Adaptive Lossless Floating-Point Compression in Embedded OLAP Databases

We disclose an innovative integration of Adaptive Lossless floating-Point (ALP) compression with embedded column-oriented storage engines. This work extends the academic ALP algorithm (Afroozehet al., VLDB 2023) with novel database-specific optimizations: adaptive pattern detection for decimal-origin vs. high-precision data, Frame-of-Reference encoding with dynamic bit-width packing, and SIMD-compatible decoding architecture enabling 2-10x compression ratios with sub-microsecond decompression latency. The implementation integrates seamlessly with multi-version concurrency control (MVCC) version chains and incremental materialized view (MV) refresh workflows, establishing prior art for compression techniques in time-travel and incremental update scenarios.