def item_get_elements(self, s, type, name, filters=None):
if filters:
# Convert filter elements to strings
filters = {dim: as_str_list(ele) for dim, ele in filters.items()}
try:
# Retrieve the cached value with this exact set of filters
return self.cache_get(s, type, name, filters)
except KeyError:
pass # Cache miss
try:
# Retrieve a cached, unfiltered value of the same item
unfiltered = self.cache_get(s, type, name, None)
except KeyError:
pass # Cache miss
else:
# Success; filter and return
return filtered(unfiltered, filters)
# Failed to load item from cache
# Retrieve the item
item = self._get_item(s, type, name, load=True)
idx_names = list(item.getIdxNames())
idx_sets = list(item.getIdxSets())
# Get list of elements, using filters if provided
if filters is not None:
jFilter = java.HashMap()
for idx_name, values in filters.items():
# Retrieve the elements of the index set as a list
idx_set = idx_sets[idx_names.index(idx_name)]
elements = self.item_get_elements(s, 'set', idx_set).tolist()
# Filter for only included values and store
filtered_elements = filter(lambda e: e in values, elements)
jFilter.put(idx_name, to_jlist(filtered_elements))
jList = item.getElements(jFilter)
else:
jList = item.getElements()
if item.getDim() > 0:
# Mapping set or multi-dimensional equation, parameter, or variable
columns = copy(idx_names)
# Prepare dtypes for index columns
dtypes = {}
for idx_name, idx_set in zip(columns, idx_sets):
# NB using categoricals could be more memory-efficient, but
# requires adjustment of tests/documentation. See
# https://github.com/iiasa/ixmp/issues/228
# dtypes[idx_name] = CategoricalDtype(
# self.item_get_elements(s, 'set', idx_set))
dtypes[idx_name] = str
# Prepare dtypes for additional columns
if type == 'par':
columns.extend(['value', 'unit'])
dtypes['value'] = float
# Same as above
# dtypes['unit'] = CategoricalDtype(self.jobj.getUnitList())
dtypes['unit'] = str
elif type in ('equ', 'var'):
columns.extend(['lvl', 'mrg'])
dtypes.update({'lvl': float, 'mrg': float})
# Prepare empty DataFrame
result = pd.DataFrame(index=pd.RangeIndex(len(jList)),
columns=columns)
# Copy vectors from Java into DataFrame columns
# NB [:] causes JPype to use a faster code path
for i in range(len(idx_sets)):
result.iloc[:, i] = item.getCol(i, jList)[:]
if type == 'par':
result.loc[:, 'value'] = item.getValues(jList)[:]
result.loc[:, 'unit'] = item.getUnits(jList)[:]
elif type in ('equ', 'var'):
result.loc[:, 'lvl'] = item.getLevels(jList)[:]
result.loc[:, 'mrg'] = item.getMarginals(jList)[:]
# .loc assignment above modifies dtypes; set afterwards
result = result.astype(dtypes)
elif type == 'set':
# Index sets
# dtype=object is to silence a warning in pandas 1.0
result = pd.Series(item.getCol(0, jList), dtype=object)
elif type == 'par':
# Scalar parameters
result = dict(value=item.getScalarValue().floatValue(),
unit=str(item.getScalarUnit()))
elif type in ('equ', 'var'):
# Scalar equations and variables
result = dict(lvl=item.getScalarLevel().floatValue(),
mrg=item.getScalarMarginal().floatValue())
# Store cache
self.cache(s, type, name, filters, result)
return result
def test_filtered():
df = pd.DataFrame()
assert df is utils.filtered(df, filters=None)
