再清理,接回也只搬个人记录;刮削来的标签、封面和资料留在旧行快照里,新行要从头刮削。不带个人记录的旧行会被直接删除。++## 决策++扫描与推送发现登记一个还没有账本行的视频之前,先按文件身份接回;接不上时,再由 `record_rehome.CodeRelocator` 按番号判断。以下条件全部成立,就把旧行的 `path`、`name` 改成新文件,不新建行:++1. 新文件名解析出番号(`normalise_code_key`),并且新文件所在目录里这个番号的视频只有它一个。+2. 同一扫描根下,这个番号的在库视频(`disposal IS NULL`)恰好一条。+3. 这条旧行的路径 `stat` 明确返回「文件不存在」。读取失败、超时或权限错误都不算。++旧行的 id、标签、封面、资料与个人记录原样保留。探测列(时长、分辨率、编码等)清空,等探测补上。增量扫描把这些行并进探测范围,但不并进新影片范围,所以不会触发 `reattach`、想要清单对账、目录规则和「新增」标记。++已标 `vanished` 的行仍按 ADR-0087 决策二接回。同一番号有多条候选、同目录有多个同番号视频,或旧行在别的媒体根下,这里都不处理,仍交给资源同步与 `reattach`。++## 理由++- 番号是跨版本稳定的键(ADR-0087)。旧文件不在、同番号新文件出现,在单人馆藏里几乎就是换版本或整理目录。报告者手工修复的做法也正是把旧行改到新路径,再删掉重复行。+- 只在同一扫描根里找:别的媒体根可能只是没挂上,那里的 `stat` 同样会返回「不存在」,不能当作文件已删除。+- 这一步不记 ADR-0052 批次,与文件身份接回一致。改指只改路径定位,没有写入新的真相字段;旧路径已不存在,撤回后会指向一个打不开的文件。资料与个人记录都没有移动,不需要快照。++## 被否决的方案++- **扫描时把没标 `vanished` 的旧行也纳入 `reattach`。** `reattach` 会搬走个人记录、删除旧行,刮削标签与封面照样丢失,没有解决报告里的主要损失。+- **「检查文件」顺带标 `vanished`。** 检查只读盘、不改任何东西([运行与配置](../OPERATIONS.md)「资源同步」),用户确认清理之前不该有写入。+- **按番号加时长判断。** 登记时新文件还没探测,拿不到时长。同目录只能有一个同番号视频、同根只能有一条同番号候选,这两条已经排除了分卷与多版本并存的情况。++## 后果++- `ScanResult.relocated_ids` 与 `IngestResult.relocated` 记录被改指的行,库处理的增量扫描据此补探测。+- 预览帧、九宫格等派生产物的来源签名含路径与文件代际,改指后自然失效,会重新生成。+- 同目录里旧字幕的 `asset_subtitle` 行不随之迁移,新目录的字幕会在扫描时重新配对。这一点与文件身份接回相同。+- 网盘目录若偶发对单个文件答「不存在」,同时同番号的新文件刚好出现,旧行会被改指到新文件;之后旧文件再被扫到时,会登记成一条新行。diff --git a/src/peach/library_processing.py b/src/peach/library_processing.pyindex e801c69ec..8f20c84ca 100644--- a/src/peach/library_processing.py+++ b/src/peach/library_processing.py@@ -2033,7 +2033,8 @@ def _scan_sources(config, db_path, *, locations, asset_ids, stage, guard, updateif stage == SCAN_STAGE:result.probed += probe_unmeasured(db_path, ffprobe, source, root, config=config,- on_junk=count_junk, asset_ids=list(registered.new_ids) if incremental else None,+ on_junk=count_junk,+ asset_ids=[*registered.new_ids, *registered.relocated_ids] if incremental else None,report=lambda done, total: update(stage='读取时长与分辨率',checked=result.probed + done, total=result.probed + total, probed=result.probed + done))reattach_in(db_path, registered.new_ids)diff --git a/src/peach/media_probe.py b/src/peach/media_probe.pyindex 87a561eed..5a60ac2cc 100644--- a/src/peach/media_probe.py+++ b/src/peach/media_probe.py@@ -35,10 +35,10 @@ TIMEOUT_SECONDS = 20.0#: 入库时同时探几个。网盘挂载上这是延迟受限的元数据读取,开多了只会把 CloudDrive 压住。WORKERS = 4-UPDATE = (- "UPDATE asset SET duration=?,width=?,height=?,vcodec=?,fps=?,has_audio=?,"- "ctx_length=?,ctx_orient=?,ctx_quality=? WHERE id=?"-)+#: 探测写的列;文件换了版本时 `record_rehome.CodeRelocator` 按这份清空。+PROBED_COLUMNS = ("duration", "width", "height", "vcodec", "fps", "has_audio",+ "ctx_length", "ctx_orient", "ctx_quality")+UPDATE = f"UPDATE asset SET {','.join(f'{column}=?' for column in PROBED_COLUMNS)} WHERE id=?"_UNMEASURED = f"medium='video' AND duration IS NULL AND location=? AND {ACTIVE_ASSET_SQL}"diff --git a/src/peach/record_rehome.py b/src/peach/record_rehome.pyindex 831f55a66..815d36625 100644--- a/src/peach/record_rehome.py+++ b/src/peach/record_rehome.py@@ -10,6 +10,9 @@同一 `(location, path)` 的文件回来了不归这里管:扫描的 upsert 直接清掉那一行的 `disposal`。