Expand description
Codec auto-detection from column type and data distribution.
Analyzes up to the first 1024 values of a column to select the optimal
codec chain. Called at flush time when ColumnCodec::Auto is configured.
Selection strategy:
- Partitions ≥ 1024 values → cascading codecs (ALP, FastLanes, etc.)
- Partitions < 1024 values → single-step codecs (Gorilla, Delta, etc.)
- f64 with >95% ALP encodability →
AlpFastLanesLz4 - f64 with ≤95% ALP encodability →
Gorilla(fallback) - i64 timestamps/counters →
DeltaFastLanesLz4 - Symbol columns →
FastLanesLz4(small integer IDs)
Functions§
- detect_
codec - Detect the optimal codec for a column based on its type and data.
- detect_
f64_ codec - Detect the optimal codec for an f64 column by analyzing the data.
- detect_
i64_ codec - Detect the optimal codec for an i64 column by analyzing the data.