deepclean-mydisk 0.0.0-stage → 0.12.0

package/resources/scripts/photos.py

@@ -0,0 +1,268 @@
+#!/usr/bin/env python3
+"""相似照片扫描（dHash）——只读分析，供插件「相似照片」页面展示与清理。
+
+实现：系统解码器将图片缩成 9x8 → 灰度像素 → dHash64 → 汉明距离分组。
+不依赖第三方库；macOS 用 sips，Windows 用 PowerShell/.NET 系统图片解码器。
+"""
+import argparse
+import json
+import os
+import struct
+import subprocess
+import sys
+import tempfile
+import time
+from pathlib import Path
+
+EXTS = {".jpg", ".jpeg", ".png", ".heic", ".heif", ".gif", ".webp", ".bmp", ".tiff"}
+SKIP_DIRS = {"node_modules", ".git", ".Trash", "Library", "Photos Library.photoslibrary"}
+PACKAGE_SUFFIXES = (
+    ".photoslibrary", ".photolibrary", ".sparsebundle", ".fcpbundle", ".logicx",
+    ".band", ".imovielibrary", ".tvlibrary", ".aplibrary",
+)
+CLOUD_ATTRS = 0x400000 | 0x40000 | 0x1000  # RECALL_ON_DATA_ACCESS | RECALL_ON_OPEN | OFFLINE
+CLOUD_DIR_NAMES = ("onedrive", "dropbox", "google drive", "icloud drive")
+
+
+def cloud_roots():
+    """云盘根：照搬 tree.py 的规则（平台固定位置 + 环境变量 + DEEPCLEAN_CLOUD_ROOTS）。"""
+    roots = []
+    home = Path.home()
+    if sys.platform == "darwin":
+        roots += [home / "Library" / "CloudStorage", home / "Library" / "Mobile Documents"]
+    else:
+        for name in ("OneDrive", "OneDriveConsumer", "OneDriveCommercial"):
+            value = os.environ.get(name)
+            if value:
+                roots.append(Path(value))
+        try:
+            for child in home.iterdir():
+                if child.name.lower().startswith(CLOUD_DIR_NAMES):
+                    roots.append(child)
+        except OSError:
+            pass
+    extra = os.environ.get("DEEPCLEAN_CLOUD_ROOTS")
+    if extra:
+        try:
+            roots += [Path(p) for p in json.loads(extra) if isinstance(p, str)]
+        except (ValueError, TypeError):
+            pass
+    out = []
+    for root in roots:
+        try:
+            out.append(str(root.resolve()))
+        except OSError:
+            continue
+    return out
+
+
+def is_cloud_path(path, roots):
+    try:
+        target = os.path.normcase(str(Path(path).resolve()))
+    except OSError:
+        return False
+    for root in roots:
+        base = os.path.normcase(str(Path(root).resolve()))
+        if target == base or target.startswith(base + os.sep):
+            return True
+    return False
+
+
+def is_package_dir(name):
+    low = name.lower()
+    return any(low.endswith(suffix) for suffix in PACKAGE_SUFFIXES)
+
+
+def _progress(pct, label):
+    try:
+        print(f"PROGRESS {pct} {label}", file=sys.stderr, flush=True)
+    except Exception:
+        pass
+
+
+def bmp_gray(path):
+    """读 sips 产出的 9x8 BMP，返回 9x8 灰度（0-255）。"""
+    b = open(path, "rb").read()
+    if b[:2] != b"BM":
+        raise ValueError("not BMP")
+    off = struct.unpack_from("<I", b, 10)[0]
+    w, h = struct.unpack_from("<ii", b, 18)
+    bpp = struct.unpack_from("<H", b, 28)[0]
+    if w < 2 or abs(h) < 1 or bpp not in (24, 32):
+        raise ValueError(f"bad BMP w={w} h={h} bpp={bpp}")
+    top_down = h < 0
+    rows = abs(h)
+    step = bpp // 8
+    stride = ((w * bpp + 31) // 32) * 4
+    gray = []
+    for y in range(rows):
+        row = y if top_down else rows - 1 - y
+        base = off + row * stride
+        line = []
+        for x in range(w):
+            i = base + x * step
+            blue, green, red = b[i], b[i + 1], b[i + 2]
+            line.append(int(0.299 * red + 0.587 * green + 0.114 * blue))
+        gray.append(line)
+    return gray
+
+
+def windows_command(mode):
+    if mode not in ("hash", "thumbnail"):
+        raise ValueError("Unknown image operation")
+    helper = Path(__file__).with_name("photos-windows.ps1")
+    command = "& {\n" + helper.read_text(encoding="utf-8") + "\n} -Mode " + mode
+    import base64
+    encoded = base64.b64encode(command.encode("utf-16-le")).decode("ascii")
+    return ["powershell.exe", "-NoProfile", "-NonInteractive", "-EncodedCommand", encoded]
+
+
+def windows_hashes(files):
+    """Decode the batch in one PowerShell process; paths travel as UTF-8 JSON, not shell code."""
