UNIPipe.py 2.74 KB
Newer Older
赵小蒙's avatar
赵小蒙 committed
1
2
import json

3
from loguru import logger
赵小蒙's avatar
赵小蒙 committed
4
5
from magic_pdf.dict2md.mkcontent import mk_universal_format, mk_mm_markdown
from magic_pdf.dict2md.ocr_mkcontent import make_standard_format_with_para, ocr_mk_mm_markdown_with_para
6
7
from magic_pdf.filter.pdf_classify_by_type import classify
from magic_pdf.filter.pdf_meta_scan import pdf_meta_scan
kernel.h@qq.com's avatar
kernel.h@qq.com committed
8
9
from magic_pdf.rw.AbsReaderWriter import AbsReaderWriter
from magic_pdf.rw.DiskReaderWriter import DiskReaderWriter
赵小蒙's avatar
赵小蒙 committed
10
from magic_pdf.libs.commons import join_path
11
from magic_pdf.libs.json_compressor import JsonCompressor
赵小蒙's avatar
赵小蒙 committed
12
from magic_pdf.pipe.AbsPipe import AbsPipe
赵小蒙's avatar
赵小蒙 committed
13
from magic_pdf.user_api import parse_union_pdf, parse_ocr_pdf
14
15


赵小蒙's avatar
赵小蒙 committed
16
class UNIPipe(AbsPipe):
17

kernel.h@qq.com's avatar
kernel.h@qq.com committed
18
19
20
    def __init__(self, pdf_bytes: bytes, model_list: list, image_writer: AbsReaderWriter, img_parent_path: str):
        self.pdf_type = self.PIP_OCR
        super().__init__(pdf_bytes, model_list, image_writer, img_parent_path)
21

赵小蒙's avatar
赵小蒙 committed
22
23
    def pipe_classify(self):
        self.pdf_type = UNIPipe.classify(self.pdf_bytes)
24

赵小蒙's avatar
赵小蒙 committed
25
    def pipe_parse(self):
kernel.h@qq.com's avatar
kernel.h@qq.com committed
26
        if self.pdf_type == self.PIP_TXT:
kernel.h@qq.com's avatar
kernel.h@qq.com committed
27
            self.pdf_mid_data = parse_union_pdf(self.pdf_bytes, self.model_list, self.image_writer)
kernel.h@qq.com's avatar
kernel.h@qq.com committed
28
        elif self.pdf_type == self.PIP_OCR:
kernel.h@qq.com's avatar
kernel.h@qq.com committed
29
            self.pdf_mid_data = parse_ocr_pdf(self.pdf_bytes, self.model_list, self.image_writer)
30

赵小蒙's avatar
赵小蒙 committed
31
    def pipe_mk_uni_format(self):
kernel.h@qq.com's avatar
kernel.h@qq.com committed
32
        content_list = AbsPipe.mk_uni_format(self.get_compress_pdf_mid_data(), self.img_parent_path)
赵小蒙's avatar
赵小蒙 committed
33
        return content_list
赵小蒙's avatar
赵小蒙 committed
34

赵小蒙's avatar
赵小蒙 committed
35
    def pipe_mk_markdown(self):
kernel.h@qq.com's avatar
kernel.h@qq.com committed
36
        markdown_content = AbsPipe.mk_markdown(self.get_compress_pdf_mid_data(), self.img_parent_path)
赵小蒙's avatar
赵小蒙 committed
37
        return markdown_content
38
39
40

if __name__ == '__main__':
    # 测试
赵小蒙's avatar
赵小蒙 committed
41
42
    drw = DiskReaderWriter(r"D:/project/20231108code-clean")

赵小蒙's avatar
赵小蒙 committed
43
44
    pdf_file_path = r"linshixuqiu\19983-00.pdf"
    model_file_path = r"linshixuqiu\19983-00.json"
赵小蒙's avatar
赵小蒙 committed
45
46
    pdf_bytes = drw.read(pdf_file_path, AbsReaderWriter.MODE_BIN)
    model_json_txt = drw.read(model_file_path, AbsReaderWriter.MODE_TXT)
赵小蒙's avatar
赵小蒙 committed
47
48
49
50
    model_list = json.loads(model_json_txt)
    write_path = r"D:\project\20231108code-clean\linshixuqiu\19983-00"
    img_bucket_path = "imgs"
    img_writer = DiskReaderWriter(join_path(write_path, img_bucket_path))
赵小蒙's avatar
赵小蒙 committed
51

赵小蒙's avatar
赵小蒙 committed
52
53
54
55
56
57
58
59
    pipe = UNIPipe(pdf_bytes, model_list, img_writer, img_bucket_path)
    pipe.pipe_classify()
    pipe.pipe_parse()
    md_content = pipe.pipe_mk_markdown()
    try:
        content_list = pipe.pipe_mk_uni_format()
    except Exception as e:
        logger.exception(e)
赵小蒙's avatar
赵小蒙 committed
60
61

    md_writer = DiskReaderWriter(write_path)
赵小蒙's avatar
赵小蒙 committed
62
63
64
    md_writer.write(md_content, "19983-00.md", AbsReaderWriter.MODE_TXT)
    md_writer.write(json.dumps(pipe.pdf_mid_data, ensure_ascii=False, indent=4), "19983-00.json", AbsReaderWriter.MODE_TXT)
    md_writer.write(str(content_list), "19983-00.txt", AbsReaderWriter.MODE_TXT)