Stumpy

Latest version: v1.12.0

Safety actively analyzes 613603 Python packages for vulnerabilities to keep your Python projects secure.

Scan your dependencies

Page 1 of 5

1.12.0

---------------------------
* bugfixes
- Fixed prescrump, scrump, and scraamp miscalculation of AB-joins
- Fixed bug in `is False` by converting to `bool` type
- Fixed bug in `snippet` that caused loss in precision, add unit tests
- Fixed RTD incompatibility with urllib3
- Fixed loss of precision in distances computed for self-matches
- Fixed libiomp5.dylib Github Actions location
- Fixed incorrect post-processing in naive.mass_PI
- Fixed incorrect number of contiguous windows in snippets
* features
- Improved matrix profile performance (15-20%) with uint64 indexing
- Added top-k nearest neighbor feature
- Added rolling_isconstant function
- Added `subseq_isconstant` to API for transparent handling of constant time series subsequences
- Added parallelized rolling_nanstd
- Added abstraction layer for distributed client functions (e.g., Dask, Ray, etc)
- Added initial support for `numba` function caching
- Added Python 3.11 support
- Added `query_idx` to improve distances computed for self-matches in `motifs` function
- Added `mmotifs` for multi-dimensional motif discovery
- Added function `process_isconstant`
* tasks
- Refactored mpdist
- Added MERLIN notebook reproducer
- Added negative index checks to mmotifs
- Fixed mybinder badge and links
- Renamed nave.mass to naive.mass_PI
- Added `row_wise` parameter to naive.stump
- Added input type (list) check to ostinato
- Added ability to return fully filled bfs indices
- Updated setup.py to enable Github dependency graph tracking
- Improved stability of prescrimp
- Split coverage tests for more verbosity
- Added more features to custom test function
- Removed explicit cancellation of dask futures
- Removed max `dask`/`distributed` version requirement
- Optimized stumpi and aampi class init
- Improve aampi update behavior with constant sequences returning nan
- Added pytest notebook link checking
- Refactored `match` function
- Added warnings to motifs and aamp_motifs
- Added `numba -s` step in Github Actions workflow
- Added explicit link to OpenMP for MacOS Github Actions workflow
- Moved to Actions/Checkout V3 in Github Actions
- Improved numba function signatures
- Added '--editable' install mode to setup.sh
- Updated Github Action codecov/codecov-actionv1 to v3
- Maintained 100% code coverage
- Used unittest.mock.patch to prevent overwriting of config variables during testing
- Added Python 3.11 to test matrix
- Added ability to detect missing parameter definitions in docstrings (docstring.py)
- Updated pytest flags to report skips and added additional summary
- Replaced logging.warning with warnings.warn
- Improved multi-line warnings
- Ensured `bfs_indices` are sent to correct GPU device
- Refactored window size check in mass/mass_absolute functions
- Added animated GIF to README
- Updated minimum black version
- Removed `_parallel_rolling_func` as it conflicted with `numba` caching
- Removed mamba timeout
- Improved warnings
- Added missing p-norm param to idx_to_mp and floss functions
- Added test failure when coverage is below 100%
- Added boolean array test for rolling_isfinite
- Added isort, resolved circular dependencies, updated examples
- Only build HTML for RTD
- Added `mp` param to stumpi to allow pre-computed matrix profile as input
- Added various unit tests
- Removed codecov as dependency
- Added ability to test the execution of tutorial notebooks
- Added check for negative matrix profile indices
- Added `test_precision.py` for all issues related to loss-of-precision
- Fixed tls deprecation warning
- Replaced np.int with np.int64
- Specified fastmath flags to include nan/inf values in inputs/outputs
- Replaced bool dtype with np.bool_
- Replaced np.newaxis with np.expand_dims
- Added check for docstring and parameter mismatch
- Update URLs for minimum version references
- Added explicit shell declaration in Github Actions workflow
- Show OpenMP libraries in Github Actions workflow
- Added `pip.sh` script for setting up dev environment using `pip`
- Removed parallel=True in `core._compute_multi_PI`
- Updated coverage testing to include all modules
- Improved code consistency for `T_A` and `T_A` definitions
- Refactored `test.sh` and include missing test files in unit tests
- Added minimum dependency compatibility script (min.py)
- Updated minimum dependency bumping instructions
- Bumped minimum Python version to 3.8
- Update PyPI downloads badge
* documentation
- Improved/updated various docstrings
- Added shapelet discovery tutorial
- Clarified unanchored chain description
- Fixed typos
- Improved `core._get_QT docstring`
- Fixed imbalanced tree representation in docstring
- Improved scrump documentation
- Made light mode default and remove theme switcher from header nav bar
- Improved dataframe layout display in tutorials
- Added multi-dimensional motif and match tutorial
= Added T_subseq_isfinite to docstring
- Added tutorial for "Discovering motifs under uniform scaling"
- Updated docs for using a dask client
- Fixed malformed link in floss docstring
- Added missing parameter section in various docstrings
- Added Minkowski docstring for Euclidean distance
- Added missing parameters for GPU functions in docstrings
- Improved documentation for `P` in motifs function