+    result = subprocess.run(
+        windows_command("hash"),
+        input=json.dumps([f["path"] for f in files], ensure_ascii=False),
+        capture_output=True, text=True, encoding="utf-8", errors="replace", timeout=540)
+    if result.returncode:
+        raise RuntimeError("Windows image decoder failed: " + result.stderr.strip()[:500])
+    rows = json.loads(result.stdout.lstrip("\ufeff"))
+    return {r["path"]: int(r["hash"], 16) for r in rows if r.get("hash")}
+
+
+def dhash(path):
+    with tempfile.TemporaryDirectory(prefix="dc-photo-") as td:
+        out = os.path.join(td, "x.bmp")
+        subprocess.run(["sips", "-z", "8", "9", "-s", "format", "bmp", str(path), "--out", out],
+                       capture_output=True, timeout=30, check=True)
+        g = bmp_gray(out)
+    bits = 0
+    for y in range(8):
+        for x in range(8):
+            bits = (bits << 1) | (1 if g[y][x] > g[y][x + 1] else 0)
+    return bits
+
+
+def hamming(a, b):
+    return bin(a ^ b).count("1")
+
+
+def collect(dirs, min_kb, limit, stats=None):
+    """收集候选图片；云盘目录与 App 包目录跳过（云端去重=删云端文件，包目录移动会破坏 App）。"""
+    stats = stats if stats is not None else {"cloud": 0, "package": 0}
+    clouds = cloud_roots()
+    files = []
+    for root in dirs:
+        root = Path(root)
+        if not root.is_dir():
+            continue
+        if is_cloud_path(root, clouds):
+            stats["cloud"] += 1
+            continue
+        for dirpath, dirnames, filenames in os.walk(root):
+            kept = []
+            for d in dirnames:
+                if d in SKIP_DIRS or d.startswith("."):
+                    continue
+                child = Path(dirpath) / d
+                if is_cloud_path(child, clouds):
+                    stats["cloud"] += 1
+                    continue
+                if is_package_dir(d):
+                    stats["package"] += 1
+                    continue
+                kept.append(d)
+            dirnames[:] = kept
+            for name in filenames:
+                if Path(name).suffix.lower() not in EXTS:
+                    continue
+                p = Path(dirpath) / name
+                try:
+                    st = p.stat()
+                except OSError:
+                    continue
+                if getattr(st, "st_file_attributes", 0) & CLOUD_ATTRS:
+                    stats["cloud"] += 1
+                    continue
+                if st.st_size < min_kb * 1024:
+                    continue
+                files.append({"path": str(p), "kb": st.st_size // 1024, "mtime": int(st.st_mtime), "size": st.st_size})
+    files.sort(key=lambda f: -f["size"])
+    return files[:limit], stats
+
+
+def main():
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--json", required=True)
+    ap.add_argument("--limit", type=int, default=600)
+    ap.add_argument("--min-kb", type=int, default=200)
+    ap.add_argument("--home")
+    args = ap.parse_args()
+    home = Path(args.home).expanduser() if args.home else Path.home()
+    dirs = [home / "Pictures", home / "Downloads", home / "Desktop", home / "Documents"]
+    t0 = time.time()
+    _progress(5, "查找图片")
+    files, stats = collect(dirs, args.min_kb, args.limit)
+    total = len(files)
+    _progress(15, f"计算指纹（0/{total}）")
+    if sys.platform == "win32" and files:
+        hashes = windows_hashes(files)
+        for f in files:
+            f["hash"] = hashes.get(f["path"])
+    else:
+        for i, f in enumerate(files):
+            try:
+                f["hash"] = dhash(f["path"])
+            except (FileNotFoundError, PermissionError):
+                raise RuntimeError("Image decoder unavailable: sips")
+            except (OSError, ValueError, struct.error, subprocess.SubprocessError):
+                f["hash"] = None
+            if i % 25 == 0:
+                _progress(15 + int(70 * i / max(1, total)), f"计算指纹（{i}/{total}）")
+    good = [f for f in files if f["hash"] is not None]
+    if files and not good:
+        raise RuntimeError("No images could be decoded; check image formats and the system decoder")
+    _progress(88, "分组相似照片")
+    parent = list(range(len(good)))
+
+    def find(x):
+        while parent[x] != x:
+            parent[x] = parent[parent[x]]
+            x = parent[x]
+        return x
+
+    def union(a, b):
+        ra, rb = find(a), find(b)
+        if ra != rb:
+            parent[rb] = ra
+
+    for i in range(len(good)):
+        for j in range(i + 1, len(good)):
+            if hamming(good[i]["hash"], good[j]["hash"]) <= 6:
+                union(i, j)
+    groups = {}
+    for i, f in enumerate(good):
+        groups.setdefault(find(i), []).append(f)
+    result = []
+    for gi, items in groups.items():
+        if len(items) < 2:
+            continue
+        items.sort(key=lambda x: -x["size"])
+        total_kb = sum(x["kb"] for x in items)
+        waste = total_kb - items[0]["kb"]
+        result.append({
+            "id": f"g{gi}",
+            "items": [{k: it[k] for k in ("path", "kb", "mtime", "size")} for it in items],
+            "totalKb": total_kb,
+            "wastedKb": waste,
+        })
+    result.sort(key=lambda g: -g["wastedKb"])
+    out = {"generatedAt": int(time.time() * 1000), "scanned": total,
+           "compared": len(good), "skipped": total - len(good), "groups": result,
+           "skippedCloud": stats.get("cloud", 0), "skippedPackages": stats.get("package", 0),
+           "elapsed": round(time.time() - t0, 1)}
+    Path(args.json).write_text(json.dumps(out, ensure_ascii=False, indent=1), encoding="utf-8")
+    _progress(100, "完成")
+
+
+if __name__ == "__main__":
+    main()