Bind keyset

Coming soon

This capability will be available in an upcoming release.

Please use BIND_VALUE(), BIND_TABLE() or BIND_SQL() to define the entity space manually.

FeatureQL's key selection approach varies between online and offline evaluation scenarios, requiring different strategies for defining the scope of feature computation.

Online vs. Offline evaluation patterns

Online evaluation:

  • Precise key specification: "Calculate revenue for user_id = 1"
  • Real-time feature serving with specific entity instances
  • Typically involves BIND_VALUE() or small BIND_TABLE() operations

Offline analytics:

  • Subset selection: "Calculate revenue for all orders from yesterday"
  • Batch processing across entity populations
  • Requires efficient key space definition and filtering

The BIND_KEYSET() function is designed specifically for offline analytics scenarios, enabling sophisticated key selection based on business criteria while maintaining FeatureQL's entity-centric feature evaluation principles.

Last update at: 2025/10/13 10:23:46