def allocate_dropped_sector_data(df_load, target_sector_level):
"""
Determine rows of data that will be lost if subset data at target sector level
Equally allocate parent NAICS to child NAICS where child NAICS missing
:param df: df, FBS format
:param target_sector_level: str, target NAICS level for FBS output
:return: df, with all child NAICS at target sector level
"""
# exclude nonsectors
df = replace_NoneType_with_empty_cells(df_load)
rows_lost = pd.DataFrame()
for i in range(2, sector_level_key[target_sector_level]):
# create df of i length
df_x1 = df.loc[
(df[fbs_activity_fields[0]].apply(lambda x: len(x) == i))
& (df[fbs_activity_fields[1]] == '')]
df_x2 = df.loc[(df[fbs_activity_fields[0]] == '') & (
df[fbs_activity_fields[1]].apply(lambda x: len(x) == i))]
df_x3 = df.loc[
(df[fbs_activity_fields[0]].apply(lambda x: len(x) == i))
& (df[fbs_activity_fields[1]].apply(lambda x: len(x) == i))]
df_x = pd.concat([df_x1, df_x2, df_x3], ignore_index=True, sort=False)
# create df of i + 1 length
df_y1 = df.loc[
df[fbs_activity_fields[0]].apply(lambda x: len(x) == i + 1)
| df[fbs_activity_fields[1]].apply(lambda x: len(x) == i + 1)]
df_y2 = df.loc[
df[fbs_activity_fields[0]].apply(lambda x: len(x) == i + 1)
& df[fbs_activity_fields[1]].apply(lambda x: len(x) == i + 1)]
df_y = pd.concat([df_y1, df_y2], ignore_index=True, sort=False)
# create temp sector columns in df y, that are i digits in length
df_y.loc[:, 'spb_tmp'] = df_y[fbs_activity_fields[0]].apply(
lambda x: x[0:i])
df_y.loc[:, 'scb_tmp'] = df_y[fbs_activity_fields[1]].apply(
lambda x: x[0:i])
# don't modify household sector lengths or gov't transport
df_y = df_y.replace({'F0': 'F010', 'F01': 'F010'})
# merge the two dfs
df_m = pd.merge(df_x,
df_y[[
'Class', 'Context', 'FlowType', 'Flowable',
'Location', 'LocationSystem', 'Unit', 'Year',
'spb_tmp', 'scb_tmp'
]],
how='left',
left_on=[
'Class', 'Context', 'FlowType', 'Flowable',
'Location', 'LocationSystem', 'Unit', 'Year',
'SectorProducedBy', 'SectorConsumedBy'
],
right_on=[
'Class', 'Context', 'FlowType', 'Flowable',
'Location', 'LocationSystem', 'Unit', 'Year',
'spb_tmp', 'scb_tmp'
])
# extract the rows that are not disaggregated to more specific naics
rl = df_m[(df_m['scb_tmp'].isnull())
& (df_m['spb_tmp'].isnull())].reset_index(drop=True)
# clean df
rl = replace_strings_with_NoneType(rl)
rl_list = rl[['SectorProducedBy',
'SectorConsumedBy']].drop_duplicates().values.tolist()
# match sectors with target sector length sectors
# import cw and subset to current sector length and target sector length
cw_load = load_sector_length_crosswalk()
nlength = list(sector_level_key.keys())[list(
sector_level_key.values()).index(i)]
cw = cw_load[[nlength, target_sector_level]].drop_duplicates()
# add column with counts
cw['sector_count'] = cw.groupby(nlength)[nlength].transform('count')
# merge df & conditionally replace sector produced/consumed columns
rl_m = pd.merge(rl,
cw,
how='left',
left_on=[fbs_activity_fields[0]],
right_on=[nlength])
rl_m.loc[rl_m[fbs_activity_fields[0]] != '',
fbs_activity_fields[0]] = rl_m[target_sector_level]
rl_m = rl_m.drop(columns=[nlength, target_sector_level])
rl_m2 = pd.merge(rl_m,
cw,
how='left',
left_on=[fbs_activity_fields[1]],
right_on=[nlength])
rl_m2.loc[rl_m2[fbs_activity_fields[1]] != '',
fbs_activity_fields[1]] = rl_m2[target_sector_level]
rl_m2 = rl_m2.drop(columns=[nlength, target_sector_level])
# create one sector count column
rl_m2['sector_count_x'] = rl_m2['sector_count_x'].fillna(
