Merge branch 'master' of github.com:papayalove/Magic-PDF

.github/workflows/benchmark.yml

@@ -5,7 +5,13 @@ name: PDF
 on:
   push:
     branches:
-      - master
+      - "master"
+    paths-ignore:
+      - "cmds/**"
+      - "**.md"
+  pull_request:
+    branches:
+      - "master"
     paths-ignore:
       - "cmds/**"
       - "**.md"
@@ -34,16 +40,26 @@ jobs:
           pip install -r requirements.txt
         fi
 
-
-    - name: benchmark
+    - name: config-net-reset
+      run: |
+        export http_proxy=""
+        export https_proxy=""
+    - name: get-benchmark-result
       run: |
         echo "start test"
-        cd tools && python ocr_badcase.py pdf_json_label_0306.json ocr_dataset.json json_files.zip badcase.json overall.json base_data.json
+        cd tools && python text_badcase.py pdf_json_label_0306.json pdf_json_label_0229.json json_files.zip text_badcase text_overall base_data_text.json --s3_bucket_name llm-process-pperf --s3_file_directory qa-validate/pdf-datasets/badcase --AWS_ACCESS_KEY 7X9CWNHIVOHH3LXRD5WK  --AWS_SECRET_KEY IHLyTsv7h4ArzReLWUGZNKvwqB7CMrRi6e7ZyUt0 --END_POINT_URL http://p-ceph-norm-inside.pjlab.org.cn:80
+        python ocr_badcase.py pdf_json_label_0306.json ocr_dataset.json json_files.zip ocr_badcase ocr_overall base_data_ocr.json --s3_bucket_name llm-process-pperf --s3_file_directory qa-validate/pdf-datasets/badcase --AWS_ACCESS_KEY 7X9CWNHIVOHH3LXRD5WK  --AWS_SECRET_KEY IHLyTsv7h4ArzReLWUGZNKvwqB7CMrRi6e7ZyUt0 --END_POINT_URL http://p-ceph-norm-inside.pjlab.org.cn:80
+  
   notify_to_feishu:
     if: ${{ always() && !cancelled() && contains(needs.*.result, 'failure') && (github.ref_name == 'master') }}
     needs: [pdf-test]
-    runs-on: [pdf]
+    runs-on: pdf
     steps:
     - name: notify
       run: |
-        curl -X POST -H "Content-Type: application/json" -d '{"msg_type":"post","content":{"post":{"zh_cn":{"title":"'${{ github.repository }}' GitHubAction Failed","content":[[{"tag":"text","text":""},{"tag":"a","text":"Please click here for details ","href":"https://github.com/'${{ github.repository }}'/actions/runs/'${GITHUB_RUN_ID}'"},{"tag":"at","user_id":"'${{ secrets.USER_ID }}'"}]]}}}}'  ${{ secrets.WEBHOOK_URL }}
+        curl  ${{ secrets.WEBHOOK_URL }} -H 'Content-Type: application/json'  -d '{
+        "msgtype": "text",
+        "text": {
+            "content": "'${{ github.repository }}' GitHubAction Failed!\n 细节请查看：https://github.com/'${{ github.repository }}'/actions/runs/'${GITHUB_RUN_ID}'"
+        } 
+        }'                                                                                                                            

demo/ocr_demo.py

@@ -115,8 +115,9 @@ def ocr_parse_pdf_core(pdf_bytes, model_output_json_list, book_name, start_page_
 if __name__ == '__main__':
     pdf_path = r"/home/cxu/workspace/Magic-PDF/ocr_demo/j.1540-627x.2006.00176.x.pdf"
     json_file_path = r"/home/cxu/workspace/Magic-PDF/ocr_demo/j.1540-627x.2006.00176.x.json"
-    ocr_local_parse(pdf_path, json_file_path)
-    # book_name = "科数网/edu_00011318"
-    # ocr_online_parse(book_name)
+    # ocr_local_parse(pdf_path, json_file_path)
+    
+    book_name = "数学新星网/edu_00001236"
+    ocr_online_parse(book_name)
     
     pass

demo/text_demo.py

@@ -67,9 +67,7 @@ def demo_classify_by_type(book_name=None, debug_mode=True):
     img_num_list = pdf_meta["imgs_per_page"]
     text_len_list = pdf_meta["text_len_per_page"]
     text_layout_list = pdf_meta["text_layout_per_page"]
-    pdf_path = json_object.get("file_location")
     is_text_pdf, results = classify(
-        pdf_path,
         total_page,
         page_width,
         page_height,
@@ -89,7 +87,7 @@ def demo_meta_scan(book_name=None, debug_mode=True):
     s3_pdf_path = json_object.get("file_location")
     s3_config = get_s3_config_dict(s3_pdf_path)
     pdf_bytes = read_file(s3_pdf_path, s3_config)
-    res = pdf_meta_scan(s3_pdf_path, pdf_bytes)
+    res = pdf_meta_scan(pdf_bytes)
 
     logger.info(json.dumps(res, ensure_ascii=False))
     write_json_to_local(res, book_name)

magic_pdf/cli/magicpdf.py

@@ -21,28 +21,122 @@ python magicpdf.py --json  s3://llm-pdf-text/scihub/xxxx.json?bytes=0,81350
 python magicpdf.py --pdf  /home/llm/Downloads/xxxx.pdf --model /home/llm/Downloads/xxxx.json  或者 python magicpdf.py --pdf  /home/llm/Downloads/xxxx.pdf
 """
 
+import click
+from magic_pdf.libs.config_reader import get_s3_config
+from magic_pdf.libs.path_utils import (
+    parse_s3path,
+    parse_s3_range_params,
+    remove_non_official_s3_args,
+)
+from magic_pdf.libs.config_reader import get_local_dir
+from magic_pdf.io.S3ReaderWriter import S3ReaderWriter, MODE_BIN
+from magic_pdf.io.DiskReaderWriter import DiskReaderWriter
+from magic_pdf.spark.spark_api import parse_union_pdf, parse_txt_pdf, parse_ocr_pdf
+import os
+import json as json_parse
+from datetime import datetime
 
 
+parse_pdf_methods = click.Choice(["ocr", "txt", "auto"])
+
+
+def get_pdf_parse_method(method):
+    if method == "ocr":
+        return parse_ocr_pdf
+    elif method == "txt":
+        return parse_txt_pdf
+    return parse_union_pdf
+
+
+def prepare_env():
+    local_parent_dir = os.path.join(
+        get_local_dir(), "magic-pdf", datetime.now().strftime("%Y-%m-%d-%H-%M-%S")
+    )
+
+    local_image_dir = os.path.join(local_parent_dir, "images")
+    local_md_dir = os.path.join(local_parent_dir, "md")
+    os.makedirs(local_image_dir, exist_ok=True)
+    os.makedirs(local_md_dir, exist_ok=True)
+    return local_image_dir, local_md_dir
 
-import click
 
 @click.group()
 def cli():
     pass
 
+
 @cli.command()
-@click.option('--json', type=str, help='输入一个S3路径')
-def json_command(json):
-    # 这里处理json相关的逻辑
-    print(f'处理JSON: {json}')
+@click.option("--json", type=str, help="输入一个S3路径")
+@click.option(
+    "--method",
+    type=parse_pdf_methods,
+    help="指定解析方法。txt: 文本型 pdf 解析方法， ocr: 光学识别解析 pdf, auto: 程序智能选择解析方法",
+    default="auto",
+)
+def json_command(json, method):
+    if not json.startswith("s3://"):
+        print("usage: python magipdf.py --json s3://some_bucket/some_path")
+        os.exit(1)
+
+    def read_s3_path(s3path):
+        bucket, key = parse_s3path(s3path)
+
+        s3_ak, s3_sk, s3_endpoint = get_s3_config(bucket)
+        s3_rw = S3ReaderWriter(
+            s3_ak, s3_sk, s3_endpoint, "auto", remove_non_official_s3_args(s3path)
+        )
+        may_range_params = parse_s3_range_params(json)
+        if may_range_params is None or 2 != len(may_range_params):
+            byte_start, byte_end = 0, None
+        else:
+            byte_start, byte_end = int(may_range_params[0]), int(may_range_params[1])
+        return s3_rw.read_jsonl(
+            remove_non_official_s3_args(s3path), byte_start, byte_end, MODE_BIN
+        )
+
+    jso = json_parse.loads(read_s3_path(json).decode("utf-8"))
+    pdf_data = read_s3_path(jso["file_location"])
+    local_image_dir, _ = prepare_env()
+
+    local_image_rw = DiskReaderWriter(local_image_dir)
+    parse = get_pdf_parse_method(method)
+    parse(pdf_data, jso["doc_layout_result"], local_image_rw, is_debug=True)
+
 
