Optional UCL Support?

[techdragon]

Support getting configuration values from UCL files.

Specifications

It would be nice to include support for UCL / libUCL based configuration files. These are common enough on FreeBSD that it would be useful. https://github.com/vstakhov/libucl

[issue-label-bot]

Issue-Label Bot is automatically applying the label feature_request to this issue, with a confidence of 0.99. Please mark this comment with 👍 or 👎 to give our bot feedback!

Links: app homepage, dashboard and code for this bot.

[naorlivne]

Had a look at it & the only issue I have with it is that the python package in the given link is in fact just a wrapper around the C codebase which means a enduser will have to ensure he has a C compiler to install it & that it will take considerably longer then a dedicated Python parser (both are things you generally want to avoid in today containerized microservice world).

Is there any Python parser your familiar with that doesn't require C compiling for UCL? I'd like to keep the install to just be a pip install without any external dependencies.

[techdragon]

@naorlivne would having pre-built wheels for the python library, with the c library already compiled and included also be a solution?

[naorlivne]

There are a few problem with pre-build wheels as I see it:

• Will require parse_it to be pre-built based on the underlying CPU architecture (X64\ARMv8\ARMvX\etc) so multiple versions will be needed
• Will need to have some hashing to ensure unmodified version of the precompiled c library to reassure endusers the libucl version is unmodified
• Extra complexity during CI/CD

On a personal note I wouldn't see it as a problem if it was a service\product that runs as a full software or inside a container but as an open source library that could be used anywhere it's a bit more of a problem.

Take it you don't know of any pure Python UCL parser instead?

[techdragon]

Very good points there, but I meant the libUCL python library could be improved by having them produce pre-built wheels with their C library instead of leaving it up to end users. I've got some sprint days at a python conference coming up so I figure I could improve their python packaging situation a bit first, then once that was better the library would be more suitable for use in parse_it.

[naorlivne]

If you can get libUCL to work with pip install even if it has precompiled C in it on all CPU architectures (On a personal note I have a few ARM based systems also using it so it will be easy for me to test it on) then that will take care of all the issues I had with it & I see no reason why not to include it.