下面是针对我的特定用例的最常见答案的时间比较:合并5个大单页pdf文件的列表。每个测试我都运行了两次。

(免责声明:我在Flask中运行这个函数，您的里程可能会有所不同)

博士TL;

pdfrw是我测试的3个pdf文件组合库中最快的一个。

PyPDF2

start = time.time()
merger = PdfFileMerger()
for pdf in all_pdf_obj:
    merger.append(
        os.path.join(
            os.getcwd(), pdf.filename # full path
                )
            )
formatted_name = f'Summary_Invoice_{date.today()}.pdf'
merge_file = os.path.join(os.getcwd(), formatted_name)
merger.write(merge_file)
merger.close()
end = time.time()
print(end - start) #1 66.50084733963013 #2 68.2995400428772

PyMuPDF

start = time.time()
result = fitz.open()

for pdf in all_pdf_obj:
    with fitz.open(os.path.join(os.getcwd(), pdf.filename)) as mfile:
        result.insertPDF(mfile)
formatted_name = f'Summary_Invoice_{date.today()}.pdf'

result.save(formatted_name)
end = time.time()
print(end - start) #1 2.7166640758514404 #2 1.694727897644043

PDFrw

start = time.time()
result = fitz.open()

writer = PdfWriter()
for pdf in all_pdf_obj:
    writer.addpages(PdfReader(os.path.join(os.getcwd(), pdf.filename)).pages)

formatted_name = f'Summary_Invoice_{date.today()}.pdf'
writer.write(formatted_name)
end = time.time()
print(end - start) #1 0.6040127277374268 #2 0.9576816558837891

我在linux终端上通过利用subprocess(假设目录中存在one.pdf和two.pdf)使用pdf unite，目的是将它们合并为three.pdf

 import subprocess
 subprocess.call(['pdfunite one.pdf two.pdf three.pdf'],shell=True)

pdfrw库可以很容易地做到这一点，假设您不需要保存书签和注释，并且您的pdf文件没有加密。Cat.py是一个示例拼接脚本，而子集.py是一个示例页面子集脚本。

串联脚本的相关部分——假设input是一个输入文件名列表，outfn是一个输出文件名:

from pdfrw import PdfReader, PdfWriter

writer = PdfWriter()
for inpfn in inputs:
    writer.addpages(PdfReader(inpfn).pages)
writer.write(outfn)

正如你所看到的，省略最后一页是很容易的，例如:

    writer.addpages(PdfReader(inpfn).pages[:-1])

免责声明:我是pdfrw的主要作者。

使用字典以获得更大的灵活性(例如sort, dedup):

import os
from PyPDF2 import PdfFileMerger
# use dict to sort by filepath or filename
file_dict = {}
for subdir, dirs, files in os.walk("<dir>"):
    for file in files:
        filepath = subdir + os.sep + file
        # you can have multiple endswith
        if filepath.endswith((".pdf", ".PDF")):
            file_dict[file] = filepath
# use strict = False to ignore PdfReadError: Illegal character error
merger = PdfFileMerger(strict=False)

for k, v in file_dict.items():
    print(k, v)
    merger.append(v)

merger.write("combined_result.pdf")

使用正确的python解释器:

conda activate py_envs

pip install PyPDF2

Python代码:

from PyPDF2 import PdfMerger

#set path files
import os
os.chdir('/ur/path/to/folder/')
cwd = os.path.abspath('')
files = os.listdir(cwd)

def merge_pdf_files():
    merger = PdfMerger()
    pdf_files = [x for x in files if x.endswith(".pdf")]
    [merger.append(pdf) for pdf in pdf_files]
    with open("merged_pdf_all.pdf", "wb") as new_file:
        merger.write(new_file)

if __name__ == "__main__":
    merge_pdf_files()

下面是针对我的特定用例的最常见答案的时间比较:合并5个大单页pdf文件的列表。每个测试我都运行了两次。

(免责声明:我在Flask中运行这个函数，您的里程可能会有所不同)

博士TL;

pdfrw是我测试的3个pdf文件组合库中最快的一个。

PyPDF2

start = time.time()
merger = PdfFileMerger()
for pdf in all_pdf_obj:
    merger.append(
        os.path.join(
            os.getcwd(), pdf.filename # full path
                )
            )
formatted_name = f'Summary_Invoice_{date.today()}.pdf'
merge_file = os.path.join(os.getcwd(), formatted_name)
merger.write(merge_file)
merger.close()
end = time.time()
print(end - start) #1 66.50084733963013 #2 68.2995400428772

PyMuPDF

start = time.time()
result = fitz.open()

for pdf in all_pdf_obj:
    with fitz.open(os.path.join(os.getcwd(), pdf.filename)) as mfile:
        result.insertPDF(mfile)
formatted_name = f'Summary_Invoice_{date.today()}.pdf'

result.save(formatted_name)
end = time.time()
print(end - start) #1 2.7166640758514404 #2 1.694727897644043

PDFrw

start = time.time()
result = fitz.open()

writer = PdfWriter()
for pdf in all_pdf_obj:
    writer.addpages(PdfReader(os.path.join(os.getcwd(), pdf.filename)).pages)

formatted_name = f'Summary_Invoice_{date.today()}.pdf'
writer.write(formatted_name)
end = time.time()
print(end - start) #1 0.6040127277374268 #2 0.9576816558837891