 @cli.command()
-@click.option('--pdf', type=click.Path(exists=True), required=True, help='PDF文件的路径')
-@click.option('--model', type=click.Path(exists=True), help='模型的路径')
-def pdf_command(pdf, model):
+@click.option(
+    "--pdf", type=click.Path(exists=True), required=True, help="PDF文件的路径"
+)
+@click.option("--model", type=click.Path(exists=True), help="模型的路径")
+@click.option(
+    "--method",
+    type=parse_pdf_methods,
+    help="指定解析方法。txt: 文本型 pdf 解析方法， ocr: 光学识别解析 pdf, auto: 程序智能选择解析方法",
+    default="auto",
+)
+def pdf_command(pdf, model, method):
     # 这里处理pdf和模型相关的逻辑
-    print(f'处理PDF: {pdf}')
-    print(f'加载模型: {model}')
+    if model is None:
+        model = pdf.replace(".pdf", ".json")
+        if not os.path.exists(model):
+            print(f"make sure json file existed and place under {os.dirname(pdf)}")
+            os.eixt(1)
+
+    def read_fn(path):
+        disk_rw = DiskReaderWriter(os.path.dirname(path))
+        return disk_rw.read(os.path.basename(path), MODE_BIN)
+
+    pdf_data = read_fn(pdf)
+    jso = json_parse.loads(read_fn(model).decode("utf-8"))
+
+    local_image_dir, _ = prepare_env()
+    local_image_rw = DiskReaderWriter(local_image_dir)
+    parse = get_pdf_parse_method(method)
+    parse(pdf_data, jso["doc_layout_result"], local_image_rw, is_debug=True)
+
 
-if __name__ == '__main__':
+if __name__ == "__main__":
+    """
+    python magic_pdf/cli/magicpdf.py json-command --json s3://llm-pdf-text/pdf_ebook_and_paper/format/v070/part-66028dd46437-000076.jsonl?bytes=0,308393
+    """
     cli()

magic_pdf/dict2md/mkcontent.py

@@ -2,6 +2,7 @@ import math
 from loguru import logger
 
 from magic_pdf.libs.boxbase import find_bottom_nearest_text_bbox, find_top_nearest_text_bbox
+from magic_pdf.libs.commons import join_path
 from magic_pdf.libs.ocr_content_type import ContentType
 
 TYPE_INLINE_EQUATION = ContentType.InlineEquation
@@ -227,7 +228,7 @@ def __insert_before_para(text, type, element, content_list):
         logger.error(f"Can't find the location of image {element.get('image_path')} in the markdown file, search target is {text}")
          
 
-def mk_universal_format(para_dict: dict):
+def mk_universal_format(para_dict: dict, img_buket_path):
     """
     构造统一格式 https://aicarrier.feishu.cn/wiki/FqmMwcH69iIdCWkkyjvcDwNUnTY
     """
@@ -249,7 +250,7 @@ def mk_universal_format(para_dict: dict):
             for img in all_page_images:
                 content_node = {
                     "type": "image",
-                    "img_path": img['image_path'],
+                    "img_path": join_path(img_buket_path, img['image_path']),
                     "img_alt":"",
                     "img_title":"",
                     "img_caption":""
@@ -258,7 +259,7 @@ def mk_universal_format(para_dict: dict):
             for table in all_page_tables:
                 content_node = {
                     "type": "table",
-                    "img_path": table['image_path'],
+                    "img_path": join_path(img_buket_path, table['image_path']),
                     "table_latex": table.get("text"),
                     "table_title": "",
                     "table_caption": "",

magic_pdf/dict2md/ocr_mkcontent.py

-from magic_pdf.libs.commons import s3_image_save_path, join_path
+from magic_pdf.libs.commons import join_path
 from magic_pdf.libs.language import detect_lang
 from magic_pdf.libs.markdown_utils import ocr_escape_special_markdown_char
 from magic_pdf.libs.ocr_content_type import ContentType
@@ -56,7 +56,7 @@ def ocr_mk_mm_markdown(pdf_info_dict: dict):
                         if not span.get('image_path'):
                             continue
                         else:
-                            content = f"![]({join_path(s3_image_save_path, span['image_path'])})"
+                            content = f"![]({span['image_path']})"
                     else:
                         content = ocr_escape_special_markdown_char(span['content'])  # 转义特殊符号
                         if span['type'] == ContentType.InlineEquation:
@@ -123,7 +123,7 @@ def ocr_mk_markdown_with_para_core(paras_of_layout, mode):
                         content = f"\n$$\n{span['content']}\n$$\n"
                     elif span_type in [ContentType.Image, ContentType.Table]:
                         if mode == 'mm':
-                            content = f"\n![]({join_path(s3_image_save_path, span['image_path'])})\n"
+                            content = f"\n![]({span['image_path']})\n"
                         elif mode == 'nlp':
                             pass
                     if content != '':
@@ -138,10 +138,10 @@ def ocr_mk_markdown_with_para_core(paras_of_layout, mode):
     return page_markdown
 
 
-def para_to_standard_format(para):
+def para_to_standard_format(para, img_buket_path):
     para_content = {}
     if len(para) == 1:
-        para_content = line_to_standard_format(para[0])
+        para_content = line_to_standard_format(para[0], img_buket_path)
     elif len(para) > 1:
         para_text = ''
         inline_equation_num = 0
@@ -171,7 +171,7 @@ def para_to_standard_format(para):
         }
     return para_content
 
-def make_standard_format_with_para(pdf_info_dict: dict):
+def make_standard_format_with_para(pdf_info_dict: dict, img_buket_path: str):
     content_list = []
     for _, page_info in pdf_info_dict.items():
         paras_of_layout = page_info.get("para_blocks")
@@ -179,12 +179,12 @@ def make_standard_format_with_para(pdf_info_dict: dict):
             continue
         for paras in paras_of_layout:
             for para in paras:
-                para_content = para_to_standard_format(para)
+                para_content = para_to_standard_format(para, img_buket_path)
                 content_list.append(para_content)
     return content_list
 
 
-def line_to_standard_format(line):
+def line_to_standard_format(line, img_buket_path):
     line_text = ""
     inline_equation_num = 0
     for span in line['spans']:
@@ -195,13 +195,13 @@ def line_to_standard_format(line):
                 if span['type'] == ContentType.Image:
                     content = {
                         'type': 'image',
-                        'img_path': join_path(s3_image_save_path, span['image_path'])
+                        'img_path': join_path(img_buket_path, span['image_path'])
                     }
                     return content
                 elif span['type'] == ContentType.Table:
                     content = {
                         'type': 'table',
-                        'img_path': join_path(s3_image_save_path, span['image_path'])
+                        'img_path': join_path(img_buket_path, span['image_path'])
                     }
                     return content
         else:

magic_pdf/filter/pdf_classify_by_type.py

@@ -15,6 +15,7 @@ from collections import Counter
 
 import click
 import numpy as np
+from loguru import logger
 
 from magic_pdf.libs.commons import mymax, get_top_percent_list
 from magic_pdf.filter.pdf_meta_scan import scan_max_page, junk_limit_min
@@ -298,7 +299,7 @@ def classify_by_img_narrow_strips(page_width, page_height, img_sz_list):
     return narrow_strip_pages_ratio < 0.5
 
 
-def classify(pdf_path, total_page: int, page_width, page_height, img_sz_list: list, text_len_list: list, img_num_list: list, text_layout_list: list):
+def classify(total_page: int, page_width, page_height, img_sz_list: list, text_len_list: list, img_num_list: list, text_layout_list: list):
     """
     这里的图片和页面长度单位是pts
     :param total_page:
@@ -323,7 +324,7 @@ def classify(pdf_path, total_page: int, page_width, page_height, img_sz_list: li
     elif not any(results.values()):
         return False, results
     else:
-        print(f"WARNING: {pdf_path} is not classified by area and text_len, by_image_area: {results['by_image_area']}, by_text: {results['by_text_len']}, by_avg_words: {results['by_avg_words']}, by_img_num: {results['by_img_num']}, by_text_layout: {results['by_text_layout']}, by_img_narrow_strips: {results['by_img_narrow_strips']}", file=sys.stderr)  # 利用这种情况可以快速找出来哪些pdf比较特殊，针对性修正分类算法
+        logger.warning(f"pdf is not classified by area and text_len, by_image_area: {results['by_image_area']}, by_text: {results['by_text_len']}, by_avg_words: {results['by_avg_words']}, by_img_num: {results['by_img_num']}, by_text_layout: {results['by_text_layout']}, by_img_narrow_strips: {results['by_img_narrow_strips']}", file=sys.stderr)  # 利用这种情况可以快速找出来哪些pdf比较特殊，针对性修正分类算法
         return False, results
 