rl_m2['sector_count_y'])
rl_m3 = rl_m2.rename(columns={'sector_count_x': 'sector_count'})
rl_m3 = rl_m3.drop(columns=['sector_count_y'])
# calculate new flow amounts, based on sector count,
# allocating equally to the new sector length codes
rl_m3['FlowAmount'] = rl_m3['FlowAmount'] / rl_m3['sector_count']
rl_m3 = rl_m3.drop(columns=['sector_count'])
# append to df
if len(rl) != 0:
vLogDetailed.warning(
'Data found at %s digit NAICS not represented in current '
'data subset: {}'.format(' '.join(map(str, rl_list))), str(i))
rows_lost = rows_lost.append(rl_m3, ignore_index=True)
if len(rows_lost) != 0:
vLogDetailed.info(
'Allocating FlowAmounts equally to each %s associated with '
'the sectors previously dropped', target_sector_level)
# add rows of missing data to the fbs sector subset
df_w_lost_data = pd.concat([df, rows_lost], ignore_index=True, sort=True)
df_w_lost_data = replace_strings_with_NoneType(df_w_lost_data)
return df_w_lost_data
def check_for_missing_sector_data(df, target_sector_level):
"""
Modeled after datachecks.py check_if_losing_sector_data
Allocates flow amount equally across child NAICS when parent NAICS is not target_level
:param df:
:param target_sector_level:
:return:
"""
from flowsa.flowbyfunctions import replace_NoneType_with_empty_cells, replace_strings_with_NoneType
# temporarily replace null values with empty cells
df = replace_NoneType_with_empty_cells(df)
activity_field = "SectorProducedBy"
rows_lost = pd.DataFrame()
cw_load = load_sector_length_crosswalk_w_nonnaics()
for i in range(3, sector_level_key[target_sector_level]):
# create df of i length
df_subset = df.loc[df[activity_field].apply(lambda x: len(x) == i)]
# import cw and subset to current sector length and target sector length
nlength = list(sector_level_key.keys())[list(
sector_level_key.values()).index(i)]
cw = cw_load[[nlength, target_sector_level]].drop_duplicates()
# add column with counts
cw['sector_count'] = cw.groupby(nlength)[nlength].transform('count')
# merge df & replace sector produced columns
df_x = pd.merge(df_subset,
cw,
how='left',
left_on=[activity_field],
right_on=[nlength])
df_x[activity_field] = df_x[target_sector_level]
df_x = df_x.drop(columns=[nlength, target_sector_level])
# calculate new flow amounts, based on sector count, allocating equally to the new sector length codes
df_x['FlowAmount'] = df_x['FlowAmount'] / df_x['sector_count']
df_x = df_x.drop(columns=['sector_count'])
# replace null values with empty cells
df_x = replace_NoneType_with_empty_cells(df_x)
# append to df
sector_list = df_subset[activity_field].drop_duplicates()
if len(df_x) != 0:
log.warning('Data found at ' + str(i) +
' digit NAICS to be allocated'
': {}'.format(' '.join(map(str, sector_list))))
rows_lost = rows_lost.append(df_x, ignore_index=True, sort=True)
if len(rows_lost) == 0:
log.info('No data loss from NAICS in dataframe')
else:
log.info('Allocating FlowAmounts equally to each ' +
target_sector_level)
# add rows of missing data to the fbs sector subset
df_allocated = pd.concat([df, rows_lost], ignore_index=True, sort=True)
df_allocated = df_allocated.loc[df_allocated[activity_field].apply(
lambda x: len(x) == sector_level_key[target_sector_level])]
df_allocated.reset_index(inplace=True)
# replace empty cells with NoneType (if dtype is object)
df_allocated = replace_strings_with_NoneType(df_allocated)
return df_allocated
def check_if_losing_sector_data(df, df_subset, target_sector_level):
