Modular version of the validator

[harisont]

On several occasions, @ellepannitto and I have felt the need to import the validator in external programs (similar to what @bguil does in Arborator-Grew) and/or to only use specific checks, non necessarily coinciding with the existing validation levels. Similarly, there are cases in which we want to perform additional checks, e.g. for treebank-specific MISC features, while still relying on the official validator for the remaining fields.

As far as we know, validate.py is not packaged as a library. We wonder if a modular version of the validator exists, and if not, if there is interest for such a version of it. If so, we would be willing to code it reusing the existing code and keeping the interface unchanged.

[dan-zeman]

As far as we know, validate.py is not packaged as a library. We wonder if a modular version of the validator exists

I am not aware of any other version of the validator.

[bguil]

To be fair, I have to say that the integration of the validator in ArboratorGrew was done by @khansadaoudi!
I let her explain how it was done, but for sure a library version would have made this integration easier.

[khansadaoudi]

@harisont the integration of the validator in ArboratorGrew is dependent on the project language. If the project language is indicated it will validate on the language level, otherwise it will validate the level 3. For the integration, I had to make few updates to the script you can find them here and they are commented. But As @bguil said, if we had a library version the integration would have made the integration easier.

[AngledLuffa]

I'd be interested in lending some time to turn these scripts into a pip installable package. The scorer would be another useful component to be able to say "import udtools" (that piece can already be used as a library, but isn't available as an installable package)

[harisont]

Thank you @AngledLuffa!

@dan-zeman are you on board too?
Since you are the main maintainer of the validator, we want to ensure that the "library version" of the validator could eventually become the official validator. In other words, we are willing and happy to work on this if it is useful for all parts: ending up with two versions needing each its maintenance would only create duplication of work.

[dan-zeman]

@dan-zeman are you on board too?

Yes, I guess I am.

I have been thinking for a long time that the code would benefit from a major rewrite. But I never had enough time, and also it is critical to not break anything that relies on the script. That's why I prefer incremental changes as much as possible.

If we can make it importable as a module while making sure that it can be still called as a script with the same arguments and options as before, and its output has the same form as before, that would be great and I'm willing to participate in the endeavor.

[harisont]

Great!

@ellepannitto and I can start setting up a repository some time this week, but we will a few days in a row to focus on the implementation in the beginning of September, when I'll be with her in Bologna.

Maybe before that we can have a call to coordinate the work a bit?

If we can make it importable as a module while making sure that it can be still called as a script with the same arguments and options as before, and its output has the same form as before, that would be great and I'm willing to participate in the endeavor.

And yes, this is exactly what we're aiming for!

[dan-zeman]

@ellepannitto and I can start setting up a repository some time this week

Is there a reason for not doing it in just a new branch of the UD tools repository?

[harisont]

I was thinking a fork but no, simply a branch works just as well. I hadn't realized I had push rights.

[dan-zeman]

I made some changes to the validator code in the master branch recently. I occasionally merge it to the validator-refactoring branch so that the validate.py there does not go totally out of sync. But I did not do anything with the copy of the code you have in the validator subfolder, so that one is outdated.

Let me know if you want to discuss the modifications that will be needed to make the validator importable as a library.

[harisont]

Good to know @dan-zeman!

@ellepannitto might have started actively looking into this already, but we plan to get a large chunk of the work done together between the end of August and the beginning of September. We'll get in touch then if not before.

[AngledLuffa]

Any progress on this? I was looking for something similar

[ellepannitto]

@harisont and I are planning to make most of the work on this next week, while she will be visiting Bologna next week and we will have time to actually do some coding together and set the ball rolling.
Since (I think) we're all in SyntaxFest now, shall we maybe have a small meeting to gather suggestions, see how to collaborate and if we can do anything to make future work on this smoother? Given it won't be all finished in a week obviously...