 
@@ -350,7 +351,7 @@ def main(json_file):
                 is_needs_password = o['is_needs_password']
                 if is_encrypted or total_page == 0 or is_needs_password:  # 加密的，需要密码的，没有页面的，都不处理
                     continue
-                tag = classify(pdf_path, total_page, page_width, page_height, img_sz_list, text_len_list, text_layout_list)
+                tag = classify(total_page, page_width, page_height, img_sz_list, text_len_list, text_layout_list)
                 o['is_text_pdf'] = tag
                 print(json.dumps(o, ensure_ascii=False))
     except Exception as e:

magic_pdf/filter/pdf_meta_scan.py

@@ -287,7 +287,7 @@ def get_language(doc: fitz.Document):
     return language
 
 
-def pdf_meta_scan(s3_pdf_path: str, pdf_bytes: bytes):
+def pdf_meta_scan(pdf_bytes: bytes):
     """
     :param s3_pdf_path:
     :param pdf_bytes: pdf文件的二进制数据
@@ -298,8 +298,8 @@ def pdf_meta_scan(s3_pdf_path: str, pdf_bytes: bytes):
     is_encrypted = doc.is_encrypted
     total_page = len(doc)
     if total_page == 0:
-        logger.warning(f"drop this pdf: {s3_pdf_path}, drop_reason: {DropReason.EMPTY_PDF}")
-        result = {"need_drop": True, "drop_reason": DropReason.EMPTY_PDF}
+        logger.warning(f"drop this pdf, drop_reason: {DropReason.EMPTY_PDF}")
+        result = {"_need_drop": True, "_drop_reason": DropReason.EMPTY_PDF}
         return result
     else:
         page_width_pts, page_height_pts = get_pdf_page_size_pts(doc)
@@ -322,7 +322,6 @@ def pdf_meta_scan(s3_pdf_path: str, pdf_bytes: bytes):
 
         # 最后输出一条json
         res = {
-            "pdf_path": s3_pdf_path,
             "is_needs_password": is_needs_password,
             "is_encrypted": is_encrypted,
             "total_page": total_page,
@@ -350,7 +349,7 @@ def main(s3_pdf_path: str, s3_profile: str):
     """
     try:
         file_content = read_file(s3_pdf_path, s3_profile)
-        pdf_meta_scan(s3_pdf_path, file_content)
+        pdf_meta_scan(file_content)
     except Exception as e:
         print(f"ERROR: {s3_pdf_path}, {e}", file=sys.stderr)
         logger.exception(e)

magic_pdf/io/DiskReaderWriter.py

@@ -5,9 +5,11 @@ from loguru import logger
 
 MODE_TXT = "text"
 MODE_BIN = "binary"
+
+
 class DiskReaderWriter(AbsReaderWriter):
 
-    def __init__(self, parent_path, encoding='utf-8'):
+    def __init__(self, parent_path, encoding="utf-8"):
         self.path = parent_path
         self.encoding = encoding
 
@@ -20,10 +22,10 @@ class DiskReaderWriter(AbsReaderWriter):
             logger.error(f"文件 {abspath} 不存在")
             raise Exception(f"文件 {abspath} 不存在")
         if mode == MODE_TXT:
-            with open(abspath, 'r', encoding = self.encoding) as f:
+            with open(abspath, "r", encoding=self.encoding) as f:
                 return f.read()
         elif mode == MODE_BIN:
-            with open(abspath, 'rb') as f:
+            with open(abspath, "rb") as f:
                 return f.read()
         else:
             raise ValueError("Invalid mode. Use 'text' or 'binary'.")
@@ -37,20 +39,21 @@ class DiskReaderWriter(AbsReaderWriter):
             directory_path = os.path.dirname(abspath)
             os.makedirs(directory_path)
         if mode == MODE_TXT:
-            with open(abspath, 'w', encoding=self.encoding) as f:
+            with open(abspath, "w", encoding=self.encoding) as f:
                 f.write(content)
                 logger.info(f"内容已成功写入 {abspath}")
 
         elif mode == MODE_BIN:
-            with open(abspath, 'wb') as f:
+            with open(abspath, "wb") as f:
                 f.write(content)
                 logger.info(f"内容已成功写入 {abspath}")
         else:
             raise ValueError("Invalid mode. Use 'text' or 'binary'.")
 
-    def read_jsonl(self, path: str, byte_start=0, byte_end=None, encoding='utf-8'):
+    def read_jsonl(self, path: str, byte_start=0, byte_end=None, encoding="utf-8"):
         return self.read(path)
 
+
 # 使用示例
 if __name__ == "__main__":
     file_path = "io/test/example.txt"
@@ -63,5 +66,3 @@ if __name__ == "__main__":
     content = drw.read(path=file_path)
     if content:
         logger.info(f"从 {file_path} 读取的内容: {content}")
-
-

magic_pdf/libs/commons.py

@@ -24,7 +24,7 @@ error_log_path = "s3://llm-pdf-text/err_logs/"
 # json_dump_path = "s3://pdf_books_temp/json_dump/" # 这条路径仅用于临时本地测试,不能提交到main
 json_dump_path = "s3://llm-pdf-text/json_dump/"
 
-s3_image_save_path = "s3://mllm-raw-media/pdf2md_img/" # TODO 基础库不应该有这些存在的路径，应该在业务代码中定义
+# s3_image_save_path = "s3://mllm-raw-media/pdf2md_img/" # 基础库不应该有这些存在的路径，应该在业务代码中定义
 
 
 def get_top_percent_list(num_list, percent):
@@ -120,28 +120,8 @@ def read_file(pdf_path: str, s3_profile):
             return f.read()
 
 
-def get_docx_model_output(pdf_model_output, pdf_model_s3_profile, page_id):
-    if isinstance(pdf_model_output, str):
-        model_output_json_path = join_path(pdf_model_output, f"page_{page_id + 1}.json")  # 模型输出的页面编号从1开始的
-        if os.path.exists(model_output_json_path):
-            json_from_docx = read_file(model_output_json_path, pdf_model_s3_profile)
-            model_output_json = json.loads(json_from_docx)
-        else:
-            try:
-                model_output_json_path = join_path(pdf_model_output, "model.json")
-                with open(model_output_json_path, "r", encoding="utf-8") as f:
-                    model_output_json = json.load(f)
-                    model_output_json = model_output_json["doc_layout_result"][page_id]
-            except:
-                s3_model_output_json_path = join_path(pdf_model_output, f"page_{page_id + 1}.json")
-                s3_model_output_json_path = join_path(pdf_model_output, f"{page_id}.json")
-                #s3_model_output_json_path = join_path(pdf_model_output, f"page_{page_id }.json")
-                # logger.warning(f"model_output_json_path: {model_output_json_path} not found. try to load from s3: {s3_model_output_json_path}")
-
-                s = read_file(s3_model_output_json_path, pdf_model_s3_profile)
-                return json.loads(s)
-
-    elif isinstance(pdf_model_output, list):
+def get_docx_model_output(pdf_model_output, page_id):
+
     model_output_json = pdf_model_output[page_id]
 
     return model_output_json

magic_pdf/libs/config_reader.py

@@ -2,6 +2,7 @@
 根据bucket的名字返回对应的s3 AK， SK，endpoint三元组
 
 """
+
 import json
 import os
 
@@ -10,11 +11,7 @@ from loguru import logger
 from magic_pdf.libs.commons import parse_bucket_key
 
 
-def get_s3_config(bucket_name: str):
-    """
-    ~/magic-pdf.json 读出来
-    """
-
+def read_config():
     home_dir = os.path.expanduser("~")
 
     config_file = os.path.join(home_dir, "magic-pdf.json")
@@ -24,6 +21,14 @@ def get_s3_config(bucket_name: str):
 
     with open(config_file, "r") as f:
         config = json.load(f)
+    return config
+
+
+def get_s3_config(bucket_name: str):
+    """
+    ~/magic-pdf.json 读出来
+    """
+    config = read_config()
 
     bucket_info = config.get("bucket_info")
     if bucket_name not in bucket_info:
@@ -49,5 +54,10 @@ def get_bucket_name(path):
     return bucket
 
 
-if __name__ == '__main__':
+def get_local_dir():
+    config = read_config()
+    return config.get("temp-output-dir", "/tmp")
+
+
+if __name__ == "__main__":
     ak, sk, endpoint = get_s3_config("llm-raw")

magic_pdf/libs/detect_language_from_model.py

+from collections import Counter
+
+from magic_pdf.libs.language import detect_lang
+
+def get_language_from_model(model_list: list):
+    language_lst = []
+    for ocr_page_info in model_list:
+        page_text = ""
+        layout_dets = ocr_page_info["layout_dets"]
+        for layout_det in layout_dets:
+            category_id = layout_det["category_id"]
+            allow_category_id_list = [15]
+            if category_id in allow_category_id_list:
+                page_text += layout_det["text"]
+        page_language = detect_lang(page_text)
+        language_lst.append(page_language)
+    # 统计text_language_list中每种语言的个数
+    count_dict = Counter(language_lst)
+    # 输出text_language_list中出现的次数最多的语言
+    language = max(count_dict, key=count_dict.get)
+    return language

magic_pdf/libs/drop_reason.py

@@ -8,7 +8,7 @@ class DropReason:
     HIGH_COMPUTATIONAL_lOAD_BY_SVGS = "high_computational_load_by_svgs" # 特殊的SVG图，计算量太大，从而丢弃
     HIGH_COMPUTATIONAL_lOAD_BY_TOTAL_PAGES = "high_computational_load_by_total_pages" # 计算量超过负荷，当前方法下计算量消耗过大
     MISS_DOC_LAYOUT_RESULT = "missing doc_layout_result" # 版面分析失败
-    Exception = "exception" # 解析中发生异常
+    Exception = "_exception" # 解析中发生异常
     ENCRYPTED = "encrypted" # PDF是加密的
     EMPTY_PDF = "total_page=0" # PDF页面总数为0
     NOT_IS_TEXT_PDF = "not_is_text_pdf" # 不是文字版PDF，无法直接解析

