CHANGELOG.md

Changelog

All issue numbers are relative to https://github.com/pywr/pywr/issues unless otherwise stated.

v1.25.0 (2024-06-19)

Features

• Add support for defining net or gross loss in LossLink. (#1124) (ed8eb8b), closes #1124

Other

• Pin numpy<2 for now. (#1130) (2672645), closes #1130
• Use delvewheel to repair Windows wheels. (#1125) (691172c), closes #1125 #1123

Documentation

• Add OtherModelXXX parameters to API docs. (#1126) (bdb257d), closes #1126

v1.24.0 (2024-03-20)

Features

• Cache file hashes to avoid recomputation when the same file is referenced multiple times (#1118) (4afc270)

Fixes

• update warnings check in orphaned components test (#1119) (582a764)

Other

• black formatting updates (#1119) (7d1b60c)

v1.23.0 (2023-11-13)

Features

• Support parameter loss_factor in LossLink. (#1113) (cb583aa), closes #1113 #1091

v1.22.1 (2023-10-20)

Fixes

• Fix issue with dataframe scenario indexing (#1111) (e6818a8), closes #1111 #1110

v1.22.0 (2023-10-06)

Other

• Updates for Python 3.12 in CI (#1107) (a0924ea), closes #1107
• Support Cython 3.x (#1104) (1981de1), closes #1104
• Add Zulip link to README. (#1108) (651e949), closes #1108

v1.21.0 (2023-06-30)

Features

• Pass kwargs to model load via optimisation wrappers (#1096) (3557e4b)
• Add metadata to TablesRecorder arrays. (#1083) (4dac394)

Fixes

• Overload get_xxx_flow for PiecewiseLink. (#1088) (51dcd57), closes #1087
• Add pywr.parameter imports from control_curves module. (#1084) (94bcdad)

Other

• Support pandas v2.x (#1089) (68ee7cc)
• Use ubuntu-latest image for building wheels. (#1090) (cce65e3)

v1.20.1

Bug Fixes

• Fix bug loading timestep_offset in DataFrameParameter from JSON. (#1078)

v1.20.0

New Features

• Allow storage max volume to be set as a constant parameter without specifying both initial volume and initial volume pc. (#1071)
• Add timestep offset to DataFrameParameter and TablesArrayParameter. (#1073)
• Update TablesRecorder to save scenario slice data. (#1072)
• Add RollingMeanFlowNodeParameter and deprecate RollingMeanFlowNodeRecorder. (#1074)

Bug Fixes

• Fix Cython typing warning in MonthlyProfile. (#1075)

Miscellaneous

• Release wheels for Python 3.11.

v1.19.0

New Features

• Add WeightedAverageProfileParameter. (#1066)
• Update to AnnualCountIndexThresholdRecorder to include a range of days to record. (#1061)

Bug Fixes

• Fix __init__ argument handling in BisectionSearchModel. (#1067)

Miscellaneous

• Remove matplotlib, jinja2 and ipython from core dependencies. (#1040)
• Fix LicenceParameter example in documentation. (#1068)
• Remove outdated documentation on constant factors in AggregatedNode. (#1060)

v1.18.0

New Features

• Hash checking is now not case-sensitive. (#1046)
• Add MonthlyVirtualStorage node that resets after a given number of months.
• Add StorageParameter that provides the current volume from a storage node.
• Add support for running multiple Pywr models in an interleaved fashion using a new MultiModel class and associated parameters.

Bug Fixes

• Update load method of several parameters (ControlCurveInterpolatedParameter, ControlCurveParameter, ArrayIndexedScenarioMonthlyFactorsParameter, several interpolated parameters) to ensure all data is loaded. (#1048)

v1.17.2

Bug Fixes

• Fix a bug where the solver specified using a keyword to Model.load would be ignored. (#1044)

v1.17.1

Miscellaneous

• Release wheels for Python 3.9 (Windows) and Python 3.10 (Linux and Windows).

v1.17.0

New Features

• Networks with missing connections will now more reliably raise ModelStructureError. (#1025)
• Improve GLPK API (#1021):
  • GLPK errors are now handled by default by raising GLPKInternalError.
  • NaN checks are performed by default.
  • Added runtime option to use "unsafe" API with no error or NaN handling (see new documentation).
  • Added documentation describing error handling changes.
• Improve handling of NaN values in Storage.current_pc (i.e. when max_volume is zero).
  • Added a new method Storage.get_current_pc to guarantee a finite value.
• Added min_output to RectifierParameter

