Python中textract提取.doc文件页眉页脚失效，求替代方案

Got it, the core issue here is that textract’s handling of legacy .doc files (the old binary Word format) is way more limited than its .docx support. Unlike .docx which is XML-based and easy to parse, .doc uses a complex binary structure, and the default backend textract uses for .doc (usually antiword) doesn’t extract headers or footers out of the box. That’s why you’re seeing the discrepancy between the two file types.

Here are two reliable workarounds to fix this:

Option 1: Use pywin32 (Windows-only) to leverage Word’s COM API #

This method is the most reliable because it directly uses Microsoft Word’s own API to extract content—so it’ll capture headers, footers, and all other elements just like opening the file in Word.

import win32com.client as win32

def extract_doc_content(filename):
    # Initialize Word application
    word = win32.gencache.EnsureDispatch('Word.Application')
    word.Visible = False  # Run in background to avoid popping up Word windows
    
    try:
        doc = word.Documents.Open(filename)
        
        # Extract main body text
        main_text = doc.Content.Text
        
        # Extract headers and footers (loop through all sections)
        headers = []
        footers = []
        for section in doc.Sections:
            # Get primary header/footer for the section
            headers.append(section.Headers(win32.constants.wdHeaderFooterPrimary).Range.Text)
            footers.append(section.Footers(win32.constants.wdHeaderFooterPrimary).Range.Text)
        
        # Combine all content into one string
        full_content = (
            "--- HEADER ---\n" + "\n".join(headers) +
            "\n--- MAIN TEXT ---\n" + main_text +
            "\n--- FOOTER ---\n" + "\n".join(footers)
        )
        # Encode to ASCII (matching your original code's behavior)
        return full_content.encode('ascii', errors='replace')
    
    finally:
        # Clean up: close document and quit Word
        doc.Close()
        word.Quit()

Pros: Perfectly extracts all Word content including headers/footers; preserves more formatting context if needed.
Cons: Only works on Windows (requires Microsoft Word installed).

Option 2: Convert .doc to .docx first (cross-platform) #

If you need a solution that works across Windows, macOS, or Linux, convert the .doc file to .docx first using LibreOffice/OpenOffice, then reuse your existing textract code for .docx. This works because .docx’s XML structure is easy for textract to parse, including headers and footers.

import subprocess
import os
import textract

def convert_doc_to_docx(doc_path):
    # Generate output .docx path
    docx_path = os.path.splitext(doc_path)[0] + ".docx"
    # Call LibreOffice in headless mode to convert
    subprocess.run(
        ["libreoffice", "--headless", "--convert-to", "docx", doc_path],
        check=True,
        stdout=subprocess.DEVNULL,
        stderr=subprocess.DEVNULL
    )
    return docx_path

def extract_doc_with_conversion(filename):
    if filename.lower().endswith('.doc'):
        # Convert to .docx temporarily
        docx_path = convert_doc_to_docx(filename)
        # Use your existing textract logic
        content = textract.process(docx_path, encoding='ascii')
        # Optional: Delete the temporary .docx file after extraction
        os.remove(docx_path)
        return content
    elif filename.lower().endswith('.docx'):
        # Use your original code directly
        try:
            return textract.process(filename, encoding='ascii')
        except Exception as e:
            raise RuntimeError(f"Couldn't open the file: {filename}") from e
    else:
        raise RuntimeError("Unsupported file format - only .doc and .docx are allowed")

Pros: Cross-platform; reuses your existing .docx handling code.
Cons: Requires installing LibreOffice/OpenOffice; adds a conversion step (minor performance hit).

Quick Note on Textract and .doc #

Just to clarify: Textract relies on external tools like antiword for .doc files, and antiword doesn’t support header/footer extraction by design. So tweaking your existing textract code won’t solve this—you need to switch to one of the methods above.

内容的提问来源于stack exchange，提问作者Kiran Kumar Kotari