magic_pdf/libs/hash_utils.py

+import hashlib
+
+
+def compute_md5(file_bytes):
+    hasher = hashlib.md5()
+    hasher.update(file_bytes)
+    return hasher.hexdigest().upper()
+
+
+def compute_sha256(input_string):
+    hasher = hashlib.sha256()
+    # 在Python3中，需要将字符串转化为字节对象才能被哈希函数处理
+    input_bytes = input_string.encode('utf-8')
+    hasher.update(input_bytes)
+    return hasher.hexdigest()

magic_pdf/libs/path_utils.py

+
+
+from s3pathlib import S3Path
+
+def remove_non_official_s3_args(s3path):
+    """
+    example: s3://abc/xxxx.json?bytes=0,81350 ==> s3://abc/xxxx.json
+    """
+    arr = s3path.split("?")
+    return arr[0]
+
+def parse_s3path(s3path: str):
+    p = S3Path(remove_non_official_s3_args(s3path))
+    return p.bucket, p.key
+
+def parse_s3_range_params(s3path: str):
+    """
+    example: s3://abc/xxxx.json?bytes=0,81350 ==> [0, 81350]
+    """
+    arr = s3path.split("?bytes=")
+    if len(arr) == 1:
+        return None
+    return arr[1].split(",")

magic_pdf/libs/pdf_image_tools.py

-import os
-from pathlib import Path
-from typing import Tuple
-import io
 
-# from app.common.s3 import get_s3_client
 from magic_pdf.libs.commons import fitz
 from loguru import logger
-from magic_pdf.libs.commons import parse_bucket_key, join_path
+from magic_pdf.libs.commons import join_path
+from magic_pdf.libs.hash_utils import compute_sha256
 
 
-def cut_image(bbox: Tuple, page_num: int, page: fitz.Page, save_parent_path: str, s3_return_path=None, img_s3_client=None, upload_switch=True):
+def cut_image(bbox: tuple, page_num: int, page: fitz.Page, return_path, imageWriter):
     """
     从第page_num页的page中，根据bbox进行裁剪出一张jpg图片，返回图片路径
     save_path：需要同时支持s3和本地, 图片存放在save_path下，文件名是: {page_num}_{bbox[0]}_{bbox[1]}_{bbox[2]}_{bbox[3]}.jpg , bbox内数字取整。
     """
     # 拼接文件名
-    filename = f"{page_num}_{int(bbox[0])}_{int(bbox[1])}_{int(bbox[2])}_{int(bbox[3])}.jpg"
-    # 拼接路径
-    image_save_path = join_path(save_parent_path, filename)
-    s3_img_path = join_path(s3_return_path, filename) if s3_return_path is not None else None
-    # 打印图片文件名
-    # print(f"Saved {image_save_path}")
-
-    #检查坐标
-    # x_check = int(bbox[2]) - int(bbox[0])
-    # y_check = int(bbox[3]) - int(bbox[1])
-    # if x_check <= 0 or y_check <= 0:
-    #
-    #     if image_save_path.startswith("s3://"):
-    #         logger.exception(f"传入图片坐标有误，x1<x0或y1<y0,{s3_img_path}")
-    #         return s3_img_path
-    #     else:
-    #         logger.exception(f"传入图片坐标有误，x1<x0或y1<y0,{image_save_path}")
-    #         return image_save_path
+    filename = f"{page_num}_{int(bbox[0])}_{int(bbox[1])}_{int(bbox[2])}_{int(bbox[3])}"
 
+    # 老版本返回不带bucket的路径
+    img_path = join_path(return_path, filename) if return_path is not None else None
+
+    # 新版本生成平铺路径
+    img_hash256_path = f"{compute_sha256(img_path)}.jpg"
 
     # 将坐标转换为fitz.Rect对象
     rect = fitz.Rect(*bbox)
@@ -42,39 +26,17 @@ def cut_image(bbox: Tuple, page_num: int, page: fitz.Page, save_parent_path: str
     # 截取图片
     pix = page.get_pixmap(clip=rect, matrix=zoom)
 
-    if image_save_path.startswith("s3://"):
-        if not upload_switch:
-            pass
-        else:
-            # 图片保存到s3
-            bucket_name, bucket_key = parse_bucket_key(image_save_path)
-            # 将字节流上传到s3
     byte_data = pix.tobytes(output='jpeg', jpg_quality=95)
-            file_obj = io.BytesIO(byte_data)
-            if img_s3_client is not None:
-                img_s3_client.upload_fileobj(file_obj, bucket_name, bucket_key)
-                # 每个图片上传任务都创建一个新的client
-                # img_s3_client_once = get_s3_client(image_save_path)
-                # img_s3_client_once.upload_fileobj(file_obj, bucket_name, bucket_key)
-            else:
-                logger.exception("must input img_s3_client")
-        return s3_img_path
-    else:
-        # 保存图片到本地
-        # 先检查一下image_save_path的父目录是否存在，如果不存在，就创建
-        parent_dir = os.path.dirname(image_save_path)
-        if not os.path.exists(parent_dir):
-            os.makedirs(parent_dir)
-        pix.save(image_save_path, jpg_quality=95)
-        # 为了直接能在markdown里看，这里把地址改为相对于mardown的地址
-        pth = Path(image_save_path)
-        image_save_path = f"{pth.parent.name}/{pth.name}"
-        return image_save_path
-
-
-def save_images_by_bboxes(book_name: str, page_num: int, page: fitz.Page, save_path: str,
-                            image_bboxes: list, images_overlap_backup:list, table_bboxes: list, equation_inline_bboxes: list,
-                            equation_interline_bboxes: list, img_s3_client) -> dict:
+
+    imageWriter.write(data=byte_data, path=img_hash256_path, mode="binary")
+
+    return img_hash256_path
+
+
+def save_images_by_bboxes(page_num: int, page: fitz.Page, pdf_bytes_md5: str,
+                          image_bboxes: list, images_overlap_backup: list, table_bboxes: list,
+                          equation_inline_bboxes: list,
+                          equation_interline_bboxes: list, imageWriter) -> dict:
     """
     返回一个dict, key为bbox, 值是图片地址
     """
@@ -85,53 +47,30 @@ def save_images_by_bboxes(book_name: str, page_num: int, page: fitz.Page, save_p
     interline_eq_info = []
 
     # 图片的保存路径组成是这样的： {s3_or_local_path}/{book_name}/{images|tables|equations}/{page_num}_{bbox[0]}_{bbox[1]}_{bbox[2]}_{bbox[3]}.jpg
-    s3_return_image_path = join_path(book_name, "images")
-    image_save_path = join_path(save_path, s3_return_image_path)
-
-    s3_return_table_path = join_path(book_name, "tables")
-    table_save_path = join_path(save_path, s3_return_table_path)
-
-    s3_return_equations_inline_path = join_path(book_name, "equations_inline")
-    equation_inline_save_path = join_path(save_path, s3_return_equations_inline_path)
-
-    s3_return_equation_interline_path = join_path(book_name, "equation_interline")
-    equation_interline_save_path = join_path(save_path, s3_return_equation_interline_path)
 
+    def return_path(type):
+        return join_path(pdf_bytes_md5, type)
 
     for bbox in image_bboxes:
-        if any([bbox[0]>=bbox[2], bbox[1]>=bbox[3]]):
+        if any([bbox[0] >= bbox[2], bbox[1] >= bbox[3]]):
             logger.warning(f"image_bboxes: 错误的box, {bbox}")
             continue
 
-        image_path = cut_image(bbox, page_num, page, image_save_path, s3_return_image_path, img_s3_client)
+        image_path = cut_image(bbox, page_num, page, return_path("images"), imageWriter)
         image_info.append({"bbox": bbox, "image_path": image_path})
 
     for bbox in images_overlap_backup:
-        if any([bbox[0]>=bbox[2], bbox[1]>=bbox[3]]):
+        if any([bbox[0] >= bbox[2], bbox[1] >= bbox[3]]):
             logger.warning(f"images_overlap_backup: 错误的box, {bbox}")
             continue
-        image_path = cut_image(bbox, page_num, page, image_save_path, s3_return_image_path, img_s3_client)
+        image_path = cut_image(bbox, page_num, page, return_path("images"), imageWriter)
         image_backup_info.append({"bbox": bbox, "image_path": image_path})
 
     for bbox in table_bboxes:
-        if any([bbox[0]>=bbox[2], bbox[1]>=bbox[3]]):
+        if any([bbox[0] >= bbox[2], bbox[1] >= bbox[3]]):
             logger.warning(f"table_bboxes: 错误的box, {bbox}")
             continue
-        image_path = cut_image(bbox, page_num, page, table_save_path, s3_return_table_path, img_s3_client)
+        image_path = cut_image(bbox, page_num, page, return_path("tables"), imageWriter)
         table_info.append({"bbox": bbox, "image_path": image_path})
 
