vole-document 0.1.0-alpha.20

Persistent procedural document runtime: byte-exact reconstruction plus a content-addressed procedural seed DAG, queryable observations with provenance, and selective late materialization.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245

#!/usr/bin/env python3
# Structural probe for real100-v1 documents (runs in the pinned `realcorpus`
# Docker service). Read-only: it inspects the *downloaded bytes* with Poppler,
# qpdf and the Python stdlib and prints objective structural facts. Those facts
# justify the manifest's `structural_tags`; the probe never runs the codec and
# never consults a codec result, so a tag can never be influenced by a loss.
#
#   docker compose run --rm --no-TTY realcorpus \
#     python3 tools/realcorpus/probe.py real100-v1/documents/**/*.pdf
#
# Output: one TSV row per file (header first):
#   path format pages bytes text_chars images fonts objstm xref_stream
#   encrypted zip_members media_files spine tables headings max_heading
#   list_items text_chars2 suggested_tags

import json
import os
import re
import subprocess
import sys
import zipfile

IMAGE_EXT = (".png", ".jpg", ".jpeg", ".gif", ".tif", ".tiff", ".bmp",
             ".webp", ".svg", ".emf", ".wmf")


def run(cmd):
    try:
        r = subprocess.run(cmd, capture_output=True, text=True, timeout=180)
        return r.stdout
    except Exception:  # noqa: BLE001
        return ""


def read_head(path, n=1 << 22):
    with open(path, "rb") as f:
        return f.read(n)


def probe_pdf(path):
    facts = {"format": "pdf", "pages": "", "text_chars": "", "images": "",
             "fonts": "", "objstm": "", "xref_stream": "", "encrypted": "",
             "producer": ""}
    info = run(["pdfinfo", path])
    m = re.search(r"^Pages:\s+(\d+)", info, re.M)
    if m:
        facts["pages"] = m.group(1)
    if re.search(r"^Encrypted:\s+yes", info, re.M):
        facts["encrypted"] = "yes"
    mp = re.search(r"^Producer:\s+(.*)$", info, re.M)
    if mp:
        facts["producer"] = mp.group(1).strip()[:80]
    txt = run(["pdftotext", "-q", path, "-"])
    facts["text_chars"] = str(len(re.sub(r"\s+", "", txt)))
    imgs = run(["pdfimages", "-list", path])
    n = 0
    for line in imgs.splitlines()[2:]:
        if line.strip():
            n += 1
    facts["images"] = str(n)
    fonts = run(["pdffonts", path])
    facts["fonts"] = str(max(0, len(fonts.splitlines()) - 2))
    # A PDF with a cross-reference *stream* (rather than a classic table) and/or
    # compressed object streams is the "complex" family the corpus must cover.
    # Scan the tail (xref lives near the end) plus a head slice.
    blob = read_head(path, 1 << 20)
    try:
        with open(path, "rb") as f:
            f.seek(max(0, os.path.getsize(path) - (1 << 20)))
            blob += f.read(1 << 20)
    except OSError:
        pass
    facts["objstm"] = "yes" if b"/ObjStm" in blob else "no"
    facts["xref_stream"] = "yes" if (b"/XRef" in blob or b"/Type/XRef" in blob) else "no"
    return facts


def probe_docx(path):
    facts = {"format": "docx", "zip_members": "", "media_files": "",
             "tables": "", "headings": "", "max_heading": "", "list_items": "",
             "text_chars2": ""}
    try:
        with zipfile.ZipFile(path) as z:
            names = z.namelist()
            facts["zip_members"] = str(len(names))
            facts["media_files"] = str(sum(1 for n in names
                                           if n.startswith("word/media/")
                                           and n.lower().endswith(IMAGE_EXT)))
            doc = z.read("word/document.xml") if "word/document.xml" in names else b""
    except (zipfile.BadZipFile, KeyError, OSError):
        return facts
    facts["tables"] = str(doc.count(b"<w:tbl>"))
    facts["list_items"] = str(doc.count(b"<w:numPr>"))
    heads = re.findall(rb'w:val="Heading(\d)"', doc)
    facts["headings"] = str(len(heads))
    facts["max_heading"] = str(max((int(h) for h in heads), default=0))
    text = re.sub(rb"<[^>]+>", b" ", doc)
    facts["text_chars2"] = str(len(re.sub(rb"\s+", b"", text)))
    return facts


