Internet 上主流的 pdf 文档基本都是 A4 版面，在 kindle paperwhite 上直接阅读的体验一言难尽。如何尽可能简单地将A4幅面的文档转换为合适的格式?
2023 更新：由 kpw 更换到了 掌阅的 smart X 后，8寸屏幕的 pdf 阅读体验相对可以接受了。
首先明确的是，这里指的pdf是以可复制的文字格式为主要组成部分的文档，而非扫描文档。
面向kindle的文档管理，首选的是Calibre，它已经可以将pdf文件转换为epub文档，但是转换过程中会把一整段分为很多独立的行，发送到kindle后分段错乱更明显。
参考epub电子书–目录结构介绍，Epub的格式非常容易解析，它实质上是一个包含所需标记文本、图片、元信息等的一个zip压缩包。通过修改其中的标记文本，比较容易将被分开的多行合并。
对超文本的解析走了一点弯路，最开始是打算暴力匹配的，通过正则匹配标签，然后将连续的多段合并，这样处理的问题主要是，章节标题不容易处理。
接下来简单介绍一下思路，主要用到两个包，zipfile和BeautifulSoup，后者版本是4。
main函数分三部分，解压文件，处理超文本，打包回epub格式。
def main():
# extract epub file into appointed dir
extract_epub("data/src.epub", "data/src")
# format all the .html file.
file_list = [x for x in os.listdir("data/src") if ".html" in x]
for file in file_list:
print(f"Handle file: {file}")
res = handle_document("data/src/" + file)
with open("data/src/" + file, 'w') as f:
f.writelines(res)
# packaging target dir into epub file.
creat_epub("data/src", "data/dst.epub", "data/src/", log=False)
print("Create target File.")
解压文件部分的代码：
def extract_epub(file, target_dir):
# extract epub file into appointed dir
with zipfile.ZipFile(file, 'r') as myzip:
if not os.path.exists(target_dir):
os.makedirs(target_dir)
myzip.extractall(target_dir)
print(f"Excart file into [{target_dir}]")
处理超文本部分的代码：
def handle_document(file):
# load file, get root and body
(soup, lines) = load_document(file, "body")
# define ad string
ad_string = "更多电子书资料请搜索「书行天下」：http://www.sxpdf.com"
delete_ad(lines, ad_string)
# merge splited lines
handle_lines(lines)
for line in lines:
if not len(line.contents):
line.decompose()
return soup.prettify()
处理文本部分的代码完成三件事：
- 删除单行广告（这里的文件广告行是独立一行，所以比较容易处理）
- 将被分开的多行合并，准确的讲，是将这几行的所有tag按顺序合并到一个
<p>标签中。 - 删除多余的空白
<p>标签。
其中第二部分占据了主要的工作量，代码如下：
这里的lines就是所有的<p>标签，遍历后合并，被合并的标签中置空，留待后续处理。
def handle_lines(lines):
# re-format lines, merget splited lines into one line.
index = 0
toc = [x[0] for x in get_toc()]
paragraph = [0, []]
# clear cache lines
def clear_cache():
if paragraph[0]:
lines[paragraph[0]].string = ""
for tag in paragraph[1]:
lines[paragraph[0]].append(tag)
paragraph[0] = 0
paragraph[1] = []
while (index < len(lines)):
line = ''.join(list(lines[index].stripped_strings))
# merge splited title
(is_title, completed, title) = is_toc_prefix(toc, line)
if is_title and not completed:
t = [x for x in lines[index].find_all(name="a") if x.string]
if t:
t[0].string = title
lines[index + 1].decompose()
index += 2
clear_cache()
continue
elif is_title and completed:
index += 1
clear_cache()
continue
# single line contain img, omit:
if lines[index].img:
index += 1
clear_cache()
continue
# merge splited normal lines
if (25 < len(line)):
if not paragraph[0]:
paragraph[0] = index
# whether single space between every char?
if line.count(' ')/len(line) > 0.4 and " " not in line:
if not lines[index].string:
# may have tag <a> in lines[index]
# handle this situation
t = [x for x in lines[index].contents if x.string]
t[0].string = line.replace(' ', '')
else:
lines[index].string = line.replace(' ', '')
paragraph[1].extend(lines[index].contents)
if not paragraph[0] and index != paragraph[0]:
lines[index].decompose()
else:
paragraph[1].extend(lines[index].contents)
clear_cache()
