使用PyMuPDF提取PDF表格时函数调用致Jupyter Kernel崩溃

问题描述 #

单独运行PyMuPDF提取PDF表格的脚本时完全正常，能成功获取目标表格，但将逻辑封装成函数调用时，Jupyter内核直接崩溃，单元格无限运行，日志显示：

11:46:41.424 [error] Disposing session as kernel process died ExitCode: 3221225477, Reason:

已尝试打印调试、更新PyMuPDF，且PDF仅两页，排除内存问题。

正常运行的脚本 #

import pandas as pd
import pymupdf

doc = pymupdf.open(filePath)
pageNum = 1

page = doc[pageNum]

# "afuhfahuashsadhhadasdhukads" 是故意设置的不存在文本，让left_anchor为空列表
left_anchor = page.search_for("afuhfahuashsadhhadasdhukads") 
top_anchor = page.search_for("Characteristics")
right_anchor = page.search_for("Market Allocation")
bottom_anchor = page.search_for("Top 10 holdings")

x0 = left_anchor[0].x1 if len(left_anchor) > 0 else page.rect.x0
y0 = top_anchor[0].y1 if len(top_anchor) > 0 else page.rect.y0
x1 = right_anchor[0].x0 if len(right_anchor) > 0 else page.rect.x1
y1 = bottom_anchor[0].y0 if len(bottom_anchor) > 0 else page.rect.y1

clip = pymupdf.Rect(x0, y0, x1, y1)

find_table = page.find_tables(
    clip = clip, 
    vertical_strategy = "text", 
    horizontal_strategy = "lines"
)

table = find_table.tables[0].to_pandas()

table

导致崩溃的函数及调用 #

import pymupdf

def get_table(doc, page_num, left_anchor_text=None, top_anchor_text=None, right_anchor_text=None, bottom_anchor_text=None, vertical_strategy=None, horizontal_strategy=None):

    page = doc[page_num]

    left_anchor = page.search_for(left_anchor_text)
    top_anchor = page.search_for(top_anchor_text)
    right_anchor = page.search_for(right_anchor_text)
    bottom_anchor = page.search_for(bottom_anchor_text)

    x0 = left_anchor[0].x1 if len(left_anchor) > 0 else page.rect.x0
    y0 = top_anchor[0].y1 if len(top_anchor) > 0 else page.rect.y0
    x1 = right_anchor[0].x0 if len(right_anchor) > 0 else page.rect.x1
    y1 = bottom_anchor[0].y0 if len(bottom_anchor) > 0 else page.rect.y1

    clip = pymupdf.Rect(x0, y0, x1, y1)

    ft = page.find_tables(
        clip = clip, 
        vertical_strategy = vertical_strategy if vertical_strategy else "lines",
        horizontal_strategy = horizontal_strategy if horizontal_strategy else "lines", 
    )

    table = ft.tables[0].to_pandas()
    
    return table

document = pymupdf.open(filePath)

get_table(
    doc=document, 
    page_num=1, 
    top_anchor_text="Performance", 
    bottom_anchor_text="Calendar Year Performance",
    vertical_strategy="text", 
    horizontal_strategy="lines"
)

问题分析 #

ExitCode 3221225477对应Windows系统的访问违规错误（Access Violation），通常是代码试图访问未授权的内存地址，比如空指针引用。对比两段代码，核心差异在于：

• 函数版本中，left_anchor_text、right_anchor_text默认传None（调用时未传入这两个参数），而单独脚本中是传了一个不存在的字符串
• PyMuPDF的page.search_for方法传入None时，可能触发底层C扩展的空指针错误，直接导致进程崩溃

解决方案 #

修改函数，确保传给page.search_for的参数永远是字符串类型，未传入的锚点文本用空字符串替代None，同时增加边界检查避免索引越界：

import pymupdf
import pandas as pd

def get_table(doc, page_num, left_anchor_text="", top_anchor_text="", right_anchor_text="", bottom_anchor_text="", vertical_strategy="lines", horizontal_strategy="lines"):
    page = doc[page_num]

    # 处理空锚点文本，避免传入None触发底层错误
    left_anchor = page.search_for(left_anchor_text) if left_anchor_text else []
    top_anchor = page.search_for(top_anchor_text) if top_anchor_text else []
    right_anchor = page.search_for(right_anchor_text) if right_anchor_text else []
    bottom_anchor = page.search_for(bottom_anchor_text) if bottom_anchor_text else []

    x0 = left_anchor[0].x1 if len(left_anchor) > 0 else page.rect.x0
    y0 = top_anchor[0].y1 if len(top_anchor) > 0 else page.rect.y0
    x1 = right_anchor[0].x0 if len(right_anchor) > 0 else page.rect.x1
    y1 = bottom_anchor[0].y0 if len(bottom_anchor) > 0 else page.rect.y1

    clip = pymupdf.Rect(x0, y0, x1, y1)

    ft = page.find_tables(
        clip = clip, 
        vertical_strategy = vertical_strategy,
        horizontal_strategy = horizontal_strategy, 
    )

    # 检查是否找到表格，避免索引越界
    if ft.tables:
        return ft.tables[0].to_pandas()
    else:
        return pd.DataFrame()  # 未找到表格返回空DataFrame

document = pymupdf.open(filePath)

get_table(
    doc=document, 
    page_num=1, 
    top_anchor_text="Performance", 
    bottom_anchor_text="Calendar Year Performance",
    vertical_strategy="text", 
    horizontal_strategy="lines"
)

额外优化说明 #

• 函数参数默认值直接设为合理的默认策略，简化代码逻辑
• 明确处理空锚点文本的情况，直接返回空列表，避免调用search_for(None)
• 增加表格存在性检查，防止ft.tables[0]索引越界报错

内容的提问来源于stack exchange，提问作者OBramalam