Miscellaneous

• Documentation improvements:
  • RiverSplitWithGauge docstring. (#1030)
  • Added section describing how to use Dataframe checksums. (#1028)
  • Added ControlCurveParameter to API docs. (#1027)

v1.16.1

Bug Fixes

• Fix internal node names for PiecewiseLink. (#1024)

v1.16.0

New Features

• Supply initial_volume_pc in VirtualStorage and its subclasses. (#1015)
• Add residual_days option to UniformDrawdownProfileParameter. (#1013)

Bug Fixes

• Fix a JSON formatting error in the documentation. (#1009)
• Removed a circular import in pywr.recorders.calibration. (#1012)

Miscellaneous

• Removed some older folders related to Travis. (#1016)
• Re-formatted the Python code using the Black formatter and enforced this for future modifications. (#1010)

v1.15.3

Bug Fixes

• Fix bug where constraints are not applied between VirtualStorage and some complex node types (e.g PiecewiseLink and AggregatedNode). Previously models would run and silently ignore constraints that users might reasonably expect to be applied. Now these models will error until a proper fix is implemented. See issue #1001 to see details and track progress. (#1002)

v1.15.2

Bug Fixes

• Fix bug with AggregatedParameter.remove() and AggregatedIndexParameter.remove() methods using the wrong method internally. (#997)

v1.15.1

Bug Fixes

• Fix AnnualTotalFlowRecorder's accounting of flow when a timestep ends in the following year. (#992)
• Fix divide by zero errors in NumpyArrayNodeSuppliedRatioRecorder and NumpyArrayNodeCurtailmentRatioRecorder when the max flow of the parameter they monitor returned zero. (#993)
• Fix the load method of FlowDurationCurveDeviationRecorder to handle scenario data correctly. (#989)

v1.15.0

New Features

• Performance improvements for the GLPK solvers. Several new options have been added to the GLPK solvers that reduce the number of linear programme updates required each time-step. These options are currently off by default, but can be enabled by either passing keyword arguments to the solver instances or via environment variables. (#912, #983, #985)
  • Update constant and fixed flows only once during reset (PYWR_SOLVER_GLPK_FIXED_FLOWS_ONCE=True)
  • Update constant and fixed costs only once during reset (PYWR_SOLVER_GLPK_FIXED_COSTS_ONCE=True)
  • Update constant and fixed aggregated node factors only once during reset (PYWR_SOLVER_GLPK_FIXED_FACTORS_ONCE=True)
• Improve memory usage in DataFrameParameter when running a subset of the defined scenarios. The parameter now only retains the data for the subset of scenarios required for the simulation instead of all scenarios. (#981)
• Update MonthlyProfileParameter and RbfProfileParameter to allow variable lower and upper bounds to be defined. (#986)

Bug Fixes

• Fix use of deprecated np.int and np.float types. (#982)

v1.14.0

New Features

• Allow target flow-duration curves to be supplied in an external file for FlowDurationDeviationRecorder. (#977)
• Allow only one of upper or lower targets to be specified in FlowDurationDeviationRecorder. (#977)

Bug Fixes

• Fix bug where event recorders were not registered and could not be loaded from JSON. (#976)

v1.13.1

Bug Fixes

• Allow loading normalised KDE and NumpyArrayNormalisedStorageRecorder recorder from JSON. (#974)

v1.13.0

New Features

• Added LossLink node. (#960)
• Added MultiThresholdIndexParameter. (#969)
• Added two recorders that fit a KDE to time-series of storage volumes. (#970, #971)
• Added an optional tags dictionary to Component for storing user defined metadata. (#968)

Bug Fixes

• Support loading "area" and "level" attributes as parameters correctly. (#972)

Miscellaneous

• Added CI support for Python-3.9 on manylinux. (#954)

v1.12.0

New Features

• Add additional "activation function" parameters: BinaryStepParameter, RectifierParameter and LogisticParameter. (#965)

v1.11.0

New Features

• Refactor node loading into two stages to solve circular loading problems. (#945)
• Add additional get_all_xxx methods to core nodes to allow fetching min_flow, max_flow and cost values for all scenarios simultaneously. (#958)

Miscellaneous

• Switch to using Openpyxl for reading and writing Excel files (see changes to Pandas v1.2). (#959)
• Dropped support for Python 3.6 following lack of support in upstream libraries (NumPy and Pandas). (#959)

v1.10.0

New Features

• Add optional exclude_months to AnnualCountIndexThresholdRecorder. (#950)
• Update load methods of interpolation parameters so that they accept interp_kwargs. (#943)
• Allow InterpolateVolumeParameter to use external data. (#926)
• Added NumpyArrayNodeCostRecorder to recorder a timeseries of a node's cost attribute. (#932)
• Added to_dataframe() method to AnnualCountIndexThresholdRecorder and AnnualTotalFlowRecorder. (#939)