-    for bbox in equation_inline_bboxes:
-        if any([bbox[0]>=bbox[2], bbox[1]>=bbox[3]]):
-            logger.warning(f"equation_inline_bboxes: 错误的box, {bbox}")
-            continue
-        image_path = cut_image(bbox[:4], page_num, page, equation_inline_save_path, s3_return_equations_inline_path, img_s3_client, upload_switch=False)
-        inline_eq_info.append({'bbox':bbox[:4], "image_path":image_path, "latex_text":bbox[4]})
-
-    for bbox in equation_interline_bboxes:
-        if any([bbox[0]>=bbox[2], bbox[1]>=bbox[3]]):
-            logger.warning(f"equation_interline_bboxes: 错误的box, {bbox}")
-            continue
-        image_path = cut_image(bbox[:4], page_num, page, equation_interline_save_path, s3_return_equation_interline_path, img_s3_client, upload_switch=False)
-        interline_eq_info.append({"bbox":bbox[:4], "image_path":image_path, "latex_text":bbox[4]})
-
     return image_info, image_backup_info, table_info, inline_eq_info, interline_eq_info
\ No newline at end of file

magic_pdf/pdf_parse_by_ocr.py

-import json
-import os
 import time
-
 from loguru import logger
-
-from magic_pdf.libs.draw_bbox import draw_layout_bbox, draw_text_bbox
 from magic_pdf.libs.commons import (
-    read_file,
-    join_path,
     fitz,
-    get_img_s3_client,
     get_delta_time,
     get_docx_model_output,
 )
 from magic_pdf.libs.coordinate_transform import get_scale_ratio
 from magic_pdf.libs.drop_tag import DropTag
+from magic_pdf.libs.hash_utils import compute_md5
 from magic_pdf.libs.ocr_content_type import ContentType
-from magic_pdf.libs.safe_filename import sanitize_filename
 from magic_pdf.para.para_split import para_split
 from magic_pdf.pre_proc.construct_page_dict import ocr_construct_page_component
 from magic_pdf.pre_proc.detect_footer_by_model import parse_footers
@@ -38,38 +30,16 @@ from magic_pdf.pre_proc.remove_bbox_overlap import remove_overlap_between_bbox
 def parse_pdf_by_ocr(
         pdf_bytes,
         pdf_model_output,
-        save_path,
-        book_name,
-        pdf_model_profile=None,
-        image_s3_config=None,
+        imageWriter,
         start_page_id=0,
         end_page_id=None,
         debug_mode=False,
 ):
-
-    save_tmp_path = os.path.join(os.path.dirname(__file__), "../..", "tmp", "unittest")
-    book_name = sanitize_filename(book_name)
-    md_bookname_save_path = ""
-    if debug_mode:
-        save_path = join_path(save_tmp_path, "md")
-        pdf_local_path = join_path(save_tmp_path, "download-pdfs", book_name)
-
-        if not os.path.exists(os.path.dirname(pdf_local_path)):
-            # 如果目录不存在，创建它
-            os.makedirs(os.path.dirname(pdf_local_path))
-
-        md_bookname_save_path = join_path(save_tmp_path, "md", book_name)
-        if not os.path.exists(md_bookname_save_path):
-            # 如果目录不存在，创建它
-            os.makedirs(md_bookname_save_path)
-
-        with open(pdf_local_path + ".pdf", "wb") as pdf_file:
-            pdf_file.write(pdf_bytes)
+    pdf_bytes_md5 = compute_md5(pdf_bytes)
 
     pdf_docs = fitz.open("pdf", pdf_bytes)
     # 初始化空的pdf_info_dict
     pdf_info_dict = {}
-    img_s3_client = get_img_s3_client(save_path, image_s3_config)
 
     start_time = time.time()
 
@@ -91,16 +61,14 @@ def parse_pdf_by_ocr(
 
         # 获取当前页的模型数据
         ocr_page_info = get_docx_model_output(
-            pdf_model_output, pdf_model_profile, page_id
+            pdf_model_output, page_id
         )
 
         """从json中获取每页的页码、页眉、页脚的bbox"""
         page_no_bboxes = parse_pageNos(page_id, page, ocr_page_info)
         header_bboxes = parse_headers(page_id, page, ocr_page_info)
         footer_bboxes = parse_footers(page_id, page, ocr_page_info)
-        footnote_bboxes = parse_footnotes_by_model(
-            page_id, page, ocr_page_info, md_bookname_save_path, debug_mode=debug_mode
-        )
+        footnote_bboxes = parse_footnotes_by_model(page_id, page, ocr_page_info, debug_mode=debug_mode)
 
         # 构建需要remove的bbox字典
         need_remove_spans_bboxes_dict = {
@@ -179,7 +147,7 @@ def parse_pdf_by_ocr(
         spans, dropped_spans_by_removed_bboxes = remove_spans_by_bboxes_dict(spans, need_remove_spans_bboxes_dict)
 
         '''对image和table截图'''
-        spans = cut_image_and_table(spans, page, page_id, book_name, save_path, img_s3_client)
+        spans = cut_image_and_table(spans, page, page_id, pdf_bytes_md5, imageWriter)
 
         '''行内公式调整, 高度调整至与同行文字高度一致(优先左侧, 其次右侧)'''
         displayed_list = []
@@ -242,16 +210,4 @@ def parse_pdf_by_ocr(
     """分段"""
     para_split(pdf_info_dict, debug_mode=debug_mode)
 
-    '''在测试时,保存调试信息'''
-    if debug_mode:
-        params_file_save_path = join_path(
-            save_tmp_path, "md", book_name, "preproc_out.json"
-        )
-        with open(params_file_save_path, "w", encoding="utf-8") as f:
-            json.dump(pdf_info_dict, f, ensure_ascii=False, indent=4)
-
-        # drow_bbox
-        draw_layout_bbox(pdf_info_dict, pdf_bytes, md_bookname_save_path)
-        draw_text_bbox(pdf_info_dict, pdf_bytes, md_bookname_save_path)
-
     return pdf_info_dict

magic_pdf/pdf_parse_by_txt.py

@@ -12,6 +12,7 @@ from magic_pdf.layout.bbox_sort import (
 )
 from magic_pdf.layout.layout_sort import LAYOUT_UNPROC, get_bboxes_layout, get_columns_cnt_of_layout, sort_text_block
 from magic_pdf.libs.drop_reason import DropReason
+from magic_pdf.libs.hash_utils import compute_md5
 from magic_pdf.libs.markdown_utils import escape_special_markdown_char
 from magic_pdf.libs.safe_filename import sanitize_filename
 from magic_pdf.libs.vis_utils import draw_bbox_on_page, draw_layout_bbox_on_page
@@ -73,46 +74,20 @@ paraMergeException_msg = ParaMergeException().message
 def parse_pdf_by_txt(
     pdf_bytes,
     pdf_model_output,
-    save_path,
-    book_name,
-    pdf_model_profile=None,
-    image_s3_config=None,
+    imageWriter,
     start_page_id=0,
     end_page_id=None,
-    junk_img_bojids=[],
     debug_mode=False,
 ):
-
-    save_tmp_path = os.path.join(os.path.dirname(__file__), "../..", "tmp", "unittest")
-    md_bookname_save_path = ""
-    book_name = sanitize_filename(book_name)
-    if debug_mode:
-        save_path = join_path(save_tmp_path, "md")
-        pdf_local_path = join_path(save_tmp_path, "download-pdfs", book_name)
-
-        if not os.path.exists(os.path.dirname(pdf_local_path)):
-            # 如果目录不存在，创建它
-            os.makedirs(os.path.dirname(pdf_local_path))
-
-        md_bookname_save_path = join_path(save_tmp_path, "md", book_name)
-        if not os.path.exists(md_bookname_save_path):
-            # 如果目录不存在，创建它
-            os.makedirs(md_bookname_save_path)
-
-        with open(pdf_local_path + ".pdf", "wb") as pdf_file:
-            pdf_file.write(pdf_bytes)
+    pdf_bytes_md5 = compute_md5(pdf_bytes)
 
     pdf_docs = fitz.open("pdf", pdf_bytes)
     pdf_info_dict = {}
-    img_s3_client = get_img_s3_client(save_path, image_s3_config)  # 更改函数名和参数，避免歧义
-    # img_s3_client = "img_s3_client"  #不创建这个对象，直接用字符串占位
-
     start_time = time.time()
 
     """通过统计pdf全篇文字,识别正文字体"""
     main_text_font = get_main_text_font(pdf_docs)
 
