Module codec 

Fast-field compression codecs with auto-selection.

Four codecs are available, and the writer picks the smallest at build time:

Codec	ID	Description
Constant	0	No data bytes — all values identical
Bitpacked	1	min-subtract + global bitpack
Linear	2	Regression line, bitpack residuals
BlockwiseLinear	3	Per-512-block linear, bitpack residuals per block

Structs

BitpackedEstimator

Min-subtract + global bitpack. This is the existing codec, now behind a tag.

BlockwiseLinearEstimator

Per-512-element-block linear interpolation with per-block bitpacked residuals.

ConstantEstimator

All values are identical → zero data bytes. Value stored in the codec header.

LinearEstimator

Fits y = slope * x + intercept across all values, stores residuals bitpacked.

Enums

CodecType

Codec identifier stored in the column data region (first byte).

Constants

BLOCKWISE_LINEAR_BLOCK_SIZE

Block size for BlockwiseLinear codec (matching Tantivy).

Traits

CodecEstimator

Estimates serialized size for a given codec.

Functions

auto_read

Read a single value from auto-codec encoded data.

auto_read_batch

Batch-read out.len() consecutive values starting at start_index from auto-codec data.

bitpacked_read

Read a single value from a bitpacked-codec column.

bitpacked_read_batch

Batch-read count consecutive values starting at start_index from bitpacked data.

blockwise_linear_read

Read a single value from a blockwise-linear-codec column.

linear_read

Read a single value from a linear-codec column.

serialize_auto

Serialize values using the codec that produces the smallest output.