tesseract_get_version returns cached result after version change

[JohnReid]

The @run_once decorator on pytesseract.tesseract_get_version() means that if the version is switched via

pytesseract.pytesseract.tesseract_cmd = 'tesseract5'

for example, after pytesseract.tesseract_get_version() has been called, it no longer returns correct results.

As a complete example:

import pytesseract
pytesseract.pytesseract.tesseract_cmd = 'Tesseract/tesseract4'
print(pytesseract.get_tesseract_version())
pytesseract.pytesseract.tesseract_cmd = 'Tesseract/tesseract5'
print(pytesseract.get_tesseract_version())

gives the following (even though I have switched from version 4 to version 5)

4.1.3
 leptonica-1.78.0
  libgif 5.1.4 : libjpeg 8d (libjpeg-turbo 1.4.2) : libpng 1.2.54 : libtiff 4.0.6 : zlib 1.2.11 : libwebp 0.4.4 : libopenjp2 2.3.0
 Found AVX512BW
 Found AVX512F
 Found AVX2
 Found AVX
 Found FMA
 Found SSE
 Found libarchive 3.1.2

4.1.3
 leptonica-1.78.0
  libgif 5.1.4 : libjpeg 8d (libjpeg-turbo 1.4.2) : libpng 1.2.54 : libtiff 4.0.6 : zlib 1.2.11 : libwebp 0.4.4 : libopenjp2 2.3.0
 Found AVX512BW
 Found AVX512F
 Found AVX2
 Found AVX
 Found FMA
 Found SSE
 Found libarchive 3.1.2```

[bozhodimitrov]

This behavior is as intended.
You can use the non-cached version by accessing pytesseract.tesseract_get_version.__wrapped__()
This is a legacy thing with the pytesseract cli -- and to be honest, it should be extremely rare to switch between different tesseract versions during a single session.

[JohnReid]

Thanks for the quick response and the work-around.

I can see the behaviour is intended, it is just undocumented and totally counterintuitive. Memoising the function return value conditioned on the value of tesseract_cmd would satisfy your caching needs and also provide a consistent, intuitive interface.

Also, I don't think it is as rare as you think to switch between versions in a session, e.g. #279

[bozhodimitrov]

I see what you mean, but this still could be ambiguous. For example if the user has two different tesseract versions in a different path but with the same binary name. I don't see how we are going to distinguish, between both of them if we don't make the expensive call in a sub shell (keep in mind that get_tesseract_version is used for some of the image_to_ methods)

Maybe we can just add a cache flag and if someone needs the non-cached version, they can make the function call with this flag.

[JohnReid]

I see what you mean as well. Your last suggestion sounds the simplest and the best.

[elvinagam]

How do you actually update to tesseract 5.0?
I am using pytesseract on colab, and when I pip installed it there, I got 4.0.0. How do I switch to updated tesseract? @int3l

[stefan6419846]

As pytesseract just calls the OS package, you probably want to update Tesseract there. It seems like Google Colab is Ubuntu-based, which can be verified using !cat /etc/*release. According to https://packages.ubuntu.com/search?keywords=tesseract&searchon=names&suite=all&section=all, you will need at least Ubuntu 22.10 to get version 5 from the official package repositories.

Otherwise, you might be able to build the package yourself, which is documented in https://tesseract-ocr.github.io/tessdoc/Compiling-%E2%80%93-GitInstallation.html. With some searching on the web, you will find some custom PPAs for older Ubuntu versions which provide Tesseract 5 as well, although I will not link them here as especially with random PPAs you should have a good look at whether you can trust it or not.

[elvinagam]

DISTRIB_ID=Ubuntu
DISTRIB_RELEASE=18.04
It actually is. Thanks, I'll look for it.

btw, Is it the same procedure for windows to update to tesseract 5.0.0 though?

[stefan6419846]

Apart from the fact that Windows usually has a different package management, yes. All downloads/installers seem to be unofficial though in this case: https://tesseract-ocr.github.io/tessdoc/Downloads.html