+1 for having this new feature.
Not having to define types in the fully open case has been a bit of a
pain forever. This is nice change!
I wonder if we should also (could be later) add an option to not have to
talk about a key at all - e.g., should we also support something like this:
CREATE COLLECTION my_standalone_ds;
This would be similar to the auto-generated case - under the hood it
would be the same, actually, except even the key name would be
system-generated. (It would be recorded in the collection's metadata,
of course.) This would be for use when there just isn't a key that
makes any sense and one doesn't want to bother introducing its existence
into the user-level data model. (It would just be under the hood to be
used for hash-partitioning the data and referencing it from secondary
indexes - all innards stuff.)
Cheers,
Mike
On 10/28/24 3:07 PM, Peeyush Gupta wrote:
Initiating discussion for adding a feature to create dataset without type
specification
Feature: Create Dataset Without Type Specification
Details: Currently, users must create a type before creating a dataset even if
the type is fully open (with only primary key fields defined in it). With this
feature the users will be able to create fully open datasets without requiring
them to create a type explicitly.
APE:https://cwiki.apache.org/confluence/display/ASTERIXDB/APE+14%3A+Datasets+without+type+specification
Thanks,
Peeyush