+旧行还没标已消失时由 `CodeRelocator` 在登记之前处理(ADR-0131):旧文件不在、番号唯一对上,就把+旧行改指新文件,不新建行,也不记批次。+新版本先入库、旧文件后消失时反过来找:资源同步与「同步删除」标完已消失,`reattach_vanished`拿刚标的这几行按同一套判据在全部在库视频里找,唯一命中就当场接回,记同一种批次;候选多个分不出的照旧留成孤儿。@@ -29,7 +32,7 @@ import jsonimport sqlite3import timefrom collections import Counter, defaultdict-from collections.abc import Iterable, Sequence+from collections.abc import Callable, Iterable, Sequencefrom contextlib import closingfrom datetime import datetimefrom pathlib import PureWindowsPath@@ -94,6 +97,79 @@ def _under(path: str, creator: str) -> bool:return any(part.casefold() == folded for part in PureWindowsPath(path).parent.parts)+def _gone(native) -> bool:+ """只有盘上明确答「没有这个文件」才算;读不了、超时都不算。"""+ try:+ native.stat()+ except FileNotFoundError:+ return True+ except OSError:+ return False+ return False+++class CodeRelocator:+ """扫描登记新视频之前,把同番号、文件已不在盘上的在库旧行改指到它(ADR-0131)。++ 目录改名又换了编码时文件身份对不上;旧行还没被资源同步标已消失,`reattach` 也找不到它。+ 只在同一扫描根里找:番号对上恰好一条在库视频、新文件所在目录里这个番号的视频只有它一个、+ 旧路径 `stat` 明确答文件不在,才把旧行的路径改过来。行 id 不变,标签、封面、资料与个人记录+ 原样保留;文件换了,探测列清空等探测补上。其余情况一概不动,交给资源同步与 `reattach`。+ """++ def __init__(self, connection, location: str, root: str, walk_root, *, moved: Iterable[int] = ()):+ from .smb_media import SmbResource+ self.connection = connection+ self.location = location+ self.root = str(PureWindowsPath(root))+ self.walk_root = walk_root+ #: SMB 直连的目录对象没有本机 `stat`,问不出「不在」,不改指。+ self.enabled = not isinstance(walk_root, SmbResource)+ #: 这一轮(含断点之前)改指过的行 id。+ self.moved = list(moved)+ self._by_code: dict[str, list[dict]] | None = None++ def _olds(self, key: str) -> list[dict]:+ """扫描根下按番号分组的在库视频;第一次用到才查,一轮扫描只查一次。"""+ if self._by_code is None:+ from .media_folders import path_predicate+ scope, params = path_predicate(self.location, self.root)+ self._by_code = defaultdict(list)+ for row in _rows(self.connection, f"SELECT a.id,a.path,a.name,a.code FROM asset a WHERE {scope} "+ "AND a.medium='video' AND a.disposal IS NULL", params):+ if found := code_key(row):+ self._by_code[found].append(row)+ return self._by_code.get(key, [])++ def relocate(self, path: str, video_names: Callable[[], Iterable[str]]) -> int | None:+ """`path` 是还没登记的视频,`video_names()` 列出同目录全部视频的文件名;返回改指过来的行 id。++ 名单只在找到唯一候选之后才列:首次扫描的大目录里每个新文件都列一遍会变成平方级。+ """+ name = PureWindowsPath(path).name+ key = code_key({"name": name}) if self.enabled else ""+ olds = self._olds(key) if key else []+ if len(olds) != 1 or sum(code_key({"name": other}) == key for other in video_names()) != 1:+ return None+ old = olds[0]+ try:+ relative = PureWindowsPath(old["path"]).relative_to(self.root)+ except ValueError:+ return None+ if not _gone(self.walk_root.joinpath(*relative.parts)):+ return None+ from .media_probe import PROBED_COLUMNS+ cleared = ",".join(f"{column}=NULL" for column in PROBED_COLUMNS)+ changed = self.connection.execute(+ f"UPDATE asset SET path=?,name=?,{cleared} WHERE id=? AND path=? AND disposal IS NULL",+ (path, name, old["id"], old["path"])).rowcount+ olds.clear()+ if not changed:+ return None+ self.moved.append(old["id"])+ return old["id"]++def _closest(old: dict, candidates: list[dict]) -> dict | None:"""只有一个候选就是它;多个时取时长与旧行最接近的,时长缺或并列返回 None。"""