Merge pull request #182 from MichaelTiemannOSC/Infer-Scope2

Merging in preparation for NZAOA discussions.

ITR/__init__.py

@@ -2,6 +2,7 @@
 This package helps companies and financial institutions to assess the temperature alignment of investment and lending
 portfolios.
 """
+import pandas as pd
 from .data import osc_units
 from . import data
 from . import utils
@@ -34,3 +35,11 @@ def uarray(nom_vals, std_devs):
 
     def umean(unquantified_data):
         return mean(unquantified_data)
+
+def recombine_nom_and_std(nom: pd.Series, std: pd.Series) -> pd.Series:
+    assert HAS_UNCERTAINTIES
+    if std.sum()==0:
+        return nom
+    assert not std.isna().any()
+    return pd.Series(data=uarray(nom.values, std.values), index=nom.index, name=nom.name)
+

ITR/configs.py

@@ -102,7 +102,7 @@ class SectorsConfig:
     STEEL = "Steel"
     ALUMINUM = "Aluminum"
     ENERGY = "Energy"
-    # OIL_AND_GAS = "Oil & Gas"
+    OIL_AND_GAS = "Oil & Gas"
     COAL = "Coal"
     OIL = "Oil"
     GAS = "Gas"
@@ -133,7 +133,7 @@ def get_configured_sectors(cls) -> List[str]:
         return [SectorsConfig.POWER_UTILITY, SectorsConfig.GAS_UTILITY, SectorsConfig.UTILITY,
                 SectorsConfig.STEEL, SectorsConfig.ALUMINUM,
                 SectorsConfig.ENERGY, SectorsConfig.COAL, SectorsConfig.OIL, SectorsConfig.GAS,
-                # SectorsConfig.OIL_AND_GAS,
+                SectorsConfig.OIL_AND_GAS,
                 SectorsConfig.AUTOMOBILE, SectorsConfig.TRUCKING,
                 SectorsConfig.CEMENT,
                 SectorsConfig.BUILDINGS_CONSTRUCTION, SectorsConfig.BUILDINGS_RESIDENTIAL, SectorsConfig.BUILDINGS_COMMERCIAL,

ITR/data/__init__.py

@@ -41,8 +41,19 @@
     'RM':'MYR',
 }
 
+def escape_currency_symbols(text):
+    if '$' in text:
+        escaped_text = re.escape(text)
+    else:
+        escaped_text = text
+    if escaped_text[0].isalpha():
+        if escaped_text[-1].isalpha():
+            return r'\b' + escaped_text + r'\b'
+        return r'\b' + escaped_text
+    return escaped_text
+
 currency_keep_regexp = re.compile(fr"({'|'.join([cur_abbrev for cur_abbrev in currency_dict.values()])})")
-currency_split_regexp = re.compile(fr"(\$|US\$|{'|'.join([re.escape(currency_symbol) for currency_symbol in currency_dict])})")
+currency_split_regexp = re.compile(fr"(\$|\bUS\$|{'|'.join([escape_currency_symbols(currency_symbol) for currency_symbol in currency_dict])})")
 
 def translate_currency_symbols_1(text):
     split_text = re.split(currency_split_regexp, text)
@@ -91,4 +102,3 @@ def translate_currency_symbols(text):
 # FIXME: delay loading of pint_pandas until after we've initialized ourselves
 from pint_pandas import PintType
 PintType.ureg = ureg
-

ITR/data/data_warehouse.py

@@ -94,7 +94,7 @@ def _restore_historic_data(self):
             c.projected_targets = orig_data['projected_targets']
 
     def update_benchmarks(self, benchmark_projected_production: ProductionBenchmarkDataProvider,
-                         benchmarks_projected_ei: IntensityBenchmarkDataProvider):
+                          benchmarks_projected_ei: IntensityBenchmarkDataProvider):
         """
         Update the benchmark data used in this instance of the DataWarehouse.  If there is no change, do nothing.
         """
@@ -122,7 +122,6 @@ def update_benchmarks(self, benchmark_projected_production: ProductionBenchmarkD
         # Target projections rely both on Production benchmark data and S3 estimated data
         # Production-centric benchmarks shift S3 data after trajectory and targets have been projected
         if new_production_bm:
-            # Note that _validate_projected_trajectories overwrites fields of companies, so no need to use return value
             logger.info(f"new_production_bm calculating trajectories for {len(self.company_data._companies)} companies (times {len(EScope.get_scopes())} scopes times {self.company_data.projection_controls.TARGET_YEAR-self.company_data.projection_controls.BASE_YEAR} years)")
             self.company_data._validate_projected_trajectories(self.company_data._companies, self.benchmarks_projected_ei._EI_df)
 
