versioning and/or tag

[ThomasSaout]

I'm currently a PhD student, working under the CIFRE (French co-financing program). I'm developing a solution that uses your solver, but with the current change I need to review my deployment procedure. Could you put a version tag in this repo to avoid this problem in the future?

Thank you in advance

[ciaranm]

Was it the switch to CMake that caused problems? Part of the reason I did that was to make integration into other projects somewhat easier, so hopefully this kind of thing won't happen so much in the future. This is also why I've started moving headers into separate directories -- I still need to move them into namespaces, and convert to more modern C++, so there's probably at least one more round of breakage to go in the short term.

Can you tell me which parts of the API you're using, and want to be stable? I don't know the details of how people are using the solver, so I'm not sure which parts I should try to keep unchanged.

[ThomasSaout]

Yes the "problematic" change is the switch to CMake, indeed this change is minor just need to add "build" in some path and change the call for deploy from make to the new instructions into readme. I suggest some tagging to permit to identify stable commit and to make easy checkout to maitain solution who is configurate for older version.

I mainly use the glasgow_subgraph_solver to find eligible subgraphs, so if the usage and settings don't change, my solution will remain stable. The context of use of your solver is to extract tables from a scanned document, by modeling the characters extracted by OCR in this document as a graph, which allowed to find patterns corresponding to these tables and to use your solver to solve efficiently this problem of isomorphic subgraphs.

Thank you for your prompt reply

[ciaranm]

Ah OK. If you're just calling the binaries, I'm probably done breaking things for you for the foreseeable future. I'm currently trying to tidy things up a bit to get us to the point we can have a first stable release. I would recommend trying to update to using the new build system if you can. I'll try to remember to tag something before I start any future major changes (e.g. to the output format).

[ciaranm]

PS if you've not seen this, https://mcfelix.me/project/glyphgraph/ might be of interest. A while ago we tried something similar with using graphs for OCR and couldn't get it to work, but that doesn't mean it's not feasible, just that the way we tried wasn't the right one...

[ThomasSaout]

It's really interesting work but it's my words that lead you to a misunderstanding. I don't do OCR, but I use a professional OCR model to extract the characters, organise them into words or sentences, and there words are the vertices of the graph representation.

I can't share my work now, because of the publication process, but if you want it when it's finished I can send it to you, it's clearly easier to understand with diagrams and if it helps you show use cases for your solver.

Best regards

[ciaranm]

Thanks, I'd be very interested to see what you're doing when it's ready.