if len(candidates) == 1:diff --git a/src/peach/scan.py b/src/peach/scan.pyindex 981713467..eb5c332b6 100644--- a/src/peach/scan.py+++ b/src/peach/scan.py@@ -6,6 +6,7 @@文件消失、带着个人记录的行由资源同步标 `disposal='vanished'`(ADR-0087)。登记时把它们接回:同一路径的文件回来了,upsert 清掉那一行的 `disposal`;新登记的行由 `record_rehome.reattach`按番号或「创作者 + 文件名主干 + 时长」找它的旧行,把记录搬过来、删掉旧行,记一个可撤回的批次。+旧行还在库、文件已不在盘上时,登记前先由 `record_rehome.CodeRelocator` 按番号把它改指到新文件。接着由 `wants.reconcile` 按番号对一遍「想要」清单,命中的标「已入库」,同样记可撤回的批次。两条不变量都在这里守:@@ -39,6 +40,7 @@ import sqlite3import timefrom collections.abc import Callable, Mapping, Sequencefrom dataclasses import asdict, dataclass+from functools import partialfrom itertools import chainfrom pathlib import Path, PureWindowsPathfrom filelock import FileLock@@ -285,6 +287,28 @@ class IngestResult:reattached: int = 0#: 新行让几条「想要」标了已入库(`wants.reconcile`)。acquired: int = 0+ #: 登记前按番号改指过来的旧行(`record_rehome.CodeRelocator`);有它时 `new` 为假。+ relocated: int | None = None+++def _video_names(entries) -> list[str]:+ """同一目录里视频的文件名,按番号改指时数同目录有几个同番号视频。"""+ return [entry.name for entry in entries if medium_of(entry.name) == "video"]+++def _relocate_ingested(connection, registered, location, ledger_path, here, *, declared_roots, mounts, windows):+ """推送发现登记的视频接不上文件身份时,在它所在的媒体根里按番号找旧行(ADR-0131)。"""+ key = str(ledger_path).casefold()+ root = next((value for value in declared_roots.get(location, ())+ if PureWindowsPath(value) in ledger_path.parents), None)+ if key in registered or medium_of(ledger_path.name) != "video" or root is None:+ return None+ relocator = record_rehome.CodeRelocator(connection, location, root, walk_root_for(+ location, root, declared_roots=declared_roots, mounts=mounts or {}, windows=windows))+ moved = relocator.relocate(str(ledger_path), lambda: [name for name in here or () if medium_of(name) == "video"])+ if moved is not None:+ registered[key] = str(ledger_path)+ return moveddef _directory_stats(directory: Path) -> dict[str, tuple[int, str]]:@@ -385,6 +409,8 @@ def ingest_path(previous = (location, move_from, old_directory / old.name)if identities.relocate(connection, location, str(ledger_path), directory / name, stat, previous=previous):registered[str(ledger_path).casefold()] = str(ledger_path)+ relocated = _relocate_ingested(connection, registered, location, ledger_path, here,+ declared_roots=declared_roots, mounts=mounts, windows=windows)inserted_ids = _upsert_records(connection, [(location, str(ledger_path), name, medium,stat.st_size, mtime, now, now)], registered, conflicts)stored_path = registered[str(ledger_path).casefold()]@@ -407,7 +433,8 @@ def ingest_path(from .library_new_assets import rememberremember(db_path, inserted_ids)return IngestResult(location, str(ledger_path), True, stat.st_size, tracks,- asset_id=asset_id, new=new, reattached=reattached, acquired=acquired)+ asset_id=asset_id, new=new, reattached=reattached, acquired=acquired,+ relocated=relocated)@dataclass(frozen=True)@@ -433,6 +460,8 @@ class ScanResult:junk_isolated: int = 0junk_review: int = 0path_conflicts: int = 0+ #: 登记前按番号改指到新文件的旧行(`record_rehome.CodeRelocator`);不在 `new_ids` 里,增量探测要并进来。