1.11.1

---------------------------
* bugfixes
- Fixed 582 Allow 1D mean/stddev inputs for `stumpy.match`
* features
- N/A
* tasks
- Added mmotifs and aamp_mmotifs to __init__.py
* documentation
- Added mmotifs docstring in RTD API

1.11.0

---------------------------
* bugfixes
- Fixed 576 Incorrect stimp, stimped, gpu_stimp normalize rerouting
- Fixed bad index in `naive.stimp` normalization method
- Fixed unit tests
- Fixed `cutoff=np.inf` edge case in `_motifs` function
* features
- Added `mmotifs` and `aamp_mmotifs` functions for multi-dimensional motif discovery and unit tests
- Added all AAMP/p-norm `stimp` implementations
- Added `atol` parameter to `motifs` function
- Added parallelized `prescraamp`
- Added parallelized `prescrump`
- Added `_get_ranges` function
- Added `shoelace` formula for computing total diagonal ndists
- Added `p_norm` support
- Added `_preprocess` function
- Added minimum description length function, `mdl`
- Added Python 3.10 support
- Added AAMP/p-norm support for pan matrix profiles
* tasks
- Fixed typos
- Added `gpu_aamp_stimp` driver error
- Increased minimum dependencies
- Updated coverage testings to include all unit tests and additional modules
- Added `pip.sh` script for setting dev environment
- Replaced `argsort` with `argmin`/`argmax`
- Refactored redundant preprocessing steps
- Replaced deprecated `scipy.ndimage.filters` module
- Updated conda environment installation steps in `conda.sh` script
- Updated `subspace` vs `subspaces` definition
- Replaced `py.test` with `pytest`
- Updated minimum `black` version to 22.1.0
- Replaced elbow method with `mdl` in multi-dimensional motif tutorial
- Converted `float`/`int` type to `np.float64`/`np.int64`
- Added `dtype` check and fill value to `apply_excl_zone` function
- Added note on anti-correlated subsequences
* documentation
- Updated README and various docstrings
- Added MPdist tutorial draft
- Added annotation vector tutorial
- Added geometric time series chains tutorial draft
- Added top-K motif section to motif discovery tutorial

1.10.2

---------------------------
* bugfixes
- Fixed 501 Allow `max_distance = np.inf` in match function
* features
- Added `atol=1e-8` parameter to match and motifs functions
* tasks
- Fixed typos
- Removed conda download badge, updated PyPI download badge
- Added removal __pycache__ in test clean up phase
* documentation
- Updated pan matrix profile tutorial
- Updated docstring for match and motifs functions

1.10.1

---------------------------
* bugfixes
- Fixed 495 Reduce import time by removing Numba NJIT signatures
* features
- Added multi_distance_profile function
* tasks
- Refactored _query_mstump_profile (see multi_distance_profile)
- Added SVG STUMPY logo
* documentation
- Refactored tutorial

1.10.0

---------------------------
* bugfixes
- Raise TypeError for non np.float64 input arrays
- Fixed NumbaPerformanceWarning backwards compatibility
* features
- Converted all dtypes to np.float64/np.int64
- Added explicit NJIT function signatures
- Added Python 3.9 support
- Added core._idx_to_mp convenience function
- Added Nan/Inf support for motifs and matches functions
* tasks
- Removed "in-tree" build flag for pip
- Added pypi stats query for bigquery
- Set explicit sphinx version for RTD
- Fixed typos
- Replaced direct installation via setup.py
- Added numpydoc
- Removed ipywidgets from requirements
- Removed Azure Pipelines from CI
- Used official OSI name in the license metadata, added license name in classifiers section
- Added Github Citation BIB format
- Fixed broken tutorial links
* documentation
- Fixed multi-dimensional matrix profile description
- Added clear warning in discord section of MSTUMP tutorial
- Added API examples to all docstrings
- Updated docstrings to use numpy.ndarray
- Added interactive threshold example for STIMP
- Updated matplotlib style sheet to use URL for all tutorials

Page 1 of 5

© 2024 Safety CLI Cybersecurity Inc. All Rights Reserved.