-
     end_page_id = end_page_id if end_page_id else len(pdf_docs) - 1
     for page_id in range(start_page_id, end_page_id + 1):
         page = pdf_docs[page_id]
@@ -128,20 +103,11 @@ def parse_pdf_by_txt(
         # 对单页面非重复id的img数量做统计,如果当前页超过1500则直接return need_drop
         """
         page_imgs = page.get_images()
-        img_counts = 0
-        for img in page_imgs:
-            img_bojid = img[0]
-            if img_bojid in junk_img_bojids:  # 判断这个图片在不在junklist中
-                continue  # 如果在junklist就不用管了，跳过
-            else:
-                recs = page.get_image_rects(img, transform=True)
-                if recs:  # 如果这张图在当前页面有展示
-                    img_counts += 1
-        if img_counts >= 1500:  # 如果去除了junkimg的影响，单页img仍然超过1500的话，就排除当前pdf
-            logger.warning(
-                f"page_id: {page_id}, img_counts: {img_counts}, drop this pdf: {book_name}, drop_reason: {DropReason.HIGH_COMPUTATIONAL_lOAD_BY_IMGS}"
-            )
-            result = {"need_drop": True, "drop_reason": DropReason.HIGH_COMPUTATIONAL_lOAD_BY_IMGS}
+
+        # 去除对junkimg的依赖，简化逻辑
+        if len(page_imgs) > 1500:  # 如果当前页超过1500张图片，直接跳过
+            logger.warning(f"page_id: {page_id}, img_counts: {len(page_imgs)}, drop this pdf")
+            result = {"_need_drop": True, "_drop_reason": DropReason.HIGH_COMPUTATIONAL_lOAD_BY_IMGS}
             if not debug_mode:
                 return result
 
@@ -154,10 +120,10 @@ def parse_pdf_by_txt(
             "dict",
             flags=fitz.TEXTFLAGS_TEXT,
         )["blocks"]
-        model_output_json = get_docx_model_output(pdf_model_output, pdf_model_profile, page_id)
+        model_output_json = get_docx_model_output(pdf_model_output, page_id)
 
         # 解析图片
-        image_bboxes = parse_images(page_id, page, model_output_json, junk_img_bojids)
+        image_bboxes = parse_images(page_id, page, model_output_json)
         image_bboxes = fix_image_vertical(image_bboxes, text_raw_blocks)  # 修正图片的位置
         image_bboxes = fix_seperated_image(image_bboxes)  # 合并有边重合的图片
         image_bboxes = include_img_title(text_raw_blocks, image_bboxes)  # 向图片上方和下方寻找title，使用规则进行匹配，暂时只支持英文规则
@@ -225,22 +191,18 @@ def parse_pdf_by_txt(
         """
         ==================================================================================================================================
         """
-        if debug_mode:  # debugmode截图到本地
-            save_path = join_path(save_tmp_path, "md")
 
         # 把图、表、公式都进行截图，保存到存储上，返回图片路径作为内容
         image_info, image_backup_info, table_info, inline_eq_info, interline_eq_info = save_images_by_bboxes(
-            book_name,
             page_id,
             page,
-            save_path,
+            pdf_bytes_md5,
             image_bboxes,
             images_overlap_backup,
             table_bboxes,
             equations_inline_bboxes,
             equations_interline_bboxes,
-            # 传入img_s3_client
-            img_s3_client,
+            imageWriter
         )  # 只要表格和图片的截图
         
         """"以下进入到公式替换环节 """
@@ -253,13 +215,13 @@ def parse_pdf_by_txt(
 
         """去掉footnote, 从文字和图片中(先去角标再去footnote试试)"""
         # 通过模型识别到的footnote
-        footnote_bboxes_by_model = parse_footnotes_by_model(page_id, page, model_output_json, md_bookname_save_path, debug_mode=debug_mode)
+        footnote_bboxes_by_model = parse_footnotes_by_model(page_id, page, model_output_json, debug_mode=debug_mode)
         # 通过规则识别到的footnote
         footnote_bboxes_by_rule = parse_footnotes_by_rule(remain_text_blocks, page_height, page_id, main_text_font)
         """进入pdf过滤器，去掉一些不合理的pdf"""
         is_good_pdf, err = pdf_filter(page, remain_text_blocks, table_bboxes, image_bboxes)
         if not is_good_pdf:
-            logger.warning(f"page_id: {page_id}, drop this pdf: {book_name}, reason: {err}")
+            logger.warning(f"page_id: {page_id}, drop this pdf: {pdf_bytes_md5}, reason: {err}")
             if not debug_mode:
                 return err
 
@@ -273,8 +235,8 @@ def parse_pdf_by_txt(
 
         if is_text_block_horz_overlap:
             # debug_show_bbox(pdf_docs, page_id, [b['bbox'] for b in remain_text_blocks], [], [], join_path(save_path, book_name, f"{book_name}_debug.pdf"), 0)
-            logger.warning(f"page_id: {page_id}, drop this pdf: {book_name}, reason: {DropReason.TEXT_BLCOK_HOR_OVERLAP}")
-            result = {"need_drop": True, "drop_reason": DropReason.TEXT_BLCOK_HOR_OVERLAP}
+            logger.warning(f"page_id: {page_id}, drop this pdf: {pdf_bytes_md5}, reason: {DropReason.TEXT_BLCOK_HOR_OVERLAP}")
+            result = {"_need_drop": True, "_drop_reason": DropReason.TEXT_BLCOK_HOR_OVERLAP}
             if not debug_mode:
                 return result
 
@@ -292,24 +254,24 @@ def parse_pdf_by_txt(
         layout_bboxes, layout_tree = get_bboxes_layout(all_bboxes, page_boundry, page_id)
         
         if len(remain_text_blocks)>0 and len(all_bboxes)>0 and len(layout_bboxes)==0:
-            logger.warning(f"page_id: {page_id}, drop this pdf: {book_name}, reason: {DropReason.CAN_NOT_DETECT_PAGE_LAYOUT}")
-            result = {"need_drop": True, "drop_reason": DropReason.CAN_NOT_DETECT_PAGE_LAYOUT}
+            logger.warning(f"page_id: {page_id}, drop this pdf: {pdf_bytes_md5}, reason: {DropReason.CAN_NOT_DETECT_PAGE_LAYOUT}")
+            result = {"_need_drop": True, "_drop_reason": DropReason.CAN_NOT_DETECT_PAGE_LAYOUT}
             if not debug_mode:
                 return result
 
         """以下去掉复杂的布局和超过2列的布局"""
         if any([lay["layout_label"] == LAYOUT_UNPROC for lay in layout_bboxes]):  # 复杂的布局
-            logger.warning(f"page_id: {page_id}, drop this pdf: {book_name}, reason: {DropReason.COMPLICATED_LAYOUT}")
-            result = {"need_drop": True, "drop_reason": DropReason.COMPLICATED_LAYOUT}
+            logger.warning(f"page_id: {page_id}, drop this pdf: {pdf_bytes_md5}, reason: {DropReason.COMPLICATED_LAYOUT}")
+            result = {"_need_drop": True, "_drop_reason": DropReason.COMPLICATED_LAYOUT}
             if not debug_mode:
                 return result
 
         layout_column_width = get_columns_cnt_of_layout(layout_tree)
         if layout_column_width > 2:  # 去掉超过2列的布局pdf
-            logger.warning(f"page_id: {page_id}, drop this pdf: {book_name}, reason: {DropReason.TOO_MANY_LAYOUT_COLUMNS}")
+            logger.warning(f"page_id: {page_id}, drop this pdf: {pdf_bytes_md5}, reason: {DropReason.TOO_MANY_LAYOUT_COLUMNS}")
             result = {
-                "need_drop": True,
-                "drop_reason": DropReason.TOO_MANY_LAYOUT_COLUMNS,
+                "_need_drop": True,
+                "_drop_reason": DropReason.TOO_MANY_LAYOUT_COLUMNS,
                 "extra_info": {"column_cnt": layout_column_width},
             }
             if not debug_mode:
@@ -390,28 +352,11 @@ def parse_pdf_by_txt(
     for page_info in pdf_info_dict.values():
         is_good_pdf, err = pdf_post_filter(page_info)
         if not is_good_pdf:
-            logger.warning(f"page_id: {i}, drop this pdf: {book_name}, reason: {err}")
+            logger.warning(f"page_id: {i}, drop this pdf: {pdf_bytes_md5}, reason: {err}")
             if not debug_mode:
                 return err
         i += 1
 
