mlcommons / croissant

Croissant is a high-level format for machine learning datasets that brings together four rich layers.
https://mlcommons.org/croissant
Apache License 2.0
452 stars 41 forks source link

Uniform jsonQuery and jsonPath #746

Open ccl-core opened 2 months ago

ccl-core commented 2 months ago

In the croissant specs, we use two different terminology: jsonQuery for transforms (https://docs.mlcommons.org/croissant/docs/croissant-spec.html#transform) and jsonPath for extract (https://docs.mlcommons.org/croissant/docs/croissant-spec.html#extract).

The latter is defined here: https://goessner.net/articles/JsonPath/ . I don't know if the former defines a different query language.

It is not really clear what the difference between the two is, given that both can be used interchangeably (see e.g. a Transform operation using JsonPath here).

Possibly, renaming JsonQuery to JsonPath in the specs could avoid redundancy and make the definitions clearer.