Feature/restruct unc (#280)

* Remodel uncertainty into data and calc classes

* Replace all dicts by dataframes

* Attach sampling metadata to samples_df

* Add save/load hdf5 methods

* Convert metadata to all strings and make save hdf5 compatible

* Update CalcCostBenefit

* Add default skip values Sobol

* Add test morris sensitivity

* Remove column 'names' for method morris

* Set confidence as input parameter

* Change name of columns in cost ben

* Remove empty dataframe initialization

* Add default sensitivity metadata

* Add unit to unc_data

* Add quantile return period plot

* Correct typos

* Test for unc_data.unit

* Add kwargs and axes handling to plots

* update docstring

* Update plot labels

* Typo

* Reset sensitivity df index

* Add kwargs to plot_sensitvity_map

* Correct typo docstring

* Add default Uncfuncs Hazard and Exp (not finished)

* Add CLIMADA license

* Move UncVar to separate file

* Add unc_var import to __init__

* Add default UncVars

* Change unit name in save hdf5

* Remove old files

* Tutorial header update

* Update docstring

* Add intensity,  mdd, paa in impf_unc (not finished)

* Add docstrings

* Add default impf with x-y uncertainty

* Make load_hdf5 to static from_hdf5

* Update docstring

* Rename modules

* Rename methods and classes

* Update Docstgring

* Update Docstring

* Update docstring

* Correct typo

* Remove Exposures Noise Future

* Update the docstring

* Rename unc_calc to calc

* Fix formating

* Correct metric_names to metrics_names in sensitivity loop

* Update formating

* Set rp locators at end of loop

* Correct sensitvity attribute name typo

* Adjust docstring

* Make inputvar, calc_impact and calc_base tests

* Update tutorial (costben missing)

* Correct maximumm typo

* Check for dataframe sparcity

* Add Default input var description

* Change calc to calc_base

* Set input filename to Path

* Change size requirements for sparse dataframes

* Constant n for haz event subsampling

* Set hazard sub-sample size constant

* Return UncOutput in make_sample

* Return unc_output in calc_impact.uncertainty

* Return unc_output in cost_ben.uncertainty

* Update tutorial

* Return sens_output in calc.sensitivity

* Update docstring

* Return sens_output in calc.sensitivity

* Correct docstrin typo

* Add description of UncOutput

* Move Save/Load to OutputTest

* Remove sparse dataframe

The SparseArrays cannot be saved to hdf5.

* calc_base.Calc: rename input_var property to input_vars

* unc_output: UncOutput constructor takes samples_df as single argument

* Update climada/engine/uncertainty_quantification/__init__.py

Co-authored-by: emanuel-schmid <[email protected]>

* Use explicitly import_module from importlib

Co-authored-by: emanuel-schmid <[email protected]>

* Remove redundant param_labels code

Co-authored-by: emanuel-schmid <[email protected]>

* fixes

* PR review suggestions

* UncOutput.plot_rp_uncertainty: checking a data frame for being None must be explicit

* organize imports

* update docstrings

* uncertainty module has changed name to uncertainty_quantification

* Change default exposure noise to noise strength

* Add docstring to new output classes (impact/costben)

* add newline at end of file

* Add seed to noise helper functions

* Correct typo

* Set maximum x-points for discrete inputvar plot

* Add test for discrete inpuvar

* Add test to sve samples/uncertainty/sensitivity only.

* Update UncOutput Tutorial

* Check if sensitivity attributes in hdf5 file before reading

* Add docstring on seed

* Set logger to error in tutorial

* Add check input param distribution for duplictes

* Add haz_type and fun_id to entity

* Add seed sampling to ent fut noise

* calc_base has no business in UncOutput's administration

* refactor sensititvity method (break and disentangle)
truncate metric names (cut the _unc_df and _sens_df tails)
applly [gs]et_(unc|sens)_df where applicable

* adapt to new metric names
i.e., without the trailing /_(unc|sens)_df/

* PEP8, white space cosmetics

* Add tests Cost Benefit

* Update tutorial uncertainty

* Check for None axis instead of not axis

* Add suffix 'hdf5' automatically to from_hdf5

* Check for None axis instead of not axis

* Add exception for empty  uncertainty plots

* Add original values to uncertainty plot

* Add comment on dead sparse df code

* Make tigh layout samples plot

* Add test all plots

* Plot non-log value in legends

* Remove pre-restructuring tests

* Add get samples_df method

* Correct discount rate typo in ent docstring

* Add default entity input var examples

* Add cost benefit uncertainty tutorial

* Check for identical distributions only

* Test for distributions

* Remove print from debugging

* Add impf dict to inputvar.impf

* Add impf inputvar id dict to tutorial

* Add haz_id_dict to entity

* Add test files through API

* Update fetch API files to target_dir api

* Make all si 0 to appear as none in sensitivity map plot

* Wrap codeline nicer

* Add method to get largest si

* Cast index to int

Co-authored-by: Chahan Kropf <[email protected]>
Co-authored-by: emanuel-schmid <[email protected]>
Co-authored-by: emanuel-schmid <[email protected]>

Author: 3 people

climada/engine/impact.py

@@ -226,7 +226,7 @@ def calc(self, exposures, impact_funcs, hazard, save_mat=False):
             if not exp_step:
                 raise ValueError('Increase max_matrix_size configuration parameter to > %s'
                                  % str(num_events))
-            # separte in chunks
+            # separate in chunks
             chk = -1
             for chk in range(int(exp_iimp.size / exp_step)):
                 self._exp_impact(