A partitioned table routes rows into child tables according to a partition key. Range bounds include their lower value and exclude their upper value. January therefore runs from January 1 through values before February 1. A row on February 1 belongs in February.
Partition pruning can avoid reading children that can't match a filter. It is different from an index lookup within a retained child. Partitioning doesn't automatically accelerate a query that reads every period, and too many partitions add planning and maintenance costs.
Choose a retention and maintenance process before adopting the structure. Detaching an old partition preserves it as a standalone table; dropping it removes its data. A unique constraint on a partitioned parent generally must include its partition key. These exercises create real partitions, inspect bounds and check which physical child receives boundary rows.
The date key defines which period receives each reading.