Update C/C++ API (6.0.0)

[stweil]

Here we can discuss API changes which are incompatible and therefore postponed for a future release 6.

[stweil]

Should be drop the C API in release 6? Or later? Who is still using it?

[egorpugin]

+1 for dropping.

There are some users most likely, but they can take capi.h/cpp into their projects or maintain separate thin library of these two files.

[amitdo]

Who is still using it?

https://github.com/nguyenq/tess4j/

Most likely it is used by many companies internally.

[stweil]

I know from other software that it's possible to mark API functions in the header files as "deprecated", so anybody who uses them in a build process gets a warning message. So we might start adding such deprecation markers and then hopefully get more feedback.

[amitdo]

Previous talk about the C API:

#4074 (comment)

[zdenop]

I still miss reasonable explanation of benefit for removing feature (C-API).
AFAIR contribution came from corporate environment for they want to use tesseract in pascal/Delphi(?). Also for new languages (e.g. rust) is easy to for with C library but problem with C++ (AFAIK).
If it is problem for somebody we can make it as optional feature (with default option to build it)

[egorpugin]

AFAIR contribution came from corporate environment for they want to use tesseract in pascal/Delphi(?).

And they should maintain it.

We can mark capi as deprecated like @stweil said or unmaintained.

[egorpugin]

We can also move out C API (two files) into separate tesseract_c library.

[jbarlow83]

I agree with removal. It's undocumented and in my limited experience, a bit of minefield for memory corruption bugs. pytesseract, tesseract-rs and ocrmypdf (my project) all use Tesseract in a subprocess. pytesseract used to have C API bindings but removed them many years ago.

To mark functions deprecated:

• void myfunc() __attribute__ ((deprecated)); will mark a function as deprecated in GCC / LLVM.
• for MSVC it's probably easiest to deprecate the whole compiler okay I'll be nice, __declspec(deprecated("** this is a deprecated function **"))

[amitdo]

@zdenop,

Do you have an objection to move the C API out of the main tesseract repo, to a new repo 'tesseract-capI' ?

[stweil]

Would this be reasonable as long as we work with major version 5?

[egorpugin]

It is better to create v5 branch and do some fixes into both v5 and master. And declare master as future 'v6' branch and do api changes here.