@@ -195,7 +194,8 @@ def update_benchmarks(self, benchmark_projected_production: ProductionBenchmarkD
                 # The alignment calculation: Company Scope-Sector emissions = Total Company Scope emissions * (BM Scope Sector / SUM(All Scope Sectors of Company))
                 aligned_em = [(sector, [(scope, list(map(lambda em: (em[0], em[1] * sector_em_df.loc[(sector, scope)].squeeze() / em_tot.loc[scope].squeeze()),
                                                          [(em.year, em.value) for em in historic_dict['+'.join([orig_id, sector])].emissions[scope.name]])))
-                                        for scope in em_tot.index])
+                                        for scope in em_tot.index
+                                        if em_tot.loc[scope].squeeze().m != 0.0])
                               for sector in sectors]
 
                 # Having done all scopes and sectors for this company above, replace historic Em and EI data below
@@ -223,7 +223,7 @@ def update_benchmarks(self, benchmark_projected_production: ProductionBenchmarkD
         # Changes to production benchmark requires re-calculating targets (which are production-dependent)
         if new_production_bm:
             logger.info(f"projecting targets for {len(self.company_data._companies)} companies (times {len(EScope.get_scopes())} scopes times {self.company_data.projection_controls.TARGET_YEAR-self.company_data.projection_controls.BASE_YEAR} years)")
-            self.company_data._calculate_target_projections(benchmark_projected_production)
+            self.company_data._calculate_target_projections(benchmark_projected_production, benchmarks_projected_ei)
 
         # If our benchmark is production-centric, migrate S3 data (including estimated S3 data) into S1S2
         # If we shift before we project, then S3 targets will not be projected correctly.
@@ -611,9 +611,16 @@ def _get_cumulative_emissions(self, projected_ei: pd.DataFrame, projected_produc
             scale_factor = projected_ei.iloc[:, 0].map(lambda ei: ei.u).combine(projected_production.iloc[:, 0].map(lambda pp: pp.u),
                                                                                 lambda ei_u, pp_u: Q_(1.0, (ei_u * pp_u)).to('t CO2e').m)
             projected_t_CO2e = projected_ei.applymap(lambda x: x.m).mul(projected_production.applymap(lambda x: x.m)).mul(scale_factor, axis=0)
-        # At the last instance, convert our magnitudes to t CO2e everywhere
-        cumulative_emissions = projected_t_CO2e.cumsum(axis=1).astype('pint[t CO2e]')
-        return cumulative_emissions
+        if ITR.HAS_UNCERTAINTIES:
+            # Normalize uncertain NaNs...FIXME: should we instead allow and accept nan+/-nan?
+            projected_t_CO2e[projected_t_CO2e.applymap(lambda x: isinstance(x, ITR.UFloat) and ITR.isnan(x))] = ITR.ufloat(np.nan, 0.0)
+            # Sum both the nominal and std_dev values, because these series are completely correlated
+            nom_t_CO2e = projected_t_CO2e.apply(lambda x: ITR.nominal_values(x)).cumsum(axis=1)
+            err_t_CO2e = projected_t_CO2e.apply(lambda x: ITR.std_devs(x)).cumsum(axis=1)
+            cumulative_emissions = nom_t_CO2e.combine(err_t_CO2e, ITR.recombine_nom_and_std)
+        else:
+            cumulative_emissions = projected_t_CO2e.cumsum(axis=1)
+        return cumulative_emissions.astype('pint[t CO2e]')
 
     def _get_exceedance_year(self, df_subject: pd.DataFrame, df_budget: pd.DataFrame, budget_year: int=None) -> pd.Series:
         """

ITR/data/osc_units.py

@@ -20,6 +20,9 @@
 PA_ = PintArray
 
 ureg.define("CO2e = CO2 = CO2eq = CO2_eq")
+# openscm_units does this for all gas species...we just have to keep up.
+ureg.define("tCO2e = t CO2e")
+
 ureg.define("LNG = 3.44 / 2.75 CH4")
 # with ureg.context("CH4_conversions"):
 #     print(ureg("t LNG").to("t CO2"))
@@ -46,7 +49,8 @@
 ureg.define('percent = 1e-2 frac = pct = percentage')
 ureg.define('ppm = 1e-6 fraction')
 
-# USD are the reserve currency of the ITR tool
+# By default, USD are the reserve currency of the ITR tool.  But data template can change that
+base_currency_unit = 'USD'
 ureg.define("USD = [currency] = $")
 for currency_symbol, currency_abbrev in ITR.data.currency_dict.items():
     ureg.define(f"{currency_abbrev} = nan USD = {currency_symbol}")
@@ -690,8 +694,14 @@ def validate(cls, quantity):
             quantity = q
         if not isinstance(quantity, Quantity):
             raise TypeError('pint.Quantity required')
-        if quantity.is_compatible_with('USD'):
-            return quantity
+        try:
+            if quantity.is_compatible_with('USD'):
+                return quantity
+        except RecursionError:
+            breakpoint()
+        for currency in ITR.data.currency_dict.values():
+            if quantity.is_compatible_with(currency):
+                return quantity
         raise DimensionalityError (quantity, 'USD', dim1='', dim2='', extra_msg=f"Dimensionality must be 'dimensionless' or compatible with [{ITR.data.currency_dict.values()}]")
 
     def __repr__(self):