Bug Fixes

• Fixed a bug with RbfProfileParameter where it did not pass rbf_kwargs to the RBF functions. (#946)
• Fixes to save and draw graph functions to improve handling virtual and child nodes. (#938)

Documentation

• Updated installation docs. (#931)
• Add missing parameters (FlowParameter and DiscountFactorParameter) to API docs. (#934)

Miscellaneous

• Migrate continuous integration system from Travis and Appveyor to Github Actions. (#948, #951, #952, #953)
• Removed deprecated PiecewiseLinearControlCurve. (#947)
• Apply some PEP8 compliance fixes. (#929)
• Remove use of basestring which was only needed for Python 2. (#928)

v1.9.0

New Features

• Added BisectionSearchModel that performs a bisectional search on a single parameter instead of a standard simulation. (#915)
• Allow AggregatedNode factors to be time-varying using Parameters. (#919)
• Added RollingVirtualStorage node intended for modelling rolling licenses. (#891)
• Added SeasonalVirtualStorage node intended for modelling licenses that apply for limited periods. (#923)

Bug Fixes

• Ensure RollingMeanFlowNodeRecorder's internal memory pointer is reset correctly. (#893)
• Fix a bug where AggregatedNode would warn about small factors with any negative value. (#921)
• Fixed AggreagtedNode initial volume being incorrectly calculated when its dependent nodes used a proportional initial volume. (#922)

Miscellaneous

• Added NullSolver that performs no allocation and is intended for debugging purposes. (#924)
• Added a small tolerance for equality checking of fixed bounds in the GLPK solvers. (#925)

v1.8.0

New Features

• A change to the behaviour of Storage nodes with parameters for max_volume. Such nodes must now have both initial_volume and initial_volume_pc specified. This allows using arbitrary parameters for max_volume provided the initial condition is explicitly defined in absolute and percentage terms. (#690)
• Added DelayNode and FlowDelayParameter to allow emulation of time-of-travel. (#904)
• Added DiscountFactorParameter. (#901)
• Added support for optimising days of the year in RbfProfileParameter. (#908)
• Added NumpyArrayDailyProfileParameterRecorder for recording calculated annual profiles, and refactored the internal calculations around day of the year. Timestep now includes properties for dayofyear_index, week_index and is_leap_year. (#903)
• Added error logging in Model's load, setup, reset and finish methods. Handling of unknown component types now raises a custom TypeNotFoundError. (#896)

Miscellaneous

• Improvements to API documentation. (#905)

v1.7.2

Miscellaneous

• Release to fix packaging mistake in v1.7.1.

v1.7.1

Bug Fixes

• Fixed a bug when using draw_graph with a dictionary. (#899)
• Fixed a bug when giving non-float values to RbfProfileParameter. (#897)
• Improved handling of isolated nodes in draw_graph when using JSON data. (#895)

v1.7.0

New Features

• Improve Jupyter Notebook graphing functionality. (#868, #885)
• Added OffsetParameter. (#874)
• Added PywrRandomGenerator for use with the Platypus optimisation library. (#867, #892)
• Added RbfProfileParameter. (#873)
• Changed the signature of recorder aggregation functions to better catch exceptions. (#879)

Bug Fixes

• Removed unreachable code when CSV dataframes. (#880)
• Fixed incorrect parsing of position keyword in several nodes. (#884)

Miscellaneous

• Added IPython to install dependencies. (#870)
• Removed optimisation wrapper for inspyred. (#878)
• Exposed run statistics to optimisation wrapper. (#877)
• Added optional dependencies for docs and dev installs. (#882)

Documentation

• Added threshold parameters to API documentation. (#881)
• Corrected MeanParameterRecorder's docstring. (#772)
• Improved docstrings and made consistent argument names for InterpolatedVolumeParameter and InterpolatedFlowParameter. (#890)

v1.6.0

New Features

• AnnualTotalFlowRecorder now accepts an optional list of factors to scale the flow by. (#837)
• NumpyArrayNodeRecorder now accepts an optional factor (default=1.0) to scale the flow by. (#838, #840)
• Added UniformDrawdownProfileParameter (#836)
• Added ControlCurvePiecewiseInterpolatedParameter as a more general replacement for PiecewiseLinearControlCurve. (#857)
• Added 'count_nonzero' as an aggregation function for recorders. (#866)

