Compress (algostream.Algostream_time

Module Algostream_time_series.Compress

Lossless float column compression (Gorilla-style XOR-delta-of-bits) and integer column compression (delta + varint).

Float strategy: encode each value as Int64.bits_of_float, XOR with the previous value, write the leading-zero count + meaningful-bits block. Round-trips exactly for ALL Float64 values — NaN, Infinity, denormals, signed zero — because the encoder never interprets the value as a number.

Int64 strategy (timestamps): write the delta against the previous value as a zigzag-varint. Monotonically-increasing timestamps with sub-millisecond gaps compress to ~1-2 bytes per value.

val encode_float : 
  (float, Stdlib.Bigarray.float64_elt, Stdlib.Bigarray.c_layout)
    Stdlib.Bigarray.Array1.t ->
  bytes

Encode a Float64 Bigarray. Returns a Bytes.t you can persist or ship over the wire.

val decode_float : 
  bytes ->
  (float, Stdlib.Bigarray.float64_elt, Stdlib.Bigarray.c_layout)
    Stdlib.Bigarray.Array1.t

val encode_int64 : 
  (int64, Stdlib.Bigarray.int64_elt, Stdlib.Bigarray.c_layout)
    Stdlib.Bigarray.Array1.t ->
  bytes

val decode_int64 : 
  bytes ->
  (int64, Stdlib.Bigarray.int64_elt, Stdlib.Bigarray.c_layout)
    Stdlib.Bigarray.Array1.t