Improving Windows with PyInstaller - Ocrmypdf Distribution Not Found

[gabemorris12]

Describe the bug
Pyinstaller was used to successfully create an executable file, but upon running the .exe, it crashes.

To Reproduce
To reproduce the error, make sure you flag --consoled when running the pyinstaller command. Then run the executable from the command prompt. Then it will raise pkg_resources.DistributionNotFound error.

Expected behavior
I expected all dependencies to be present.

Screenshots

System (please complete the following information):

• OS: windows 10
• Python version: 3.8.3
• OCRmyPDF version: 10.3.3

Additional context
The .py script runs perfectly fine. There's just a problem with pyinstaller not finding the distribution.

[jbarlow83]

I think you'll have to ask for support from the PyInstaller people, since I don't know much about it.

ocrmypdf's use of cffi to access a C library, and its use of third party executables like Tesseract and Ghostscript, may be problematic for packaging.

If you can identify a specific issue with ocrmypdf that is an obstacle to generating this installer, please open a PR.

[gabemorris12]

@jbarlow83 I it to run! I should have gone to pyinstaller first, so sorry about that. The fix can be found here: pyinstaller/pyinstaller#4809

I had to create hook files for both ocrmypdf module and the pikepdf module by using the --additional-hooks-dir HooksFolderPath flag. Thanks for the quick response.

[jbarlow83]

Would you mind sharing your full instructions? Maybe I'll be able to automate it one day.

[gabemorris12]

@jbarlow83
You have to make two different .py scripts in a folder within your project. For my example, I made a folder called Hooks, and these are the two scripts that I made:
hook-ocrmypdf.py

from PyInstaller.utils.hooks import collect_all

datas, binaries, hiddenimports = collect_all('ocrmypdf')

Having this hook alone will allow the app to run; however, it is not enough. In your program, once it attempts to execute the ocrmypdf.ocr() function, it will raise another error. I cannot remember what error it was exactly, but it was an error relating to a missing dependency within pikepdf. So the other script that you have to make is this:

hook-pikepdf.py

from PyInstaller.utils.hooks import collect_all

datas, binaries, hiddenimports = collect_all('pikepdf')

Now when you package the main script with pyinstaller, you must include this flag: --additional-hooks-dir HooksFolderPath.

[woaidianqian]

pyinstaller pdf识别.py --additional-hooks-dir HooksFolderPath

it works！

[koolkunz]

Getting this error after following the instructions above:

Traceback (most recent call last):
File "test.py", line 30, in
File "ocrmypdf\api.py", line 324, in ocr
plugin_manager = get_plugin_manager(plugins)
File "ocrmypdf_plugin_manager.py", line 104, in get_plugin_manager
return OcrmypdfPluginManager(
File "ocrmypdf_plugin_manager.py", line 45, in init
self.setup_plugins()
File "ocrmypdf_plugin_manager.py", line 69, in setup_plugins
for module in sorted(
TypeError: '<' not supported between instances of 'FrozenImporter' and 'FileFinder'

can anyone help?

[jbarlow83]

@koolkunz Look up how to use pyinstaller with a project that uses pluggy as a plugin manager. pytest is an example of such a project. You may also need add ocrmypdf/builtin_plugins to the list of dynamic plugins that will be used - ocrmypdf installs some of its own plugins by default so that others can override them.

[koolkunz]

@koolkunz Look up how to use pyinstaller with a project that uses pluggy as a plugin manager. pytest is an example of such a project. You may also need add ocrmypdf/builtin_plugins to the list of dynamic plugins that will be used - ocrmypdf installs some of its own plugins by default so that others can override them.

pyinstaller --copy-metadata pikepdf --copy-metadata ocrmypdf --collect-submodules ocrmypdf --collect-datas ocrmypdf.data sample.py

This is all I had to do to get it working^.

But now I have another small issue, that is if I am running my GUI application in no console mode, ocrmypdf keeps opening multiple console windows of tesseract (I think equal to the number of jobs being processed) and ghostscript executables with no output.

I already have logging set to -1:
configure_logging(verbosity=-1, progress_bar_friendly=False)

Is there any other way to suppress these console windows from showing up at all?

edit: Added Screenshot

[insinfo]

@koolkunz
could you make this available on google drive so i can download and use it on my windows notebook? I was in need of making some PDF Searchable. In my notebook I already have the latest version of tesseract v5.3.0, ghostscript 10.01.1 and python 3.10.8

[koolkunz]

@koolkunz could you make this available on google drive so i can download and use it on my windows notebook? I was in need of making some PDF Searchable. In my notebook I already have the latest version of tesseract v5.3.0, ghostscript 10.01.1 and python 3.10.8

I'm still working on it and unless you want to distribute the exe you can just use ocrmypdf through simple python code or even just directly in the command line.

[zeekias]

I have the same problem, someone could help me?

[zeekias]

How can i disable the tesseract prompt?

[lugi777]

@koolkunz Have you found a way to prevent the console windows showing up?

[koolkunz]

@koolkunz Have you found a way to prevent the console windows showing up?

Unfortunately no, ultimately I just went with the console window option while building the exe using pynistaller

[Brisk4t]

@koolkunz Look up how to use pyinstaller with a project that uses pluggy as a plugin manager. pytest is an example of such a project. You may also need add ocrmypdf/builtin_plugins to the list of dynamic plugins that will be used - ocrmypdf installs some of its own plugins by default so that others can override them.

pyinstaller --copy-metadata pikepdf --copy-metadata ocrmypdf --collect-submodules ocrmypdf --collect-datas ocrmypdf.data sample.py

This is all I had to do to get it working^.

But now I have another small issue, that is if I am running my GUI application in no console mode, ocrmypdf keeps opening multiple console windows of tesseract (I think equal to the number of jobs being processed) and ghostscript executables with no output.

I already have logging set to -1: configure_logging(verbosity=-1, progress_bar_friendly=False)

Is there any other way to suppress these console windows from showing up at all?

edit: Added Screenshot

If anyone does stumble into this extremely niche issue, a hotfix you can use is to monkey patch the subprocess.Popen() function that OCRmyPDF and its submodules (Tesseract and GhostScript) use to prevent them from opening new windows.

# Patch OCRmyPDF to suppress all consoles
if platform.system() == "Windows":
    # Patch subprocess to prevent window popup on Windows
    class NoWindowPopen(subprocess.Popen):
        def __init__(self, *args, **kwargs):
            kwargs['creationflags'] = kwargs.get('creationflags', 0) | subprocess.CREATE_NO_WINDOW
            super().__init__(*args, **kwargs)

    subprocess.Popen = NoWindowPopen

import ocrmypdf

Note that its important that this block runs before importing ocrmypdf

A more robust solution would be to add a flag to the .ocr function call that then passes on the flag to .Popen()