+ relocated_ids: tuple[int, ...] = ()def summary(self) -> str:return (f"✓ {self.location}: {self.files:,} 文件 / "@@ -592,7 +621,8 @@ def _scan_resume_state(walk_root, resume_state):if previous.get('mapping') != mapping:previous = {}result = previous.get('result')- completed = ScanResult(**{**result, 'new_ids': tuple(result['new_ids'])}) if result else None+ completed = ScanResult(**{**result, 'new_ids': tuple(result['new_ids']),+ 'relocated_ids': tuple(result.get('relocated_ids', ()))}) if result else Nonereturn mapping, previous, completed@@ -605,8 +635,9 @@ def _record_directory_subtitles(connection, location, directory, here, conflictsreturn subtitles.record(connection, location, unambiguous, now)-def _scan_file_stat(connection, identities, registered, observations, location, ledger_path, directory, entry, *, direct):- """读取本次文件快照,接回已移动的身份,并积累提交后的身份观测。"""+def _scan_file_stat(connection, identities, registered, observations, location, ledger_path, directory, entry, *,+ direct, relocate=lambda path: None):+ """读取本次文件快照,接回已移动的身份或同番号的旧行,并积累提交后的身份观测。"""stat = entry.stat()if identities is not None:native = Path(directory) / entry.name@@ -615,6 +646,9 @@ def _scan_file_stat(connection, identities, registered, observations, location,if identities.relocate(connection, location, ledger_path, native, stat):registered[ledger_path.casefold()] = ledger_pathobservations.append((location, ledger_path, native, stat))+ if (ledger_path.casefold() not in registered and medium_of(entry.name) == "video"+ and relocate(ledger_path) is not None):+ registered[ledger_path.casefold()] = ledger_pathreturn stat@@ -678,7 +712,8 @@ def _scan_location(video_dirs.intersection_update(artwork_parents)save_checkpoint(dict(mapping=mapping, scan_time=now, pending=list(pending), files=files,total_bytes=total, sidecars=skipped, subtitles=tracks,- orphan_subtitles=orphans, new_ids=sorted(discovered), video_dirs=sorted(video_dirs),+ orphan_subtitles=orphans, new_ids=sorted(discovered), relocated_ids=sorted(relocator.moved),+ video_dirs=sorted(video_dirs),current_directory=directory, cursor=cursor, here=dict(here or {})))# 写事务只在每批 upsert 到 commit 之间打开,遍历与 stat 都在两次 commit 之间,不持锁。# 忙等时长与 `LedgerDatabase.connect` 一致。@@ -688,6 +723,8 @@ def _scan_location(registered, conflicts = _registered_paths(connection, location, root)from .file_identity import FileIdentityIndexidentities = FileIdentityIndex(db_path) if location == 'local' and not remote else None+ relocator = record_rehome.CodeRelocator(connection, location, root, walk_root,+ moved=previous.get('relocated_ids', ()))observations = []for conflict in conflicts.values():onerror(conflict)@@ -719,7 +756,8 @@ def _scan_location(continuetry:stat = _scan_file_stat(connection, identities, registered, observations, location,- ledger_path, directory, entry, direct=read_directory is None)+ ledger_path, directory, entry, direct=read_directory is None,+ relocate=partial(relocator.relocate, video_names=partial(_video_names, entries)))except