1"""Remove unreferenced files from an unpacked PPTX directory.
2 
3Usage: python clean.py <unpacked_dir>
4 
5Example:
6    python clean.py unpacked/
7 
8This script removes:
9- Orphaned slides (not in sldIdLst) and their relationships
10- [trash] directory (unreferenced files)
11- Orphaned .rels files for deleted resources
12- Unreferenced media, embeddings, charts, diagrams, drawings, ink files
13- Unreferenced theme files
14- Unreferenced notes slides
15- Content-Type overrides for deleted files
16"""
17 
18import sys
19from pathlib import Path
20 
21import defusedxml.minidom
22 
23 
24import re
25 
26 
27def get_slides_in_sldidlst(unpacked_dir: Path) -> set[str]:
28    pres_path = unpacked_dir / "ppt" / "presentation.xml"
29    pres_rels_path = unpacked_dir / "ppt" / "_rels" / "presentation.xml.rels"
30 
31    if not pres_path.exists() or not pres_rels_path.exists():
32        return set()
33 
34    rels_dom = defusedxml.minidom.parse(str(pres_rels_path))
35    rid_to_slide = {}
36    for rel in rels_dom.getElementsByTagName("Relationship"):
37        rid = rel.getAttribute("Id")
38        target = rel.getAttribute("Target")
39        rel_type = rel.getAttribute("Type")
40        if "slide" in rel_type and target.startswith("slides/"):
41            rid_to_slide[rid] = target.replace("slides/", "")
42 
43    pres_content = pres_path.read_text(encoding="utf-8")
44    referenced_rids = set(re.findall(r'<p:sldId[^>]*r:id="([^"]+)"', pres_content))
45 
46    return {rid_to_slide[rid] for rid in referenced_rids if rid in rid_to_slide}
47 
48 
49def remove_orphaned_slides(unpacked_dir: Path) -> list[str]:
50    slides_dir = unpacked_dir / "ppt" / "slides"
51    slides_rels_dir = slides_dir / "_rels"
52    pres_rels_path = unpacked_dir / "ppt" / "_rels" / "presentation.xml.rels"
53 
54    if not slides_dir.exists():
55        return []
56 
57    referenced_slides = get_slides_in_sldidlst(unpacked_dir)
58    removed = []
59 
60    for slide_file in slides_dir.glob("slide*.xml"):
61        if slide_file.name not in referenced_slides:
62            rel_path = slide_file.relative_to(unpacked_dir)
63            slide_file.unlink()
64            removed.append(str(rel_path))
65 
66            rels_file = slides_rels_dir / f"{slide_file.name}.rels"
67            if rels_file.exists():
68                rels_file.unlink()
69                removed.append(str(rels_file.relative_to(unpacked_dir)))
70 
71    if removed and pres_rels_path.exists():
72        rels_dom = defusedxml.minidom.parse(str(pres_rels_path))
73        changed = False
74 
75        for rel in list(rels_dom.getElementsByTagName("Relationship")):
76            target = rel.getAttribute("Target")
77            if target.startswith("slides/"):
78                slide_name = target.replace("slides/", "")
79                if slide_name not in referenced_slides:
80                    if rel.parentNode:
81                        rel.parentNode.removeChild(rel)
82                        changed = True
83 
84        if changed:
85            with open(pres_rels_path, "wb") as f:
86                f.write(rels_dom.toxml(encoding="utf-8"))
87 
88    return removed
89 
90 
91def remove_trash_directory(unpacked_dir: Path) -> list[str]:
92    trash_dir = unpacked_dir / "[trash]"
93    removed = []
94 
95    if trash_dir.exists() and trash_dir.is_dir():
96        for file_path in trash_dir.iterdir():
97            if file_path.is_file():
98                rel_path = file_path.relative_to(unpacked_dir)
99                removed.append(str(rel_path))
100                file_path.unlink()
101        trash_dir.rmdir()
102 
103    return removed
104 
105 
106def get_slide_referenced_files(unpacked_dir: Path) -> set:
107    referenced = set()
108    slides_rels_dir = unpacked_dir / "ppt" / "slides" / "_rels"
109 
110    if not slides_rels_dir.exists():
111        return referenced
112 
113    for rels_file in slides_rels_dir.glob("*.rels"):
114        dom = defusedxml.minidom.parse(str(rels_file))
115        for rel in dom.getElementsByTagName("Relationship"):
116            target = rel.getAttribute("Target")
117            if not target:
118                continue
119            target_path = (rels_file.parent.parent / target).resolve()
120            try:
121                referenced.add(target_path.relative_to(unpacked_dir.resolve()))
122            except ValueError:
123                pass
124 
125    return referenced
126 
127 
128def remove_orphaned_rels_files(unpacked_dir: Path) -> list[str]:
129    resource_dirs = ["charts", "diagrams", "drawings"]
130    removed = []
131    slide_referenced = get_slide_referenced_files(unpacked_dir)
132 
133    for dir_name in resource_dirs:
134        rels_dir = unpacked_dir / "ppt" / dir_name / "_rels"
135        if not rels_dir.exists():
136            continue
137 
138        for rels_file in rels_dir.glob("*.rels"):
139            resource_file = rels_dir.parent / rels_file.name.replace(".rels", "")
140            try:
141                resource_rel_path = resource_file.resolve().relative_to(unpacked_dir.resolve())
142            except ValueError:
143                continue
144 
