GEOMESA-3447 FSDS - Support allow-list for attribute partition scheme

[dlgundry]

• Added the ability to provided a comma separated list of allowed paths to AttributeScheme
• Allows AttributeScheme to be configured to only partition specific pre-defined allowed paths in FSDS

[elahrvivaz]

seems like we should also update getPartitionName (which isn't used in the converter case, but would in the regular case), so that we don't end up with values that can't be read. i think then we'd also need a default value (instead of the empty string like we have now), maybe as another option, and if it's not specified just use the first in the enumerated list.

[dlgundry]

seems like we should also update getPartitionName (which isn't used in the converter case, but would in the regular case), so that we don't end up with values that can't be read. i think then we'd also need a default value (instead of the empty string like we have now), maybe as another option, and if it's not specified just use the first in the enumerated list.

I think I have captured your comment with the latest commit

[elahrvivaz]

looks good, i think there's just a little bit of inconsistency in when we're dealing with type-encoded values and plan values. i'll try to capture that in another comment in a minute.

[elahrvivaz]

So for the attribute index, we run values through a lexicoder, which maps from a normal value to a sorted stringified value. i.e. if you're dealing with ints, 0 maps to something like 80000000 so that negative values sort correctly. there are 2 methods, AttributeIndexKey.typeEncode (which assumes a valid typed input value) and AttributeIndexKey.encodeForQuery (which will try to convert the input into the correct type). Since we're dealing with strings in the input here, we probably want to run allowedValues and defaultPartition through encodeForQuery in the factory set-up. then i think the only change is we have to run the input value through typeEncode before comparing it in getPartitionName

[elahrvivaz]

@dlgundry you'll also need to sign the Eclipse CLA as outlined here: https://github.com/locationtech/geomesa/blob/main/CONTRIBUTING.md#eclipse-contributor-agreement

[elahrvivaz]

generally, don't force-push, that makes it easier to track changes - then i'll squash down and clean up the messages on merge.

[elahrvivaz]

@fdfea any more comments?

[fdfea]

lgtm 🚀