def probe_epub(path):
    facts = {"format": "epub", "zip_members": "", "media_files": "",
             "spine": "", "nav": "", "text_chars2": ""}
    try:
        with zipfile.ZipFile(path) as z:
            names = z.namelist()
            facts["zip_members"] = str(len(names))
            facts["media_files"] = str(sum(1 for n in names
                                           if n.lower().endswith(IMAGE_EXT)))
            opf = None
            if "META-INF/container.xml" in names:
                c = z.read("META-INF/container.xml").decode("utf-8", "replace")
                m = re.search(r'full-path="([^"]+)"', c)
                if m:
                    opf = m.group(1)
            spine = 0
            nav = "no"
            total = 0
            if opf and opf in names:
                doc = z.read(opf).decode("utf-8", "replace")
                spine = len(re.findall(r"<itemref\b", doc))
                nav = "yes" if re.search(r'properties="[^"]*\bnav\b', doc) else nav
            facts["spine"] = str(spine)
            facts["nav"] = nav
            # sample text length across xhtml members (bounded)
            for n in names:
                if n.lower().endswith((".xhtml", ".html", ".htm")):
                    data = z.read(n)[:1 << 20]
                    total += len(re.sub(rb"\s+", b"", re.sub(rb"<[^>]+>", b" ", data)))
                    if total > (1 << 22):
                        break
            facts["text_chars2"] = str(total)
    except (zipfile.BadZipFile, OSError):
        return facts
    return facts


def detect(path):
    with open(path, "rb") as f:
        head = f.read(1024)
    # PDF may carry up to 1024 bytes of leading junk before `%PDF-`.
    if b"%PDF-" in head[:1024]:
        return "pdf"
    if head[:4] == b"PK\x03\x04":
        return _detect_opc(path)
    return None


def _detect_opc(path):
    try:
        with zipfile.ZipFile(path) as z:
            names = set(z.namelist())
            if "mimetype" in names and b"epub" in z.read("mimetype")[:64].lower():
                return "epub"
            if "META-INF/container.xml" in names or "word/document.xml" not in names:
                if "word/document.xml" not in names:
                    return "epub"
            if "word/document.xml" in names:
                return "docx"
            return "epub"
    except (zipfile.BadZipFile, OSError):
        return None


def suggest(f):
    tags = []
    fmt = f.get("format")

    def as_int(k):
        v = f.get(k, "")
        return int(v) if str(v).isdigit() else 0

    if fmt == "pdf":
        pages = as_int("pages")
        tc = as_int("text_chars")
        imgs = as_int("images")
        prod = (f.get("producer") or "").lower()
        # Scanned/legacy: a scan-conversion producer, or an image-heavy file
        # with no embedded fonts. Both are objective byte facts.
        if any(k in prod for k in ("capture", "scan", "ocr")):
            tags.append("scanned")
        elif pages and imgs and imgs >= pages and as_int("fonts") == 0:
            tags.append("scanned")
        if imgs and pages and imgs >= pages:
            tags.append("figure-heavy")
        if f.get("objstm") == "yes" or f.get("xref_stream") == "yes":
            tags.append("complex-xref")
        if pages and pages >= 300:
            tags.append("very-large")
    elif fmt == "docx":
        if as_int("tables") >= 3:
            tags.append("table-heavy")
        if as_int("media_files") >= 3:
            tags.append("image-heavy")
        if as_int("max_heading") >= 3:
            tags.append("deep-headings")
        if as_int("list_items") >= 10:
            tags.append("list-heavy")
    elif fmt == "epub":
        if as_int("spine") >= 10:
            tags.append("multi-chapter")
        if as_int("media_files") >= 20:
            tags.append("image-heavy")
        if f.get("nav") == "yes":
            tags.append("nav-heavy")
    return ";".join(tags)


def main():
    paths = sys.argv[1:]
    if not paths:
        print("usage: probe.py FILE...", file=sys.stderr)
        return 2
    fields = ["path", "format", "pages", "bytes", "text_chars", "images",
              "fonts", "objstm", "xref_stream", "encrypted", "zip_members",
              "media_files", "spine", "nav", "tables", "headings",
              "max_heading", "list_items", "text_chars2", "suggested_tags"]
    print("\t".join(fields))
    rc = 0
    for p in sorted(paths):
        try:
            fmt = detect(p)
            if fmt is None:
                print("%s\tUNKNOWN" % p)
                continue
            if fmt == "pdf":
                f = probe_pdf(p)
            elif fmt == "docx":
                f = probe_docx(p)
            else:
                f = probe_epub(p)
            b = str(os.path.getsize(p))
            f["bytes"] = b
            f["path"] = p
            f["suggested_tags"] = suggest(f)
            print("\t".join(str(f.get(k, "")) for k in fields))
        except Exception as e:  # noqa: BLE001
            print("%s\tERROR\t%s" % (p, e), file=sys.stderr)
            rc = 1
    return rc


if __name__ == "__main__":
    sys.exit(main())