145            if not resource_file.exists() or resource_rel_path not in slide_referenced:
146                rels_file.unlink()
147                rel_path = rels_file.relative_to(unpacked_dir)
148                removed.append(str(rel_path))
149 
150    return removed
151 
152 
153def get_referenced_files(unpacked_dir: Path) -> set:
154    referenced = set()
155 
156    for rels_file in unpacked_dir.rglob("*.rels"):
157        dom = defusedxml.minidom.parse(str(rels_file))
158        for rel in dom.getElementsByTagName("Relationship"):
159            target = rel.getAttribute("Target")
160            if not target:
161                continue
162            target_path = (rels_file.parent.parent / target).resolve()
163            try:
164                referenced.add(target_path.relative_to(unpacked_dir.resolve()))
165            except ValueError:
166                pass
167 
168    return referenced
169 
170 
171def remove_orphaned_files(unpacked_dir: Path, referenced: set) -> list[str]:
172    resource_dirs = ["media", "embeddings", "charts", "diagrams", "tags", "drawings", "ink"]
173    removed = []
174 
175    for dir_name in resource_dirs:
176        dir_path = unpacked_dir / "ppt" / dir_name
177        if not dir_path.exists():
178            continue
179 
180        for file_path in dir_path.glob("*"):
181            if not file_path.is_file():
182                continue
183            rel_path = file_path.relative_to(unpacked_dir)
184            if rel_path not in referenced:
185                file_path.unlink()
186                removed.append(str(rel_path))
187 
188    theme_dir = unpacked_dir / "ppt" / "theme"
189    if theme_dir.exists():
190        for file_path in theme_dir.glob("theme*.xml"):
191            rel_path = file_path.relative_to(unpacked_dir)
192            if rel_path not in referenced:
193                file_path.unlink()
194                removed.append(str(rel_path))
195                theme_rels = theme_dir / "_rels" / f"{file_path.name}.rels"
196                if theme_rels.exists():
197                    theme_rels.unlink()
198                    removed.append(str(theme_rels.relative_to(unpacked_dir)))
199 
200    notes_dir = unpacked_dir / "ppt" / "notesSlides"
201    if notes_dir.exists():
202        for file_path in notes_dir.glob("*.xml"):
203            if not file_path.is_file():
204                continue
205            rel_path = file_path.relative_to(unpacked_dir)
206            if rel_path not in referenced:
207                file_path.unlink()
208                removed.append(str(rel_path))
209 
210        notes_rels_dir = notes_dir / "_rels"
211        if notes_rels_dir.exists():
212            for file_path in notes_rels_dir.glob("*.rels"):
213                notes_file = notes_dir / file_path.name.replace(".rels", "")
214                if not notes_file.exists():
215                    file_path.unlink()
216                    removed.append(str(file_path.relative_to(unpacked_dir)))
217 
218    return removed
219 
220 
221def update_content_types(unpacked_dir: Path, removed_files: list[str]) -> None:
222    ct_path = unpacked_dir / "[Content_Types].xml"
223    if not ct_path.exists():
224        return
225 
226    dom = defusedxml.minidom.parse(str(ct_path))
227    changed = False
228 
229    for override in list(dom.getElementsByTagName("Override")):
230        part_name = override.getAttribute("PartName").lstrip("/")
231        if part_name in removed_files:
232            if override.parentNode:
233                override.parentNode.removeChild(override)
234                changed = True
235 
236    if changed:
237        with open(ct_path, "wb") as f:
238            f.write(dom.toxml(encoding="utf-8"))
239 
240 
241def clean_unused_files(unpacked_dir: Path) -> list[str]:
242    all_removed = []
243 
244    slides_removed = remove_orphaned_slides(unpacked_dir)
245    all_removed.extend(slides_removed)
246 
247    trash_removed = remove_trash_directory(unpacked_dir)
248    all_removed.extend(trash_removed)
249 
250    while True:
251        removed_rels = remove_orphaned_rels_files(unpacked_dir)
252        referenced = get_referenced_files(unpacked_dir)
253        removed_files = remove_orphaned_files(unpacked_dir, referenced)
254 
255        total_removed = removed_rels + removed_files
256        if not total_removed:
257            break
258 
259        all_removed.extend(total_removed)
260 
261    if all_removed:
262        update_content_types(unpacked_dir, all_removed)
263 
264    return all_removed
265 
266 
267if __name__ == "__main__":
268    if len(sys.argv) != 2:
269        print("Usage: python clean.py <unpacked_dir>", file=sys.stderr)
270        print("Example: python clean.py unpacked/", file=sys.stderr)
271        sys.exit(1)
272 
273    unpacked_dir = Path(sys.argv[1])
274 
275    if not unpacked_dir.exists():
276        print(f"Error: {unpacked_dir} not found", file=sys.stderr)
277        sys.exit(1)
278 
279    removed = clean_unused_files(unpacked_dir)
280 
281    if removed:
282        print(f"Removed {len(removed)} unreferenced files:")
283        for f in removed:
284            print(f"  {f}")
285    else:
286        print("No unreferenced files found")
287