pdftotext: Cannot install on Windows

All hope is not lost on the windows version. There is a command line utility with the same name and you can use the subprocess package to execute pdftotext

PDFtotext windows download instruction, credit @s2t2

1. Go to https://www.xpdfreader.com/download.html and click “Download the Xpdf tools”
2. Uncompress/extract the zip file, and move the folder to a location like the Desktop or the Programs directory.
3. Inside the unzipped folder, copy the file bin64/pdftotext.exe into your project repository

I “ugly” installed pdftotext successfully on windows three times over the past two days, as the subprocess method is a non-starter for me. I have a writeup on SO

https://stackoverflow.com/questions/45912641/unable-to-install-pdftotext-on-python-3-6-missing-poppler/58139729#58139729

as well as on my blog, which has screenshots

https://coder.haus/2019/09/27/installing-pdftotext-through-pip-on-windows-10/

Please try this, let me know if it works. I’m hoping to take time to do it properly and potentially generate a PR.

My solution requires Anaconda (for conda install). First, install Microsoft VC++ build tools, download poppler for windows as well as conda install poppler, and copy some of the poppler files to different locations in the Anaconda directory structure. Again, I have done this a total of 3 times and know it can be done better, but this will get you up and running.

Any chance of prebuilt binaries being offered? Is it something that could be integrated into the CI setup? I think we’d need to build windows binaries on windows though, so moving to Appveyor would be required, unfortunately. Maybe a solution like cibuildwheel can help with that. I tried to, and failed miserably, at building it on windows.

Considering the only alternative at the moment (pdfminer and it’s deraritives), which is super slow (4 orders of magnitude in my experience), inaccurate results which are in some cases impossible to parse accurately, I think It’d be great to offer prebuilt binaries with this functionality.

Now I have the latest wheel file. version 39 64 Bit pdftotext.zip

@palakjadwani this is the problem with there not being a binary download for the pdtotext package - you can find some workarounds at http://faculty.washington.edu/jwilker/559/2018/pdftotext.pdf but they are less than ideal.

No, I didn’t. My solution ended up being in a completely different direction, using different packages. This may be worth a mention in the README file for future Windows (10) users.

You would need to get poppler and its development files installed on Windows. I don’t use Windows, so I am not much help, sorry.

If you figure something out, I will gladly add it to the README here!

Using pyinstaller on Windows the Poppler DLLs are packed in the executable.

Using the poppler v21.10 the lcms2 DLL is needed (lcms color engine) Link to Poppler: https://anaconda.org/conda-forge/poppler/21.10.0/download/win-64/poppler-21.10.0-h24fffdf_0.tar.bz2 Updated Wheel: pdftotext-2.2.1-cp39-cp39-win_amd64.whl.zip

Updated DLL package: Conda_Forge_DLL_x64.zip poppler.dll v21.10.0 poppler-glib.dll v21.10.0 poppler-cpp.dll v21.10.0 freetype.dll v2.10 zlib.dll v1.2.11 libssh2.dll v1.10.0 cairo.dll v1.16.0 libtiff.dll tiff.dll v4.3.0 libzstd.dll en zstd.dll v1.5.0 libcurl v7.79.1.0 openjp2.dll v2.4.0 iconv.dll en charset.dll v1.16 libpng16.dll v1.6.37 liblzma.dll v5.2.2 libcrypto-1_1-x64.dll v1.1.1l lcms2.dll v2.12

I’ve had some trouble getting pdftotext working on Windows. But i managed with the following steps:

download poppler: https://anaconda.org/conda-forge/poppler/21.03.0/download/win-64/poppler-21.03.0-h9ff6ed8_0.tar.bz2 copy the contents from …\poppler-21.03.0-h9ff6ed8_0\Library\lib\ to …<Python-install-folder>\libs
copy the contents from …\poppler-21.03.0-h9ff6ed8_0\Library\include\poppler to …<Python-install-folder>\include\poppler copy the DLLs from …\poppler-21.03.0-h9ff6ed8_0\Library\include\bin*.dll to …<Python-install-folder>\Lib\site-packages\

Copy the DLLs to …<Python-install-folder>\Lib\site-packages charset.dll freetype.dll iconv.dll libcrypto-1_1-x64.dll libcurl.dll liblzma.dll libpng16.dll libssh2.dll openjp2.dll tiff.dll zlib.dll zstd.dll

Now you can install pdftotext with: pip install pdftotext-2.1.6-cp39-cp39-win_amd64.whl

Files are in attachment Conda_Forge_DLL_x64.zip

P.S. It would be great if the Poppler PDF rendering library based would be upgraded from the xpdf-3.0 to the xpdf-4.03 code base.

I have the wheel file of the pdftotext for cp38 version 3.8.5 64 bit. Just go to power shell and do cd [Location of the file] pip install ./[Wheel file [name] Or py -3.8 -m pip install ./[Wheel file name]

I am also attaching poppler files so you can extract these files in python destination folder. pdftotext.zip

If any one have the wheel of pdftotext of version cp39 64 bit kindly share it.

I have the wheel file of the pdftotext for cp38 version 3.8.5 64 bit. Just go to power shell and do cd [Location of the file] pip install ./[Wheel file [name] Or py -3.8 -m pip install ./[Wheel file name]

I am also attaching poppler files so you can extract these files in python destination folder. pdftotext.zip

The following fixes the issue on Windows 10.

Assumes MS VC++ Build Tools is installed. Assumes Anaconda is being used. Assumes Poppler is installed using conda install poppler.

The code update is in setup.py -

from os import path, getenv

if platform.system() in ['Windows']:
    conda_dir = getenv('CONDA_PREFIX')
    anaconda_poppler_include_dir = path.join(conda_dir, 'Library\include')
    anaconda_poppler_library_dir = path.join(conda_dir, 'Library\lib')
    include_dirs = [anaconda_poppler_include_dir]
    library_dirs = [anaconda_poppler_library_dir]

pip install completes successfully and the unit tests run successfully.

Let me know if this looks sane, and if I should create a PR for this.

Thanks. I am researching this but have not found any good guidance. I will report back my findings.