Support multiple JSON Schema Validator Backends?

As implicitly acknowledged in #46, validating against the JSON schema is expensive, slow, and when running things at scale, really making me think about skipping it, which is a red flag. If I’m thinking about bailing on validation, others aren’t far behind me.

I did a little digging, and it seems this is more a function of jsonschema being slow than schema validation being inherently slow. For example, this: https://github.com/epoberezkin/ajv#performance

Couple thoughts:

• I recognize that these are not Python libraries, but it demonstrates the variability in validator performance.
• The fastest one has a command line interface that we could consider wrapping. This wouldn’t (and shouldn’t) be the default validator, but multiple backends could allow for more flexible validator support.
• There don’t actually seem to be many “native” options for JSON schema validation in Python, so wrapping one in another language might be simplest way to speed this up.

thoughts? complete heresy?

[ejhumphrey]

Ya, to be perfectly clear, I also don’t want to dabble in subcomponent / arbitrary JObject validation … but I think it might make sense to do this at the Annotation level rather than a Jams object, since the former tend to be more like documents, and the latter more like containers.

as for profiling … I read Jams files far more than I write / modify them, so therefore the majority of my time validating Jams files is wasted (converging to “all of my time” as t -> inf). 😄

[bmcfee]

Is there any reason to keep this open, or can we close it out as a non-fixer?