Move source files which are used for training only to src/training

[stweil]

They are moved from src/classify and src/lstm to src/training.

This reduces the size of the Tesseract library.

Signed-off-by: Stefan Weil sw@weilnetz.de

[stweil]

Together with pull request #2613 this completes the symbol cleanup for the Tesseract library which now only contains symbols needed for the Tesseract executable and for the C API.

The Tesseract library is reduced significantly in size by both pull requests:

3590092	  47504	  15616	3653212	 37be5c	/tmp/libtesseract.so.5.0.0-old
3451063	  45808	  15616	3512487	 3598a7	/tmp/libtesseract.so.5.0.0-new

[stweil]

Updated PR to fix CMakeLists.txt and src/training/CMakeLists.txt, so CI should pass now.

[stweil]

@egorpugin, please review my changes for CMake. I don't know why there are three libraries for training and added the moved code to one of them:

src/training/libunicharset_training.a
src/training/libtessopt.a
src/training/libcommon_training.a

Wouldn't it be simpler to have a single library for training?

[egorpugin]

Wouldn't it be simpler to have a single library for training?

Only in case when files are used in >1 target.
Otherwise personally I prefer to keep files near users (when use count ==1).

[egorpugin]

D:/dev/tesseract/src/classify/trainingsample.cpp(27): fatal error C1083: Cannot open include file: 'intfeaturemap.h': No such file or directory

Main source file depends on training. Is it ok?

[stweil]

@egorpugin, yes, some files in src/classify require include files which are now in src/training. I just notices that sw.cpp also needs some changes. Should I try to fix that?

[egorpugin]

Yes, please. Now it is very dummy implementation copied from cmake and not makefiles (which seems is in better shape).
Are you using win or linux?
Since I got first working sw builds on linux, I can try to build tess there for the first time, so it will be usable there.

[egorpugin]

I've removed moved files from the installation, but that single error is still there.