MuTabNet/preprocess.py at master · JG1VPP/MuTabNet · GitHub

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
import argparse
import json
import pickle
import sys
from collections import defaultdict
from dataclasses import dataclass, field
from pathlib import Path

from dacite import from_dict
from more_itertools import ilen, iter_index, take
from ruamel.yaml import YAML
from tqdm import tqdm

TAB = "\t"

BOLD = "<b>"
DLOB = "</b>"


@dataclass
class Part:
    image: str
    jsonl: str


@dataclass
class Load:
    parts: list[Part]


@dataclass
class Dump:
    pkl: str


@dataclass
class Range:
    min: int = 0
    max: int = sys.maxsize


@dataclass
class Length:
    html: Range = field(default_factory=Range)
    cell: Range = field(default_factory=Range)


@dataclass
class Preprocess:
    load: Load
    dump: Dump
    replace: dict[tuple, str]


def options():
    args = argparse.ArgumentParser()
    args.add_argument("cfg")
    args = args.parse_args()

    data = YAML().load(path(args.cfg))
    return from_dict(Preprocess, data)


def path(root, *levels):
    path = Path(root).joinpath(*levels)
    return path.expanduser().absolute()


def count_head(html):
    size = next(iter_index(html, "</thead>"), 0)
    head = iter_index(take(size, html), "</td>")

    return ilen(head)


def tokenize(params, tokens):
    html = iter(tokens)
    half = list()

    for tag in html:
        close = take(tag == "<td>", html)
        half.append(tag + "".join(close))

    return half


def classify(params, tokens, grid):
    grid = iter(grid)
    html = list()

    for tag in tokens:
        if tag.startswith("<td"):
            if not (cell := next(grid)).get("bbox"):
                value = tuple(cell.get("tokens"))
                value = params.replace.get(value)
                tag = tag.replace("td", value, 1)

        html.append(tag)

    return html


def unbold(tokens, is_head_cell: bool):
    if is_head_cell:
        tokens = filter(BOLD.__ne__, tokens)
        tokens = filter(DLOB.__ne__, tokens)

    return TAB.join(tokens).strip().split(TAB)


def dump_item(params, part, split, name, html, grid):
    name = str(path(part.image, split, name))
    head = count_head(html)
    dump = []

    for n, cell in enumerate(grid):
        if bbox := cell.get("bbox"):
            text = unbold(cell["tokens"], n < head)
            dump.append(dict(bbox=bbox, text=text))

    html = classify(params, tokenize(params, html), grid)
    assert len(dump) == sum("<td" in tag for tag in html)

    return dict(img_path=name, html=html, cell=dump)


def fix_json(text):
    return text.replace(", ]", "]")


def parse(params, part, row, splits):
    name = row["filename"]
    html = row["html"]["structure"]["tokens"]
    grid = row["html"]["cells"]
    page = row["split"]

    item = dict(name=name, html=html, grid=grid)
    item = dump_item(params, part, page, **item)

    splits[page].append(item)


def jsonl(params, part, splits):
    with path(part.jsonl).open() as f:
        for line in tqdm(f, desc=part.jsonl):
            line = json.loads(fix_json(line))
            parse(params, part, line, splits)


def process(params):
    table_splits = defaultdict(list)

    for part in params.load.parts:
        jsonl(params, part, table_splits)

    data = pickle.dumps(dict(table_splits))
    path(params.dump.pkl).write_bytes(data)


if __name__ == "__main__":
    process(options())