实现文本拼PDF解析结果装标准格式

magic_pdf/pipeline.py

@@ -23,17 +23,11 @@
 
 from magic_pdf.pdf_parse_by_ocr import parse_pdf_by_ocr
 from magic_pdf.pdf_parse_for_train import parse_pdf_for_train
+from magic_pdf.spark.base import exception_handler, get_data_source
 from magic_pdf.train_utils.convert_to_train_format import convert_to_train_format
 from app.common.s3 import get_s3_config, get_s3_client
 
 
-def exception_handler(jso: dict, e):
-    logger.exception(e)
-    jso["need_drop"] = True
-    jso["drop_reason"] = DropReason.Exception
-    jso["exception"] = f"ERROR: {e}"
-    return jso
-
 
 def get_data_type(jso: dict):
     data_type = jso.get("data_type")
@@ -49,13 +43,6 @@ def get_bookid(jso: dict):
     return book_id
 
 
-def get_data_source(jso: dict):
-    data_source = jso.get("data_source")
-    if data_source is None:
-        data_source = jso.get("file_source")
-    return data_source
-
-
 def meta_scan(jso: dict, doc_layout_check=True) -> dict:
     s3_pdf_path = jso.get("file_location")
     s3_config = get_s3_config(s3_pdf_path)

magic_pdf/pipeline_txt.py

@@ -0,0 +1,37 @@
+"""
+文本型pdf转化为统一清洗格式
+"""
+
+
+
+from loguru import logger
+from magic_pdf.dict2md.mkcontent import mk_universal_format
+from magic_pdf.libs.commons import join_path
+from magic_pdf.libs.json_compressor import JsonCompressor
+from magic_pdf.spark.base import exception_handler, get_data_source
+
+
+def txt_pdf_to_standard_format(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):
+            book_name = join_path(get_data_source(jso), jso["file_id"])
+            logger.info(f"book_name is:{book_name} need drop")
+            jso["dropped"] = True
+            return jso
+    try:
+        pdf_intermediate_dict = jso["pdf_intermediate_dict"]
+        # 将 pdf_intermediate_dict 解压
+        pdf_intermediate_dict = JsonCompressor.decompress_json(pdf_intermediate_dict)
+        standard_format = mk_universal_format(pdf_intermediate_dict)
+        jso["content_list"] = standard_format
+        logger.info(f"book_name is:{get_data_source(jso)}/{jso['file_id']},content_list length is {len(standard_format)}",)
+        # 把无用的信息清空
+        jso["doc_layout_result"] = ""
+        jso["pdf_intermediate_dict"] = ""
+        jso["pdf_meta"] = ""
+    except Exception as e:
+        jso = exception_handler(jso, e)
+    return jso

magic_pdf/spark/base.py

@@ -0,0 +1,21 @@
+
+
+from loguru import logger
+
+from magic_pdf.libs.drop_reason import DropReason
+
+
+def get_data_source(jso: dict):
+    data_source = jso.get("data_source")
+    if data_source is None:
+        data_source = jso.get("file_source")
+    return data_source
+
+
+def exception_handler(jso: dict, e):
+    logger.exception(e)
+    jso["need_drop"] = True
+    jso["drop_reason"] = DropReason.Exception
+    jso["exception"] = f"ERROR: {e}"
+    return jso
+