tsdownsample 0.1.0a3

Time series downsampling in rust

tsdownsample

📈 Time series downsampling algorithms for visualization

Features ✨

• Fast: written in rust with PyO3 bindings
  • leverages optimized argminmax - which is SIMD accelerated with runtime feature detection
  • scales linearly with the number of data points
  • multithreaded with Rayon (in Rust)
    Why we do not use Python multiprocessing

    Citing the PyO3 docs on parallelism:
    

    CPython has the infamous Global Interpreter Lock, which prevents several threads from executing Python bytecode in parallel. This makes threading in Python a bad fit for CPU-bound tasks and often forces developers to accept the overhead of multiprocessing.

    In Rust - which is a compiled language - there is no GIL, so CPU-bound tasks can be parallelized (with Rayon) with little to no overhead.
• Efficient: memory efficient
  • works on views of the data (no copies)
  • no intermediate data structures are created
• Flexible: works on any type of data
  • supported datatypes are f16, f32, f64, i16, i32, i64, u16, u32, u64
  !! 🚀 f16 argminmax is 200-300x faster than numpy

  In contrast with all other data types above, f16 is *not* hardware supported (i.e., no instructions for f16) by most modern CPUs!!
  🐌 Programming languages facilitate support for this datatype by either (i) upcasting to `f32` or (ii) using a software implementation.
  💡 As for argminmax, only comparisons are needed - and thus no arithmetic operations - creating a symmetrical ordinal mapping from f16 to i16 is sufficient. This mapping allows to use the hardware supported scalar and SIMD i16 instructions - while not producing any memory overhead 🎉
  More details are described in argminmax PR #1.
• Easy to use: simple & flexible API

Install

❗🚨❗ This package is currently under development - no stable release yet ❗🚨❗

pip install tsdownsample

Usage

from tsdownsample import MinMaxLTTBDownsampler
import pandas as pd; import numpy as np

# Create a time series
y = np.random.randn(10_000_000)
s = pd.Series(y)

# Downsample to 1000 points
s_ds = MinMaxLTTBDownsampler.downsample(s, n_out=1000)

---

👤 Jeroen Van Der Donckt