OSError as error:onerror(error)continue@@ -766,7 +804,7 @@ def _scan_location(checked = junk.isolate(db_path, roots=[(location, root)], asset_ids=list(new_ids) if new_only else None)result = ScanResult(location, root, files, total, time.time() - started, gone,tracks, orphans, skipped, new_ids, reattached, acquired,- checked['isolated'], checked['review'], len(conflicts))+ checked['isolated'], checked['review'], len(conflicts), tuple(sorted(relocator.moved)))save_checkpoint({'mapping': mapping, 'result': asdict(result)})progress(files, total, force=True)report(result.summary())diff --git a/tests/test_processing_control.py b/tests/test_processing_control.pyindex 88070872b..da5a95072 100644--- a/tests/test_processing_control.py+++ b/tests/test_processing_control.py@@ -797,6 +797,23 @@ class ProcessingControlTests(unittest.TestCase):self.assertEqual((result.scanned, result.scanned_bytes), (4, 60))self.assertEqual(result.junk_isolated, 2)+ def test_incremental_scan_probes_rows_moved_onto_new_files(self):+ """按番号改指到新文件的旧行不算新行,但文件换了,增量扫描照样要探它。"""+ from peach.library_processing import _scan_sources+ config = PeachConfig(self.root, self.root / 'config.toml', present=True, locations={'local': ('R:\\Media',)})+ scanned = SimpleNamespace(files=2, total_bytes=30, new_ids=(5,), relocated_ids=(7,), junk_isolated=0)+ with patch('peach.library_processing.scan_location', return_value=scanned), \+ patch('peach.library_processing._online_source_root', return_value=self.root), \+ patch('peach.library_processing._ffprobe_path', return_value=None), \+ patch('peach.library_processing.probe_unmeasured', return_value=0) as probe, \+ patch('peach.library_processing.reattach_in') as reattach:+ result = _scan_sources(config, self.db, locations=config.locations, asset_ids=None,+ stage='scan', guard=Mock(), update=lambda **values: None,+ issue=lambda *_args, **_kwargs: None, incremental=True)+ self.assertEqual(probe.call_args.kwargs['asset_ids'], [5, 7])+ self.assertEqual(result.new_ids, (5,), "改指的旧行不进新影片范围")+ reattach.assert_called_once_with(self.db, (5,))+def test_library_scan_reports_conflicting_records_and_processes_both_roots(self):import sqlite3from peach.library_processing import _scan_sourcesdiff --git a/tests/test_record_rehome.py b/tests/test_record_rehome.pyindex 0e40b6ed6..862b0fb6e 100644--- a/tests/test_record_rehome.py+++ b/tests/test_record_rehome.py@@ -318,6 +318,81 @@ class RecordRehomeTests(unittest.TestCase):with self.assertRaisesRegex(ValueError, "已消失"):record_rehome.move(connection, 9, 7, source=record_rehome.MANUAL_SOURCE, rule="manual")+ # 以下:旧行还在库、文件已不在盘上,新文件同番号,登记前把旧行改指过去(ADR-0131)。+ OLD = "R:\\media\\演员\\FNS-257\\FNS-257.mp4"++ def add_old(self, asset_id=7, path=OLD):+ """一条在库旧行,盘上没有它的文件;带刮削标签、探测结果与一次播放。"""