Skip to content

Latest commit

 

History

History
120 lines (91 loc) · 6.79 KB

CHANGELOG.md

File metadata and controls

120 lines (91 loc) · 6.79 KB
Raw

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased

Changed

  • Moved making a github release to developer documentation and pointed to it in CONTRIBUTING.md (#78)
  • Added absolute link to README.md and added CONTRIBUTING.md to index.rst in docs (#78)

0.5.0 (2024-06-11)

Changed

  • Moved to ruff formatter instead of black (#70)
  • Do import sorting with ruff instead of isort (#70)
  • Option for "greedy" mode which creates an interval if any data is available within the time bound (#75)
  • map_to_data now tries to infer the time bounds for the data, and uses those to determine which anchor years are valid (#75)
  • Tests for greedy and safe mode, including inferring data frequency (#75)

Fixed

  • Fixed issue with calendar generation when the (rightmost) target period crossed into the new year (#70).
  • Fixed issue with Pandas 2.2 where 'M' changed to 'ME' (#72)
  • Fixed issue where pd.Interval now shows time where it is not wanted (#72). The issue was raised with Pandas ([#57748] pandas-dev/pandas#57748).
  • Added tests for latest version of macos (with M1 chips) (#73)
  • Fixed a bug in map_to_data that lead to anchor years being included when they should not have (#75)

0.4.2 (2024-01-19)

Changed

  • Consistent output type of train-test split as input (#62).
  • Corrected s2spy.time to lilio (#65)

0.4.1 (2023-09-11)

Added

  • Python 3.11 support (#60).

Changed

  • Improved dask integration guide (#59).

Removed

  • Python 3.8 support (#60).

0.4.0 (2023-04-21)

Changed

  • Bokeh >= 3.0.0 is now required for interactive plotting, instead of <= 2.4.3.
    • Do note that dask.distributed still requires bokeh < 3.

Added

  • Lilio's resample now supports use of dask. A tutorial on how to make use of this has been added to the documentation (#52).
  • When resampling xarray data, information such as the lilio version, calendar anchor date, the exact calendar, etc. are now added to the attributes, as well as descriptions and full names for anchor_year, i_interval, and is_target (#49)

0.3.1 (2023-02-15)

Changed

  • lilio.resample will now return xr.DataArray if the input is xr.DataArray.
  • For compatibility with DataArray output, the Dataset/DataArray returned by resample now has the coordinates "left_bound" and "right_bound" instead of a single "intervals" coordinate with the "bounds" dimension.
  • If your input data has an frequency of less than twice the Calendar's smallest interval length, a UserWarning will be raised.
  • If your input data has a frequency less than the Calendar's smallest interval length, a ValueError will be raised.
  • In the output of resample(), the column/coordinate target has been renamed to is_target to avoid clashing with a possibly commonly used name by users.
  • The input data into resample() is now checked for the existance of reserved names, such as "anchor_year" or "i_interval", to avoid overwriting these and cause unpredictable behavior.

0.3.0 (2023-02-08)

First release of Lilio as a split off from s2spy.

Lilio generates calendars to resample timeseries into training and target data for machine learning. It is named after the inventor of the Gregorian Calendar.

Fixed

  • Fixed a bug in Matplotlib calendar visualization related to the anchor date.

Changed

  • The CustomCalendar has been renamed to Calendar.
  • The AdventCalendar, MonthlyCalendar and WeeklyCalendar have been removed as classes. Instead there are functions that generate a standard Calendar.

Added

  • Resampling now supports many methods (e.g. median, min, std) as well as user-defined functions.
  • A "calendar shifter" to create a list of staggered calendars.

Dev changes

  • Lilio makes use of 'hatch' now.
    • Building the package has moved to hatchling
    • Environments and scripts are set up to handle linting and docs building.
  • Ruff is now used as a linter.
  • Notebooks have been moved to the docs folder. Notebooks needs to be cleaned to pass the CI.

0.2.1 (2022-09-02)

Fixed

  • Display of images on ReadtheDocs and PyPi (#97)

0.2.0 (2022-09-01)

Added

  • Improve Sphinx documentation hosted on ReadtheDocs (#32 #70)
  • Support max lags and mark target period methods in time module (#40 #43)
  • Add traintest splitting module for cross-validation (#37)
    • Support sklearn splitters for traintest module (#53)
    • Implement train/test splits iterator (#70)
  • Add Response Guided Dimensionality Reduction (RGDR) module (#68)
    • Implement correlation map function (#49)
    • Implement dbscan for RGDR (#57)
    • Support for multiple lags in RGDR (#85)
  • Update Readme (#95)

Changed

  • Refactor resample methods as functions (#50)
  • Refactor calendars to BaseCalendar class and subclasses (#60)

Removed

  • Python 3.7 support (#65)

0.1.0 (2022-06-28)

Added

  • Time module for an "advent calendar" to handle target and precursor periods.
  • Implemented resampling data to the advent calendar.
  • Example notebooks on how to use the calendar and resampling functionalities.