ImportError: dlopen

[mhluska]

Seeing the following issue on a MacBook M2 Max:

ImportError: dlopen(/Users/mhluska/project/env/lib/python3.11/site-packages/tesserocr.cpython-311-darwin.so, 0x0002): symbol not found in flat namespace '__ZN9tesseract11TessBaseAPID1Ev'

Python version 3.11.4

Using venv with tesserocr 2.6.0

Tesseract version 5.3.2

[zdenop]

You are using tesserocr build against different version of tesseract.

[mhluska]

How would I fix the issue? I've tried reinstalling both tesseract and tesserocr. Do I need a specific version of tesseract with this version of tesserocr?

[mhluska]

I managed to get it working by installing older versions:

Install tesseract 5.3.1-1 specifically (last known working version):

curl -O https://raw.githubusercontent.com/Homebrew/homebrew-core/8bedbf36570ebcbae090227aed6048f7e08f1c76/Formula/tesseract.rb
HOMEBREW_NO_INSTALLED_DEPENDENTS_CHECK=1 brew install tesseract.rb
rm tesseract.rb

Install tesserocr v2.6.0:

pip uninstall tesserocr
pip install --no-cache-dir tesserocr==2.6.0

It would be nice to know how to avoid having to do this though.

[zdenop]

Try to install tesserocr from source instead of pip... (not sure about Mac/brew, but e.g. on Linux you will need the Tesseract and Leptonica dev packages)

[rk-terence]

Using tesseract installed from homebrew, I just successfully installed and imported tesserocr from source.

First use these environment variables so that pip can locate the headers and libraries of tesseract:

CFLAGS=-I/opt/homebrew/include
LDFLAGS=-L/opt/homebrew/lib
PKG_CONFIG_PATH=/opt/homebrew/lib/pkgconfig
CPLUS_INCLUDE_PATH=/opt/homebrew/Cellar/leptonica/1.85.0/include:/opt/homebrew/include
CPPFLAGS=-I/opt/homebrew/include
LIBRARY_PATH=/opt/homebrew/lib

(It is very possible that only a a subset of these environment variables will already do. But I am not sure, so I included all of these)

Then clone this repo into your machine, globally search for lept and replace them with leptonica (I guess this is because the latest tesseract uses a newer version of leptonica that changed the library name from lept to leptonica), and run

pip install . --no-binary :all:

Then I can successfully print the result of get_languages().

My setting:

python 3.12.7
MacOS Sequoia 15.2
tesseract 5.5.0  installed using home-brew