Bug Fixes

• Fix bug draw_graph modifying model data when a data dict is given. (#832)
• Fix the __init__ method of BreakLink. (#850)
• Fix reset of AbstractNode._prev_flow. (#855)
• Fix a bug calculating of AggregatedStorage's initial volume in multiple scenarios. (#854)
• Fix resetting of AnnualVirtualStorage volume to maximum volume instead of initial volume. (#860)
• Fix cdef type issue in some control curve parameters allowing use with any AbstractStorage (instead of just Storage). (#861)
• Fix registering of ArrayIndexedScenarioParameter (#863)

Miscellaneous

• Fixed documentation building on tags. (#831)
• Updated notebook graph drawing to use d3.v5 and removed IE specific code. (#834)
• Add reference to published academic paper. (#846)
• PiecewiseLinearControlCurve has been marked as deprecated and will be removed in a future version. (#857)
• Added examples from the recently published paper. (#852)

v1.5.0

New Features

• Added ScenarioDailyProfileParameter and ScenarioWeeklyProfileParameter to provide different profiles per scenario. (#802)
• Added TimestepCountIndexParameterRecorder, AnnualCountIndexThresholdRecorder and AnnualTotalFlowRecorder (#784)
• Added daily interpolation support to MonthlyProfileParameter. (#807)
• Added __contains__ method to NamedIterator and NodeIterator (#813)

Bug fixes

• Fix resetting progress of ProgressRecorder (#816)
• Fix for draw_graph issue error when loading model object that has schematic positions (#821)

Miscellaneous

• Removed FutureWarning and UnicodeWarning warning filters (#803)
• Refactored setup.py to improve build time dependency handling and specifying build arguments (#811)
• Fix deprecated use of .labels in the tests (#814)
• Fix test warning when incrementing timestep by integers (#815)
• Fix duplicated test function names (#818)
• Support Python 3.8 (#796)
• Refactored the GLPK solvers in to a single extension module. (#822)
• Add .pxd files to the Pywr's package data so they are distributed. (#824)

Documentation

• Fixed some warning and deployment issues with the documentation (#771)
• Add missing new line from code-block sections in json.rst (#817)

v1.4.0

New Features

• Added support time-steps based on Pandas offsets (#675)
• Added InterpolatedFlowParameter (#740)
• Added support for percentile and percentileofscore aggregation functions (#777)
• Added PiecewiseIntegralParameter (#772)
• Added support for including references to Python modules in JSON format (#765)
• Added CurrentYearThresholdParameter and CurrentOrdinalDayThresholdParameter parameters (#789)

Bug fixes

• Ensure comment key doesn't get passed to Pandas read_xxx functions (#788)

Documentation

• Added some docs for AggregatedNode (#756)

v1.3.0

New Features

• Allow use of parameters as values in ControlCurveInterpolatedParameter (#750)
• Added ScenarioWrapper parameter (#763)

Bug fixes

• Fix loading PiecewiseLink with parameters from JSON (#749)
• Fixed a bug with CSVRecorder not saving volumes correctly (#767)

Miscellaneous

• Removed six as dependency (#745)
• Removed pywr.__git_hash__ (#752)
• Removed Blender node (#757)

v1.2.0

New Features

• Support for embedding dataframe data directly in to JSON (#700)
• Added NumpyArrayAreaRecorder and refactored storage recorders (#684)
• Added getter for Recorder.agg_func (#719)
• Add DivisionParameter and tests (#722)
• Add FlowParameter for tracking yesterday's flow (#724)
• Add InterpolatedQuadratureParameter (#714)
• Add new array deficit recorders (#729):
  • NumpyArrayNodeDeficitRecorder - timeseries of deficit.
  • NumpyArrayNodeSuppliedRatioRecorder - timeseries of supply / demand
  • NumpyArrayNodeCurtailmentRatioRecorder - timeseries of 1 - supply / demand

Bug fixes

• Fix a bug with hydropower parameters & recorders not applying efficiency factor (#737)
• Refactor of the code used to load named parameters and recorders to use shared functions (as they are both components) (#720)
• Fix a bug with AggregatedRecorder not returning the instance on load (#723)
• Use flow instead of max_flow in two_reservoirs test and example (#721)

Documentation

• Added API documentation for nodes (#668)
• Fix PiecewiseLink docstring's ASCII diagram (#668)

Miscellaneous

• Clean-up various warnings in tests (#695)
• Removed conda-recipe (Pywr is now in conda-forge) (#692)
• Added codecov based coverage reporting (#705)
• Updated test builds to use manylinux2010 to build wheels (#710)
• Updated installation instructions to reflect wheels in Pypi and conda-forge installation.