-    if debug_mode:
-        params_file_save_path = join_path(save_tmp_path, "md", book_name, "preproc_out.json")
-        page_draw_rect_save_path = join_path(save_tmp_path, "md", book_name, "layout.pdf")
-        # dir_path = os.path.dirname(page_draw_rect_save_path)
-        # if not os.path.exists(dir_path):
-        #     # 如果目录不存在，创建它
-        #     os.makedirs(dir_path)
-
-        with open(params_file_save_path, "w", encoding="utf-8") as f:
-            json.dump(pdf_info_dict, f, ensure_ascii=False, indent=4)
-        # 先检测本地 page_draw_rect_save_path 是否存在，如果存在则删除
-        if os.path.exists(page_draw_rect_save_path):
-            os.remove(page_draw_rect_save_path)
-        # 绘制bbox和layout到pdf
-        draw_bbox_on_page(pdf_docs, pdf_info_dict, page_draw_rect_save_path)
-        draw_layout_bbox_on_page(pdf_docs, pdf_info_dict, header, footer, page_draw_rect_save_path)
-
     if debug_mode:
         # 打印后处理阶段耗时
         logger.info(f"post_processing_time: {get_delta_time(start_time)}")
@@ -429,56 +374,28 @@ def parse_pdf_by_txt(
     para_process_pipeline = ParaProcessPipeline()
 
     def _deal_with_text_exception(error_info):
-        logger.warning(f"page_id: {page_id}, drop this pdf: {book_name}, reason: {error_info}")
+        logger.warning(f"page_id: {page_id}, drop this pdf: {pdf_bytes_md5}, reason: {error_info}")
         if error_info == denseSingleLineBlockException_msg:
-            logger.warning(f"Drop this pdf: {book_name}, reason: {DropReason.DENSE_SINGLE_LINE_BLOCK}")
-            result = {"need_drop": True, "drop_reason": DropReason.DENSE_SINGLE_LINE_BLOCK}
+            logger.warning(f"Drop this pdf: {pdf_bytes_md5}, reason: {DropReason.DENSE_SINGLE_LINE_BLOCK}")
+            result = {"_need_drop": True, "_drop_reason": DropReason.DENSE_SINGLE_LINE_BLOCK}
             return result
         if error_info == titleDetectionException_msg:
-            logger.warning(f"Drop this pdf: {book_name}, reason: {DropReason.TITLE_DETECTION_FAILED}")
-            result = {"need_drop": True, "drop_reason": DropReason.TITLE_DETECTION_FAILED}
+            logger.warning(f"Drop this pdf: {pdf_bytes_md5}, reason: {DropReason.TITLE_DETECTION_FAILED}")
+            result = {"_need_drop": True, "_drop_reason": DropReason.TITLE_DETECTION_FAILED}
             return result
         elif error_info == titleLevelException_msg:
-            logger.warning(f"Drop this pdf: {book_name}, reason: {DropReason.TITLE_LEVEL_FAILED}")
-            result = {"need_drop": True, "drop_reason": DropReason.TITLE_LEVEL_FAILED}
+            logger.warning(f"Drop this pdf: {pdf_bytes_md5}, reason: {DropReason.TITLE_LEVEL_FAILED}")
+            result = {"_need_drop": True, "_drop_reason": DropReason.TITLE_LEVEL_FAILED}
             return result
         elif error_info == paraSplitException_msg:
-            logger.warning(f"Drop this pdf: {book_name}, reason: {DropReason.PARA_SPLIT_FAILED}")
-            result = {"need_drop": True, "drop_reason": DropReason.PARA_SPLIT_FAILED}
+            logger.warning(f"Drop this pdf: {pdf_bytes_md5}, reason: {DropReason.PARA_SPLIT_FAILED}")
+            result = {"_need_drop": True, "_drop_reason": DropReason.PARA_SPLIT_FAILED}
             return result
         elif error_info == paraMergeException_msg:
-            logger.warning(f"Drop this pdf: {book_name}, reason: {DropReason.PARA_MERGE_FAILED}")
-            result = {"need_drop": True, "drop_reason": DropReason.PARA_MERGE_FAILED}
+            logger.warning(f"Drop this pdf: {pdf_bytes_md5}, reason: {DropReason.PARA_MERGE_FAILED}")
+            result = {"_need_drop": True, "_drop_reason": DropReason.PARA_MERGE_FAILED}
             return result
 
-    if debug_mode:
-        input_pdf_file = f"{pdf_local_path}.pdf"
-        output_dir = f"{save_path}/{book_name}"
-        output_pdf_file = f"{output_dir}/pdf_annos.pdf"
-
-        """
-        Call the para_process_pipeline function to process the pdf_info_dict.
-        
-        Parameters:
-        para_debug_mode: str or None
-            If para_debug_mode is None, the para_process_pipeline will not keep any intermediate results.
-            If para_debug_mode is "simple", the para_process_pipeline will only keep the annos on the pdf and the final results as a json file.
-            If para_debug_mode is "full", the para_process_pipeline will keep all the intermediate results generated during each step.
-        """
-        pdf_info_dict, error_info = para_process_pipeline.para_process_pipeline(
-            pdf_info_dict,
-            para_debug_mode="simple",
-            input_pdf_path=input_pdf_file,
-            output_pdf_path=output_pdf_file,
-        )
-        # 打印段落处理阶段耗时
-        logger.info(f"para_process_time: {get_delta_time(start_time)}")
-
-        # debug的时候不return drop信息
-        if error_info is not None:
-            _deal_with_text_exception(error_info)
-        return pdf_info_dict
-    else:
     pdf_info_dict, error_info = para_process_pipeline.para_process_pipeline(pdf_info_dict)
     if error_info is not None:
         return _deal_with_text_exception(error_info)

magic_pdf/pdf_parse_for_train.py

@@ -112,7 +112,6 @@ def parse_pdf_for_train(
     pdf_model_output,
     save_path,
     book_name,
-    pdf_model_profile=None,
     image_s3_config=None,
     start_page_id=0,
     end_page_id=None,
@@ -184,8 +183,8 @@ def parse_pdf_for_train(
                 f"page_id: {page_id}, img_counts: {img_counts}, drop this pdf: {book_name}, drop_reason: {DropReason.HIGH_COMPUTATIONAL_lOAD_BY_IMGS}"
             )
             result = {
-                "need_drop": True,
-                "drop_reason": DropReason.HIGH_COMPUTATIONAL_lOAD_BY_IMGS,
+                "_need_drop": True,
+                "_drop_reason": DropReason.HIGH_COMPUTATIONAL_lOAD_BY_IMGS,
             }
             if not debug_mode:
                 return result
@@ -200,7 +199,7 @@ def parse_pdf_for_train(
             flags=fitz.TEXTFLAGS_TEXT,
         )["blocks"]
         model_output_json = get_docx_model_output(
-            pdf_model_output, pdf_model_profile, page_id
+            pdf_model_output, page_id
         )
 
         # 解析图片
@@ -397,8 +396,8 @@ def parse_pdf_for_train(
                 f"page_id: {page_id}, drop this pdf: {book_name}, reason: {DropReason.TEXT_BLCOK_HOR_OVERLAP}"
             )
             result = {
-                "need_drop": True,
-                "drop_reason": DropReason.TEXT_BLCOK_HOR_OVERLAP,
+                "_need_drop": True,
+                "_drop_reason": DropReason.TEXT_BLCOK_HOR_OVERLAP,
             }
             if not debug_mode:
                 return result
@@ -444,8 +443,8 @@ def parse_pdf_for_train(
                 f"page_id: {page_id}, drop this pdf: {book_name}, reason: {DropReason.CAN_NOT_DETECT_PAGE_LAYOUT}"
             )
             result = {
-                "need_drop": True,
-                "drop_reason": DropReason.CAN_NOT_DETECT_PAGE_LAYOUT,
+                "_need_drop": True,
+                "_drop_reason": DropReason.CAN_NOT_DETECT_PAGE_LAYOUT,
             }
             if not debug_mode:
                 return result
@@ -457,7 +456,7 @@ def parse_pdf_for_train(
             logger.warning(
                 f"page_id: {page_id}, drop this pdf: {book_name}, reason: {DropReason.COMPLICATED_LAYOUT}"
             )
-            result = {"need_drop": True, "drop_reason": DropReason.COMPLICATED_LAYOUT}
+            result = {"_need_drop": True, "_drop_reason": DropReason.COMPLICATED_LAYOUT}
             if not debug_mode:
                 return result
 
@@ -467,8 +466,8 @@ def parse_pdf_for_train(
                 f"page_id: {page_id}, drop this pdf: {book_name}, reason: {DropReason.TOO_MANY_LAYOUT_COLUMNS}"
             )
             result = {
-                "need_drop": True,
-                "drop_reason": DropReason.TOO_MANY_LAYOUT_COLUMNS,
+                "_need_drop": True,
+                "_drop_reason": DropReason.TOO_MANY_LAYOUT_COLUMNS,
                 "extra_info": {"column_cnt": layout_column_width},
             }
             if not debug_mode:
@@ -617,8 +616,8 @@ def parse_pdf_for_train(
                 f"Drop this pdf: {book_name}, reason: {DropReason.DENSE_SINGLE_LINE_BLOCK}"
             )
             result = {
-                "need_drop": True,
-                "drop_reason": DropReason.DENSE_SINGLE_LINE_BLOCK,
+                "_need_drop": True,
+                "_drop_reason": DropReason.DENSE_SINGLE_LINE_BLOCK,
             }
             return result
         if error_info == titleDetectionException_msg:
@@ -626,27 +625,27 @@ def parse_pdf_for_train(
                 f"Drop this pdf: {book_name}, reason: {DropReason.TITLE_DETECTION_FAILED}"
             )
             result = {
-                "need_drop": True,
-                "drop_reason": DropReason.TITLE_DETECTION_FAILED,
+                "_need_drop": True,
+                "_drop_reason": DropReason.TITLE_DETECTION_FAILED,
             }
             return result
         elif error_info == titleLevelException_msg:
             logger.warning(
                 f"Drop this pdf: {book_name}, reason: {DropReason.TITLE_LEVEL_FAILED}"
             )
-            result = {"need_drop": True, "drop_reason": DropReason.TITLE_LEVEL_FAILED}
+            result = {"_need_drop": True, "_drop_reason": DropReason.TITLE_LEVEL_FAILED}
             return result
         elif error_info == paraSplitException_msg:
             logger.warning(
                 f"Drop this pdf: {book_name}, reason: {DropReason.PARA_SPLIT_FAILED}"
             )
-            result = {"need_drop": True, "drop_reason": DropReason.PARA_SPLIT_FAILED}
+            result = {"_need_drop": True, "_drop_reason": DropReason.PARA_SPLIT_FAILED}
             return result
         elif error_info == paraMergeException_msg:
             logger.warning(
                 f"Drop this pdf: {book_name}, reason: {DropReason.PARA_MERGE_FAILED}"
             )
-            result = {"need_drop": True, "drop_reason": DropReason.PARA_MERGE_FAILED}
+            result = {"_need_drop": True, "_drop_reason": DropReason.PARA_MERGE_FAILED}
             return result
 