# Important!
index += 1
# handle last paragraph
if paragraph[0]:
clear_cache()
综上，整体的代码为：
#!/usr/bin/python3
# -*- coding: utf-8 -*-
"""
Author: Shadow Cheng
"""
import os
import zipfile
from bs4 import BeautifulSoup
def load_document(file, label_name):
# load html or xml file, return DOM root and all node named [label_name]
with open(file, 'r') as f:
soup = BeautifulSoup(f, "html.parser")
body = soup.findChildren(name=label_name)[0]
lines = [x for x in body.children if x != '\n']
return (soup, lines)
def get_toc():
# Fixed, if re-format other epub file,
# check whether the toc file named "toc.ncx"
#
with open("data/src/toc.ncx", 'r') as f:
soup = BeautifulSoup(f, "html.parser")
ncx = soup.findChildren(name='navpoint')
toc = []
for item in ncx:
text = ""
src = ""
for s in item.children:
if s.name == "navlabel":
text = s.text.strip()
elif s.name == "content":
src = s.attrs['src']
toc.append([text, src])
return toc
def is_toc_prefix(toc, line):
# whether line is title of content, return (flag-1, flag-2)
# flag-1 indicates whether line in title
# flag-2 indicates whether line is completed title.
if len(line) == 0:
return (False, False, "")
for title in toc:
if (line.replace(' ', '') in title.replace(' ', '')):
if len(line) == len(title):
return (True, True, title)
else:
return (True, False, title)
return (False, False, title)
def delete_ad(lines, ad_string):
for line in lines:
if line.string == ad_string:
line.decompose()
def handle_lines(lines):
# re-format lines, merget splited lines into one line.
index = 0
toc = [x[0] for x in get_toc()]
paragraph = [0, []]
# clear cache lines
def clear_cache():
if paragraph[0]:
lines[paragraph[0]].string = ""
for tag in paragraph[1]:
lines[paragraph[0]].append(tag)
paragraph[0] = 0
paragraph[1] = []
while (index < len(lines)):
line = ''.join(list(lines[index].stripped_strings))
# merge splited title
(is_title, completed, title) = is_toc_prefix(toc, line)
if is_title and not completed:
t = [x for x in lines[index].find_all(name="a") if x.string]
if t:
t[0].string = title
lines[index + 1].decompose()
index += 2
clear_cache()
continue
elif is_title and completed:
index += 1
clear_cache()
continue
# single line contain img, omit:
if lines[index].img:
index += 1
clear_cache()
continue
# merge splited normal lines
if (25 < len(line)):
if not paragraph[0]:
paragraph[0] = index
# whether single space between every char?
if line.count(' ')/len(line) > 0.4 and " " not in line:
if not lines[index].string:
# may have tag <a> in lines[index]
# handle this situation
t = [x for x in lines[index].contents if x.string]
t[0].string = line.replace(' ', '')
else:
lines[index].string = line.replace(' ', '')
paragraph[1].extend(lines[index].contents)
if not paragraph[0] and index != paragraph[0]:
lines[index].decompose()
else:
paragraph[1].extend(lines[index].contents)
clear_cache()
# Important!
index += 1
# handle last paragraph
if paragraph[0]:
clear_cache()
def handle_document(file):
# load file, get root and body
(soup, lines) = load_document(file, "body")
# define ad string
ad_string = "更多电子书资料请搜索「书行天下」：http://www.sxpdf.com"
delete_ad(lines, ad_string)
# merge splited lines
handle_lines(lines)
for line in lines:
if not len(line.contents):
line.decompose()
return soup.prettify()
def extract_epub(file, target_dir):
# extract epub file into appointed dir
with zipfile.ZipFile(file, 'r') as myzip:
if not os.path.exists(target_dir):
os.makedirs(target_dir)
myzip.extractall(target_dir)
print(f"Excart file into [{target_dir}]")
def creat_epub(target_dir, dst_name, wipe_prefix, log=False):
# create epub file
# walk around target dir, and push every file into epub.
with zipfile.ZipFile(dst_name, 'w') as target:
for i in os.walk(target_dir, topdown=False):
for n in i[2]:
file_path = os.path.normpath(''.join((i[0], "/", n)))
file_path_in_zip = file_path.replace(wipe_prefix, '')
target.write(file_path, file_path_in_zip)
if log:
print(f"Write file:{file_path}->{file_path_in_zip}")
def main():
# extract epub file into appointed dir
extract_epub("data/src.epub", "data/src")
# format all the .html file.
file_list = [x for x in os.listdir("data/src") if ".html" in x]
for file in file_list:
print(f"Handle file: {file}")
res = handle_document("data/src/" + file)
with open("data/src/" + file, 'w') as f:
f.writelines(res)
# packaging target dir into epub file.
creat_epub("data/src", "data/dst.epub", "data/src/", log=False)
print("Create target File.")
if __name__ == "__main__":
main()
总结：
1. BS4 库对于 html 文档的处理挺有意思，值得一试；
2. Os库自带的`walk`、`makedirs`函数很方便，自己想要实现也不难，留待后续有时间的时候练手吧。