+ self.add_live(asset_id, path, code="FNS-257", duration=7208.23)+ with closing(sqlite3.connect(self.db)) as connection:+ connection.execute("UPDATE asset SET width=1920,height=1080,play_count=1 WHERE id=?", (asset_id,))+ connection.execute("INSERT INTO asset_tag(asset_id,tag,source) VALUES(?,'刮削标签','javdb')", (asset_id,))+ connection.commit()++ def scan_new(self):+ with redirect_stdout(io.StringIO()):+ return scan.scan_location(self.db, "local", "R:\\media", declared_roots=DECLARED,+ mounts=self.mounts, windows=False, new_only=True)++ def test_a_renamed_and_reencoded_folder_keeps_its_row(self):+ """#312:目录改名又换了编码,增量扫描不新建行,标签与播放留在原行,探测列等重探。"""+ self.add_old()+ new_path = self.put("演员/FNS-257 新后缀/FNS-257.mp4")+ result = self.scan_new()+ self.assertEqual((result.new_ids, result.relocated_ids), ((), (7,)))+ self.assertEqual(self.query("SELECT id,path,code,duration,width,height,play_count FROM asset"),+ [(7, new_path, "FNS-257", None, None, None, 1)])+ self.assertEqual(self.owner("asset_tag"), [7])+ self.assertEqual(self.query("SELECT count(*) FROM record_rehome"), [(0,)])++ def test_push_discovery_keeps_the_row_too(self):+ self.add_old()+ new_path = self.put("演员/FNS-257 新后缀/FNS-257.mp4")+ result = self.ingest(new_path)+ self.assertEqual((result.asset_id, result.new, result.relocated), (7, False, 7))+ self.assertEqual(self.query("SELECT id,path FROM asset"), [(7, new_path)])++ def test_an_old_file_still_on_disk_is_a_second_copy(self):+ old_path = self.put("演员/FNS-257/FNS-257.mp4")+ self.scan_all()+ new_path = self.put("演员/FNS-257 新后缀/FNS-257.mp4")+ result = self.scan_new()+ self.assertEqual(result.relocated_ids, ())+ self.assertEqual(sorted(row[0] for row in self.query("SELECT path FROM asset")), sorted([old_path, new_path]))++ def test_two_old_rows_with_the_code_are_left_alone(self):+ self.add_old(7)+ self.add_old(8, "R:\\media\\演员\\FNS-257 4K\\FNS-257.mp4")+ new_path = self.put("演员/FNS-257 新后缀/FNS-257.mp4")+ result = self.scan_new()+ self.assertEqual(result.relocated_ids, ())+ self.assertEqual(self.query("SELECT path FROM asset WHERE id IN (7,8) ORDER BY id"),+ [(self.OLD,), ("R:\\media\\演员\\FNS-257 4K\\FNS-257.mp4",)])+ self.assertEqual(len(result.new_ids), 1)+ self.assertEqual(self.query("SELECT path FROM asset WHERE id=?", (result.new_ids[0],)), [(new_path,)])++ def test_two_new_videos_with_the_code_in_one_folder_are_left_alone(self):+ self.add_old()+ self.put("演员/FNS-257 新后缀/FNS-257.mp4")+ self.put("演员/FNS-257 新后缀/FNS-257.mkv")+ result = self.scan_new()+ self.assertEqual((result.relocated_ids, len(result.new_ids)), ((), 2))+ self.assertEqual(self.query("SELECT path FROM asset WHERE id=7"), [(self.OLD,)])++ def test_an_old_row_under_another_media_folder_is_left_alone(self):+ """另一个媒体目录可能只是没挂上:它下面的行答「不在」不能作数。"""+ other = self.root / "other"+ other.mkdir()+ self.mounts["local"] = (str(self.media), str(other))+ self.add_old(path="R:\\other\\FNS-257\\FNS-257.mp4")+ self.put("演员/FNS-257 新后缀/FNS-257.mp4")+ declared = {"local": ("R:\\media", "R:\\other")}+ with redirect_stdout(io.StringIO()):+ result = scan.scan_location(self.db, "local", "R:\\media", declared_roots=declared,+ mounts=self.mounts, windows=False, new_only=True)+ self.assertEqual((result.relocated_ids, len(result.new_ids)), ((), 1))+ self.assertEqual(self.query("SELECT path FROM asset WHERE id=7"), [("R:\\other\\FNS-257\\FNS-257.mp4",)])+if __name__ == "__main__":unittest.main()