Enum EncodingType 

#[repr(u8)]
pub enum EncodingType {
    DeltaOfDelta = 0,
    Chimp = 1,
    Gorilla = 2,
    IntegerI64 = 3,
    IntegerU64 = 4,
    Dictionary = 5,
    Bitmap = 6,
    PlainF64 = 7,
    PlainI64 = 8,
    PlainU64 = 9,
    PlainBool = 10,
    PlainString = 11,
    Nullable = 12,
    Rle = 13,
    VarintI64 = 14,
    VarintU64 = 15,
    Chimp128 = 16,
    ForI64 = 17,
    ForU64 = 18,
    Patas = 19,
    Alp = 20,
    Pco = 21,
    PcoI64 = 22,
    PcoU64 = 23,
    DecimalI128 = 24,
}

Discriminant tag stored as the first byte of every EncodedBlock.

Variants

DeltaOfDelta = 0

Delta-of-delta for timestamps.

Kept as a selectable codec — it still wins on a perfect metronome — but no longer the default: see ColumnEncoder::encode_timestamps.

Chimp = 1

Chimp XOR for floats.

Gorilla = 2

Gorilla XOR for floats.

IntegerI64 = 3

Delta + ZigZag for i64.

IntegerU64 = 4

Delta + ZigZag for u64.

Dictionary = 5

Dictionary encoding for strings.

Bitmap = 6

Bitmap for booleans.

PlainF64 = 7

Plain f64.

PlainI64 = 8

Plain i64.

PlainU64 = 9

Plain u64.

PlainBool = 10

Plain bool.

PlainString = 11

Plain string.

Nullable = 12

Nullable wrapper — validity bitmap + inner encoded payload.

Payload format:

[inner_encoding: u8][total_count: u32 LE][validity_bitmap: ceil(count/8) bytes][inner_payload…]

Bit 1 in the validity bitmap means the value is present; 0 means null. Only non-null values are encoded in inner_payload.

Rle = 13

Run-length encoding for columns with repeated consecutive values.

VarintI64 = 14

Delta + ZigZag + LEB128 varint for i64.

Better than IntegerI64 for sparse or highly variable data where most deltas are small but occasional outliers would inflate the fixed bit width.

VarintU64 = 15

Delta + ZigZag + LEB128 varint for u64.

Chimp128 = 16

Chimp128 ring-buffer XOR for floats.

ForI64 = 17

Frame-of-Reference for narrow-range integers.

ForU64 = 18

Frame-of-Reference for narrow-range unsigned integers.

Patas = 19

Patas byte-aligned XOR for floats (VLDB 2023).

Alp = 20

ALP adaptive lossless floating-point compression (SIGMOD 2024).

A second opinion behind Pco for f64 columns whose values originated as decimals — which most sensor, meter and price data does. See crate::alp.

Pco = 21

Pcodec for floats (Loncaric 2025): mode recovery, delta, binned ANS. Leads every float candidate list — see crate::pco.

PcoI64 = 22

Pcodec for signed integers, and the default timestamp codec.

PcoU64 = 23

Pcodec for unsigned integers.

DecimalI128 = 24

Exact decimal mantissas (i128) — see crate::decimal.

Not a codec of its own so much as a shape: the narrow form, which is nearly every real decimal column, delegates straight to the i64 stack above and therefore inherits RLE, FOR, varint and pco.

Implementations

impl EncodingType

pub const ALL: &'static [Self]

Every encoding type, in tag order.

Complete by construction — generated from the same list as the enum itself, so a codec cannot be added without appearing here, and therefore in the fuzz corpus.

pub fn from_tag(tag: u8) -> Result<Self>

Convert a u8 tag to an EncodingType.

Errors

Returns EncodingError::UnsupportedEncoding for unknown tags.

pub fn tag(self) -> u8

Return the u8 discriminant.

Trait Implementations

impl Clone for EncodingType

fn clone(&self) -> EncodingType

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for EncodingType

impl Debug for EncodingType

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Display for EncodingType

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Eq for EncodingType

impl PartialEq for EncodingType

fn eq(&self, other: &EncodingType) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for EncodingType