Python使用PDFTables解析PDF表格时忽略列间空格问题

Hey there! Let's work through that frustrating problem where pdftables is merging columns and ignoring inter-column spaces when parsing your PDF tables. Here are a few practical solutions to try out:

1. Tweak pdftables Parsing Parameters (If Supported) #

First, check if pdftables offers parameters to adjust how it detects columns. Some PDF parsing libraries let you set a threshold for inter-column spacing—if the space between columns is too small for the default setting, the library might treat them as a single column.

If pdftables supports this, try adjusting a space_threshold or similar parameter (check the library's docs for exact names):

pdf_page = get_pdf_page(fileobj, page)
# Adjust the space threshold to a higher value to recognize column gaps
tables = page_to_tables(pdf_page, space_threshold=6)  # Test different values (e.g., 4-10) based on your PDF

2. Preprocess the PDF Page with Coordinate-Based Text Extraction #

If pdftables' automatic column detection isn't cutting it, you can manually parse the table using text block coordinates (using a library like PyMuPDF/fitz). This lets you define columns based on the actual position of text in the PDF:

import fitz  # Install with pip install pymupdf

def parse_table_by_coords(pdf_path, page_num):
    doc = fitz.open(pdf_path)
    page = doc[page_num]
    # Extract text blocks with their coordinates (x0, y0, x1, y1, text, ...)
    text_blocks = page.get_text("blocks")
    
    # Group blocks into columns using their x-coordinate (adjust the 100 value to match your PDF's column width)
    column_map = {}
    for block in text_blocks:
        x_start = block[0]
        # Assign blocks to columns based on x-position (tweak the division value as needed)
        col_index = int(x_start // 100)
        if col_index not in column_map:
            column_map[col_index] = []
        column_map[col_index].append(block[4].strip())
    
    # Convert column data into a table structure
    max_row_count = max(len(col) for col in column_map.values())
    table = []
    for row_idx in range(max_row_count):
        current_row = []
        # Iterate through columns in order
        for col_idx in sorted(column_map.keys()):
            # Add empty string if the column has no data for this row
            cell_value = column_map[col_idx][row_idx] if row_idx < len(column_map[col_idx]) else ""
            current_row.append(cell_value)
        table.append(current_row)
    
    return table

# Usage example
your_table = parse_table_by_coords("your_file.pdf", page)

This method bypasses pdftables' automatic detection and lets you take control of how columns are defined—great for PDFs with inconsistent spacing or no clear table borders.

3. Switch to a More Flexible PDF Table Parser #

If pdftables continues to struggle, consider using libraries like camelot-py or tabula-py, which offer more granular control over table parsing:

Using camelot-py (Good for both bordered and borderless tables) #

import camelot

# For borderless tables (uses text spacing to detect columns)
tables = camelot.read_pdf("your_file.pdf", pages=str(page), flavor="stream", edge_tol=500)
# If you know the exact column positions, define them explicitly
# tables = camelot.read_pdf("your_file.pdf", pages=str(page), columns=["80, 180, 280, 380"])

# Access the parsed table
print(tables[0].df)

Using tabula-py #

from tabula import read_pdf

# Parse with auto-detection, or specify column areas
tables = read_pdf("your_file.pdf", pages=page, guess=True)
# For more control, define the area of the table and column positions
# tables = read_pdf("your_file.pdf", pages=page, area=[26, 14, 560, 720], columns=[14, 100, 200, 300])

print(tables[0])

These libraries often handle tricky cases like merged columns or subtle spacing better than pdftables, especially with the right parameters tuned to your specific PDF.

内容的提问来源于stack exchange，提问作者Khushhal