All notable changes to this project will be documented in this file.
The format is inspired from Keep a Changelog and the versioning aim to respect Semantic Versioning.
Here is a template for new release sections
## [_._._] - 20XX-MM-DD
### Added
-
### Changed
-
### Removed
-
- converted units to process output
- fix numpy error due to new numpy version
- unit conversions for all units detected in reference dataset
- "None" values in possible FK column are overwritten by FK mapping
- versioning for "srd_point_draft" and "srd_range_draft"
- if no type is given in type column, table process name is used
- in case of bandwidth values, first value is used for process
- SPARQL queries
- multiple versions error when merging FKs into process
- merging parameters of empty dataframe
- reading parameters of process if no parameters are set for a process in structure
- helper set processes to process list
- unit conversion to scalar and timeseries data
- structure information to process
- merging of parameters if cell contains np.nan
- structure is read from Excel file
- function to get all processes from collection
- refactored data collection from artifact
- collection function to get artifact by name
- removed legacy sqlalchemy engine
- latest version for versions > v9
- databus endpoint, SPARQLing and related tests
- foreign keys are given in datasets instead of links
- timeseries refactoring for multiple periods
- typing error for python versions <3.10
- timeseries are turned into dataframes with timeindex and multiindex columns containing name and regions
- allow multiple (sub-)processes per table
- collection metadata can contain multiple names and subjects per artifact
- links in
get_process
function
- introduced Adapter class to handle collection, structure and links in a centralized way
- by default, process and parameter names are not derived from annotations, but from name in metadata and column names
- "additional parameters" are renamed to "links" and are read-in from CSV
- structure containing modex example in tests
- default parent directory for collections and structures dirs to current working dir
- mypy linting
- models for ES structure
- process dateframes are merged and returned as scalars and timeseries
- regions are unpacked and merged in process dataframes
- versioned collection metadata
- reading of annotation
- checks for annotations of collection
- collection metadata holds data type (Note: Collections have to be reloaded)
- data import via frictionless (converts timeseries to list-of-floats while loading)