     if debug_mode:

magic_pdf/pipeline.bak

@@ -32,8 +32,8 @@ def meta_scan(jso: dict, doc_layout_check=True) -> dict:
         if (
             "doc_layout_result" not in jso
         ):  # 检测json中是存在模型数据，如果没有则需要跳过该pdf
-            jso["need_drop"] = True
-            jso["drop_reason"] = DropReason.MISS_DOC_LAYOUT_RESULT
+            jso["_need_drop"] = True
+            jso["_drop_reason"] = DropReason.MISS_DOC_LAYOUT_RESULT
             return jso
     try:
         data_source = get_data_source(jso)
@@ -58,10 +58,10 @@ def meta_scan(jso: dict, doc_layout_check=True) -> dict:
         start_time = time.time()  # 记录开始时间
         res = pdf_meta_scan(s3_pdf_path, file_content)
         if res.get(
-            "need_drop", False
+            "_need_drop", False
         ):  # 如果返回的字典里有need_drop，则提取drop_reason并跳过本次解析
-            jso["need_drop"] = True
-            jso["drop_reason"] = res["drop_reason"]
+            jso["_need_drop"] = True
+            jso["_drop_reason"] = res["_drop_reason"]
         else:  # 正常返回
             jso["pdf_meta"] = res
             jso["content"] = ""
@@ -85,7 +85,7 @@ def classify_by_type(jso: dict, debug_mode=False) -> dict:
     if debug_mode:
         pass
     else:  # 如果debug没开，则检测是否有needdrop字段
-        if jso.get("need_drop", False):
+        if jso.get("_need_drop", False):
             return jso
     # 开始正式逻辑
     try:
@@ -113,8 +113,8 @@ def classify_by_type(jso: dict, debug_mode=False) -> dict:
         if (
             is_encrypted or is_needs_password
         ):  # 加密的，需要密码的，没有页面的，都不处理
-            jso["need_drop"] = True
-            jso["drop_reason"] = DropReason.ENCRYPTED
+            jso["_need_drop"] = True
+            jso["_drop_reason"] = DropReason.ENCRYPTED
         else:
             start_time = time.time()  # 记录开始时间
             is_text_pdf, results = classify(
@@ -139,8 +139,8 @@ def classify_by_type(jso: dict, debug_mode=False) -> dict:
                 if (
                     text_language not in allow_language
                 ):  # 如果语言不在允许的语言中，则drop
-                    jso["need_drop"] = True
-                    jso["drop_reason"] = DropReason.NOT_ALLOW_LANGUAGE
+                    jso["_need_drop"] = True
+                    jso["_drop_reason"] = DropReason.NOT_ALLOW_LANGUAGE
                     return jso
             else:
                 # 先不drop
@@ -148,8 +148,8 @@ def classify_by_type(jso: dict, debug_mode=False) -> dict:
                 jso["_pdf_type"] = "OCR"
                 jso["pdf_meta"] = pdf_meta
                 jso["classify_time"] = classify_time
-                # jso["need_drop"] = True
-                # jso["drop_reason"] = DropReason.NOT_IS_TEXT_PDF
+                # jso["_need_drop"] = True
+                # jso["_drop_reason"] = DropReason.NOT_IS_TEXT_PDF
                 extra_info = {"classify_rules": []}
                 for condition, result in results.items():
                     if not result:
@@ -162,7 +162,7 @@ def classify_by_type(jso: dict, debug_mode=False) -> dict:
 
 
 def drop_needdrop_pdf(jso: dict) -> dict:
-    if jso.get("need_drop", False):
+    if jso.get("_need_drop", False):
         logger.info(
             f"book_name is:{get_data_source(jso)}/{jso['file_id']} need drop",
             file=sys.stderr,
@@ -176,7 +176,7 @@ def pdf_intermediate_dict_to_markdown(jso: dict, debug_mode=False) -> dict:
     if debug_mode:
         pass
     else:  # 如果debug没开，则检测是否有needdrop字段
-        if jso.get("need_drop", False):
+        if jso.get("_need_drop", False):
             book_name = join_path(get_data_source(jso), jso["file_id"])
             logger.info(f"book_name is:{book_name} need drop", file=sys.stderr)
             jso["dropped"] = True
@@ -203,7 +203,7 @@ def parse_pdf(jso: dict, start_page_id=0, debug_mode=False) -> dict:
     if debug_mode:
         pass
     else:  # 如果debug没开，则检测是否有needdrop字段
-        if jso.get("need_drop", False):
+        if jso.get("_need_drop", False):
             return jso
     # 开始正式逻辑
     s3_pdf_path = jso.get("file_location")
@@ -220,8 +220,8 @@ def parse_pdf(jso: dict, start_page_id=0, debug_mode=False) -> dict:
     svgs_per_page_list = jso["pdf_meta"]["svgs_per_page"]
     max_svgs = max(svgs_per_page_list)
     if max_svgs > 3000:
-        jso["need_drop"] = True
-        jso["drop_reason"] = DropReason.HIGH_COMPUTATIONAL_lOAD_BY_SVGS
+        jso["_need_drop"] = True
+        jso["_drop_reason"] = DropReason.HIGH_COMPUTATIONAL_lOAD_BY_SVGS
     else:
         try:
             save_path = s3_image_save_path
@@ -244,10 +244,10 @@ def parse_pdf(jso: dict, start_page_id=0, debug_mode=False) -> dict:
                 debug_mode=debug_mode,
             )
             if pdf_info_dict.get(
-                "need_drop", False
+                "_need_drop", False
             ):  # 如果返回的字典里有need_drop，则提取drop_reason并跳过本次解析
-                jso["need_drop"] = True
-                jso["drop_reason"] = pdf_info_dict["drop_reason"]
+                jso["_need_drop"] = True
+                jso["_drop_reason"] = pdf_info_dict["_drop_reason"]
             else:  # 正常返回，将 pdf_info_dict 压缩并存储
                 pdf_info_dict = JsonCompressor.compress_json(pdf_info_dict)
                 jso["pdf_intermediate_dict"] = pdf_info_dict
@@ -269,7 +269,7 @@ def parse_pdf_for_model_train(jso: dict, start_page_id=0, debug_mode=False) -> d
     if debug_mode:
         pass
     else:  # 如果debug没开，则检测是否有needdrop字段
-        if jso.get("need_drop", False):
+        if jso.get("_need_drop", False):
             return jso
     # 开始正式逻辑
     s3_pdf_path = jso.get("file_location")
@@ -295,8 +295,8 @@ def parse_pdf_for_model_train(jso: dict, start_page_id=0, debug_mode=False) -> d
     svgs_per_page_list = jso["pdf_meta"]["svgs_per_page"]
     max_svgs = max(svgs_per_page_list)
     if max_svgs > 3000:
-        jso["need_drop"] = True
-        jso["drop_reason"] = DropReason.HIGH_COMPUTATIONAL_lOAD_BY_SVGS
+        jso["_need_drop"] = True
+        jso["_drop_reason"] = DropReason.HIGH_COMPUTATIONAL_lOAD_BY_SVGS
     # elif total_page > 1000:
     #     jso['need_drop'] = True
     #     jso['drop_reason'] = DropReason.HIGH_COMPUTATIONAL_lOAD_BY_TOTAL_PAGES
@@ -323,10 +323,10 @@ def parse_pdf_for_model_train(jso: dict, start_page_id=0, debug_mode=False) -> d
                 debug_mode=debug_mode,
             )
             if pdf_info_dict.get(
-                "need_drop", False
+                "_need_drop", False
             ):  # 如果返回的字典里有need_drop，则提取drop_reason并跳过本次解析
-                jso["need_drop"] = True
-                jso["drop_reason"] = pdf_info_dict["drop_reason"]
+                jso["_need_drop"] = True
+                jso["_drop_reason"] = pdf_info_dict["_drop_reason"]
             else:  # 正常返回，将 pdf_info_dict 压缩并存储
                 jso["parsed_results"] = convert_to_train_format(pdf_info_dict)
                 pdf_info_dict = JsonCompressor.compress_json(pdf_info_dict)