Extensions

TimeseriesTools exposes most of its analysis surface as package extensions. The function stubs live in core, but each method is only loaded when you using the corresponding package. This keeps the base load light and prevents the package from pulling unwanted heavy dependencies (CUDA, Makie, DSP, …) into every project.

Extension	Loaded by	Provides
DSPExt	using DSP	bandpass, highpass, lowpass, hilbert, analyticphase, analyticamplitude, instantaneousfreq, phasestitch, downsample
DataInterpolationsExt	using DataInterpolations	interpolate, upsample, resample, impute (1-D / per-axis)
DataInterpolationsNDExt	using DataInterpolationsND	interpolate, upsample, resample (joint multi-axis)
ContinuousWaveletsExt	using ContinuousWavelets	waveletspectrogram
CUDAExt	using CUDA, ContinuousWavelets	GPU-accelerated wavelet transform
AutocorrelationsExt	using Autocorrelations	fftacf, dotacf, msdist
ComplexityMeasuresExt	using ComplexityMeasures	apply estimators directly to AbstractTimeseries
GeneralizedPhaseExt	using GeneralizedPhase	generalizedphase
NaturalNeighboursExt	using NaturalNeighbours	2-D scattered interpolation
SignalDecompositionExt	using SignalDecomposition	trend/oscillation decomposition
OptimExt	using Optim	MAPPLE refinement (fit!)
BootstrapExt	using Bootstrap	bootstrapmean, bootstrapmedian, bootstrapaverage

Why extensions?

• Faster cold start: no precompile-time cost for capabilities you don't use.

• No version churn: heavy deps (CUDA, Makie, DSP) move quickly; a weak dep means TimeseriesTools doesn't force a particular version on the rest of your project.

• Discoverable failure mode: calling bandpass(x) without using DSP returns x unchanged (the stub) rather than throwing a MethodError. (See the warning at the top of each function's docstring.)