[Suggestion] Directories reorder and renaming

[amitdo]

Hello!

Nice work :-)

My suggestions:

• Rename 'Modules' to 'Lib'.
• Create a new folder named 'Progs' and move all the utilities (accci, accdist, accsum, etc.) to this folder.
• Create a new folder named 'Man' and move all the manpages to there.
• Create a new folder named 'src' and move 'Lib' and 'Progs' (and others) to there.

[eddieantonio]

@amitdo Hey, thanks!

I left these folders with the same structure because I really only ported the old code written in 1996 by Stephen Rice to modern computers, and added UTF-8 support. The only thing I did in terms of structure is move everything from analytic-tools to the root directory.

Why would you suggest these changes?

[amitdo]

Hi again!

I reviewed your commits, so I know what changes you made so far.

Why would you suggest these changes?

Well, It's a matter of style. Many open source projects have a directory layout that include 'src' and/or 'lib' dirs. Some projects use 'prog(s)' dir, but it's less common.

https://github.com/kriasoft/Folder-Structure-Conventions
https://github.com/DanBloomberg/leptonica

It's just a suggestion, feel free to not accept it... :)

BTW, a few years ago I started rewriting this project in Python. Never finished it...

[amitdo]

FYI, I added a link to this project to Tesseract wiki:
https://github.com/tesseract-ocr/tesseract/wiki/3rdParty

I hope this is ok with you.

[eddieantonio]

Heya!

Sorry if I sounded offended in my last post; I have a conference submission deadline looming and a major assignment due soon, so adding more potential work for me right now is like 😫

That said! This is a good idea! 😄 However, I like a more unix-y structure a bit better:

.
├── bin                 # Compiled binaries of utilities
├── include             # Public C/C++ API (does not really exist yet...)
├── lib
│   └── libisri.a       # Used for internal linking
│   └── libisri.so.60   # Used for dynamically linking with external applications
├── libexec             # Scripts used internally
├── man
│   └── man1            # Man pages for each binary
│   └── man3            # Man pages for public API?
├── src                 # Source of libisri and binaries
└── test                # Tests

This way, developers on Unix-y systems can simply add each directory to their respective paths:

ISRI=$HOME/isri-ocr-evaluation-tools # or wherever you decide to put this repo
export PATH=$ISRI/bin:$PATH
export MANDIR=$ISRI/man:$MANPATH
export LD_LIBRARY_PATH=$ISRI/lib:$LD_LIBRARY_PATH

I've suggested a public API, which may be a lot of work, but if implemented, you could link with ISRI libs like so:

g++ \
    -I$HOME/isri-ocr-evaluation-tools/include \ # Typically set in CPPFLAGS in a Makefile
    -L$HOME/isri-ocr-evaluation-tools/lib     \ # Typically set in LDFLAGS in a Makefile
    mycoolocrthing.cc -o mycoolocrthing       \ 
    -lisri                                      # Link with ISRI OCR evaluation tools

Edit: And yeah, it's okay for this project to be listed on the Tesseract page! My colleagues wanted to evaluate their Tesseract-based OCR engine and they found the ISRI tool to be the best for it (being dissatisfied with more modern options like ocrevalUAtion).

[amitdo]

👍

But do it only if/when you have enough free time. :)