v1.0.4

Fixes:

• built-in adaptors calling ndim on DataArray (#317)

v1.0.2 Minor update release

Minor updates:

• Update smif app readme
• Bump babel and webpack major versions, update other npm packages
• Validate self-dependencies (between timesteps is okay)
• Add validation methods for narratives
• Make illegal parameters visible in narrative configuration
• Pass path to binary filestore, extract method to parent class
• Allow adaptors to be directly included in a system of systems model
• Provide useful error message when there are missing data in a data array
• Add profiling to some places in the program, provide summary at end of modelrun
• Don't read dimension elements through API

Fixes:

• Fix and test reading from timeseries, including zero-d case
• Fix API calling old store methods
• Pin libgcc as possible cause of shared library import errors
• Fix up test_validate to use conftest configs
• Fix react-icons imports, drop reactstrap
• Silence mocha deprecation warning.
• Ensure smif npm package is private
• Fix update_model method store
• Adopt fix for DataFrame.to_dict('records') from future pandas

v1.0

Functionality:

• GUI improved usability

  • Forms now ask users to discard or save changes
  • Configuration lists can be sorted and filtered
  • Single click navigation between linked configurations
  • First steps of input validation (in SosModel configurations)

• Define model data (inputs/parameters/outputs) using arbitrary dimensions (may be spatial,
  temporal, categorical)

  • Dimension conversions can be performed by an Adaptor, represented as another
    SectorModel within a SosModel

• Data layer refactor to enable various Store implementations, separately for configuration
  data, metadata and input/parameter/results and interventions/decisions/state data.

  • DataArray and Spec handle input/parameter/results data and metadata
  • Groundwork for a PostgreSQL DbConfigStore implementation

• Separation of SosModel construction and configuration from ModelRun running:
  introduce a JobScheduler that runs directed graphs of simulation jobs, connected by
  dependency edges.

  • Initial JobScheduler is purely serial
  • Remove ModelSet, removing the capability to handle within-timestep dependency loops
  • Introduce explicit between-timestep dependencies (including model self-dependency)

v0.8.2

Fixes

• Avoid blank lines in csv output
• Avoid repeated reads of region names and project config
• Pin ruamel.yaml version
• Handle empty strategy file with empty list of strategies
• Fix unresponsive dependency selector
• Add/edit/remove facets and scenarios
• Run models in subprocess (now also on Windows)

v0.7.7: Various bug fixes

Fixed bug where valid ids of scenarios failed validation due to type mismatch

v0.7.6: Merge pull request #164 from nismod/feature/run_app_in_nismod

Feature/run app in nismod