v1.1.0

New features

• New "edge based" GLPK solver. (#672)
• Improved FlowDurationDeviationRecorder with JSON support and bug fixes when no scenario is given. (#677)

Bug fixes

• Replace deprecated calls time time.clock with time.perf_counter. (#683)
• Type optimisation in AggregatedIndexParameter. (#662)

Documentation

• Updated documentation: control curves, extending Pywr. (#652)

Miscellaneous

• Variable renames and clean up for GLPK path solver. (#672)
• Spport for Python 3.7. (#662)
• Updated url in setup.py metadata to GitHub project (#661)
• Additional cython compiler directives (#645)

v1.0.0

New features

• Added ratchet support to threshold parameters. (#655)
• Added ConstantScenarioIndexParameter. (#654)
• Added support for Platypus and Pygmo optimisation wrappers. Involved a refactor of the existing optimisation support. (#610)
• Added HydropowerTargetParameter to specify a flow target from a hydropower target. (#631)
• Renamed HydroPowerRecorder to HydropowerRecorder (#631)
• Better warning/error messages in TablesArrayParameter (#629)
• Allow solver to be defined by the environment variable PYWR_SOLVER. (#619)
• Added flow weights to AggregatedNode. (#603)
• Added additional labeling functionality to notebook graphing functions. (#612)
• New and improved variable API for Parameters. (#601, #258, #625)

Bug fixes

• Fix bug setting the area property of Storage nodes. (#657)
• Fixed bug with finally clause in the optimisation wrapper. (#649)
• Fix a bug in AnnualHarmonicSeriesParameter related to updating the amplitudes and phases values with set_double_variables (#622)

Miscellaneous

• Refactored several recorders and unified the use of temporal aggregation. This deprecated several keyword arguments in some existing Recorders. See the PR for details. (#635)
• Removed deprecated (Monthly|Daily)ProfileControlCurve. (#640)
• Dropped support for Python 2 and <3.6. Pywr is no longer tested against Python versions earlier than 3.6. (#623)
• Use new networkx.nodes_with_selfloops function. (#628)
• AbstractProfileControlCurveParameter, MonthlyProfileControlCurveParameter and DailyProfileControlCurveParameter have been removed after deprecation. (#231, #640)
• Improved documentation. (#616, #627)

v0.5.1

Miscellaneous

• Fixes to the source distribution (sdist) and inclusions of MANIFEST.in
• Changes to the build systems on travis to enable deploy to Anaconda and Pypi.

v0.5

New features

• Added build and testing for OS X via travis. (#588)
• Added data consumption warnings to DataFrameParamter and TablesArrayParameter. (#562)
• Added PiecewiseLinearControlCurve as a new cython parameter.
• Pywr now emits some logging calls during a model run.
• Improved the event handling code to allow tracking of a Parameter's value during an event.
• Added support for initialising storage volume by percentage. Can be set in json through the "initial_volume_pc" property.
• Added GZ2 and BZ2 compression support to CSVRecorder.
• Added JSON support for license parameters. (#544)
• Added a sense check to TablesArrayParameter having non-finite values.
• Added scale and offset to ConstantParameter.
• Added JSON support for license parameters. (#544)
• Added WeeklyProfileParameter (#537)
• Added InterpolationParameter. Closes #478. (#535)
• Added surface area property to Storage (#525)
• Added optional checksum when loading DataFrameParameter and TablesArrayParameter using hashlib.
• Added ProgressRecorder and JupyterProgressRecorder (#520)
• The threshold in *ThresholdParameter can now be a Parameter. (#517)
• Added HydroPowerRecorder and TotalHydroEnergyRecorder for hydropower studies (#584)

Bug fixes

• Explicitly set the frequency during dataframe resampling (#563)
• IndexedArrayParameter also accepts 'parameters' in JSON. Closes #538. (#539)
• Recursive deletion of child nodes when deleting compound nodes. (#527)
• Compatibility with NetworkX 2.x (#529)
• Changed GLPK log level to remove printing of superfluous messages to stdout. (#523)
• Fixed loading parameters for PiecewiseLink in JSON. (#519)

Deprecated features

• AbstractProfileControlCurveParameter marked for deprecation.

Miscellaneous

• Improved the online documentation including the API reference.
• Added a hydropower example.
• General tweaks and corrections to class docstrings.
• Updated conda build recipe to use the conda-forge lp_solve package.
• Updated the conda build recipe to use MSVC features.