Francisco Bischoff
16 Oct 2026
write_json()
and read_json() to avoid masking
base::write(). JSON round trips now preserve profile
vectors and the origin and partial
attributes.mstomp() and
mstomp_par()): Delegated computation to the
high-performance native C++ engine in
matrixprofiler::mstomp(), providing massive speedups
(~200x) and native multi-threading while removing legacy dependencies on
doSNOW and foreach.mode = "mpx" to tsmp() as a
first-class citizen algorithm delegating to
matrixprofiler::mpx().compute() now delegates
internally to tsmp(mode = "mpx") and pmp(),
preserving 100% compatibility with Matrix Profile Foundation unified API
(Python/Go) and standard MatrixProfile / PMP
S3 classes.mpdist() now delegates to
native C++ engine in matrixprofiler::mpdist(), delivering
major speedups and lower memory footprint to mpdist() and
find_snippet().fluss_cac() using tabulate(),
eliminating slow \(O(N)\) R loops.n_workers parameter to
find_snippet(), enabling multithreaded MPdist
computation.find_snippet() when time
series length is an exact multiple of snippet size.dist_profile(), compute(),
find_snippet() and pmp() with hardened
multi-worker validation against detectCores() == NA.::: calls in
dist_profile.R with public API calls to
matrixprofiler::mass().na_strategy = c("monolithic", "segmented", "segmented_native")
to tsmp() with end-to-end missing value protection for
discords and motifs.paa() and ipaa()) implemented and
maintained in pure R.audio dependency and audio completion
signals.NeedsCompilation: no) delegating heavy computations to
matrixprofiler.mpx().find_snippet() that uses MPdist to show
representative data.mpx() algorithm that doesn’t depends on FFT.pmp() pan-matrix profile.compute(), analyze() and
visualize(). Starting point in the unified API from
MPF.mass() to dist_profile(),
including options to different algorithms.dist_profile()
algorithms.dist_profile() allows Query with Gap (QwG) and
Approximate Distance Profile (ADP, with PAA)stompi_update() that updates the matrix profile
allowing real-time computation.floss() which can do real-time FLUSS
computation.[ for tsmp
objects.tail() and head() for
tsmp objects.find_motif() for
MultiMatrixProfile to report Motifs correctly.find_discord() and its print() and
plot() functions.plot() for motifs to show where are the
neighbors. Same for discord.valmod() for Variable Length Motif
Discovery.find_motif() for compatibility with
valmod().find_chains() not returning the longest chain.
Issue #33vars. Issue #23progress
package.tsmp()
output for example.%>% (pipe),
e.g. tsmp() %>% find_motif(). Except for SDTS that has a
proper way to work.tsmp() that handles the
several algorithms available.as.* functions to allow you to switch classes if
you want, e.g.: as.matrixprofile().as.* functions).beepr to audio
(actually beepr depends on audio, so fewer
dependencies).NEWS.md file to track changes to the
package.