"""
Determine rows of data that will be lost if subset data at target sector level
In some instances, not all
:param fbs:
:return:
"""
df = df.fillna(fbs_fill_na_dict)
# exclude nonsectors
df = df.replace({'nan': '', 'None': ''})
rows_lost = pd.DataFrame()
for i in range(2, sector_level_key[target_sector_level]):
# create df of i length
df_x1 = df.loc[
(df[fbs_activity_fields[0]].apply(lambda x: len(x) == i))
& (df[fbs_activity_fields[1]] == '')]
df_x2 = df.loc[(df[fbs_activity_fields[0]] == '') & (
df[fbs_activity_fields[1]].apply(lambda x: len(x) == i))]
df_x3 = df.loc[
(df[fbs_activity_fields[0]].apply(lambda x: len(x) == i))
& (df[fbs_activity_fields[1]].apply(lambda x: len(x) == i))]
df_x = pd.concat([df_x1, df_x2, df_x3], ignore_index=True, sort=False)
# create df of i + 1 length
df_y1 = df.loc[
df[fbs_activity_fields[0]].apply(lambda x: len(x) == i + 1)
| df[fbs_activity_fields[1]].apply(lambda x: len(x) == i + 1)]
df_y2 = df.loc[
df[fbs_activity_fields[0]].apply(lambda x: len(x) == i + 1)
& df[fbs_activity_fields[1]].apply(lambda x: len(x) == i + 1)]
df_y = pd.concat([df_y1, df_y2], ignore_index=True, sort=False)
# create temp sector columns in df y, that are i digits in length
df_y.loc[:, 'spb_tmp'] = df_y[fbs_activity_fields[0]].apply(
lambda x: x[0:i])
df_y.loc[:, 'scb_tmp'] = df_y[fbs_activity_fields[1]].apply(
lambda x: x[0:i])
# don't modify household sector lengths
df_y = df_y.replace({'F0': 'F010', 'F01': 'F010'})
# merge the two dfs
df_m = pd.merge(df_x,
df_y[[
'Class', 'Context', 'FlowType', 'Flowable',
'Location', 'LocationSystem', 'Unit', 'Year',
'spb_tmp', 'scb_tmp'
]],
how='left',
left_on=[
'Class', 'Context', 'FlowType', 'Flowable',
'Location', 'LocationSystem', 'Unit', 'Year',
'SectorProducedBy', 'SectorConsumedBy'
],
right_on=[
'Class', 'Context', 'FlowType', 'Flowable',
'Location', 'LocationSystem', 'Unit', 'Year',
'spb_tmp', 'scb_tmp'
])
# extract the rows that are not disaggregated to more specific naics
rl = df_m[(df_m['scb_tmp'].isnull()) & (df_m['spb_tmp'].isnull())]
# clean df
rl = clean_df(rl, flow_by_sector_fields, fbs_fill_na_dict)
rl_list = rl[['SectorProducedBy',
'SectorConsumedBy']].drop_duplicates().values.tolist()
# match sectors with target sector length sectors
# import cw and subset to current sector length and target sector length
cw_load = load_sector_length_crosswalk_w_nonnaics()
nlength = list(sector_level_key.keys())[list(
sector_level_key.values()).index(i)]
cw = cw_load[[nlength, target_sector_level]].drop_duplicates()
# add column with counts
cw['sector_count'] = cw.groupby(nlength)[nlength].transform('count')
# merge df & conditionally replace sector produced/consumed columns
rl_m = pd.merge(rl,
cw,
how='left',
left_on=[fbs_activity_fields[0]],
right_on=[nlength])
rl_m.loc[rl_m[fbs_activity_fields[0]] != '',
fbs_activity_fields[0]] = rl_m[target_sector_level]
rl_m = rl_m.drop(columns=[nlength, target_sector_level])
rl_m2 = pd.merge(rl_m,
cw,
how='left',
left_on=[fbs_activity_fields[1]],
right_on=[nlength])
rl_m2.loc[rl_m2[fbs_activity_fields[1]] != '',
fbs_activity_fields[1]] = rl_m2[target_sector_level]
rl_m2 = rl_m2.drop(columns=[nlength, target_sector_level])
# create one sector count column
rl_m2['sector_count_x'] = rl_m2['sector_count_x'].fillna(
rl_m2['sector_count_y'])
rl_m3 = rl_m2.rename(columns={'sector_count_x': 'sector_count'})
rl_m3 = rl_m3.drop(columns=['sector_count_y'])
# calculate new flow amounts, based on sector count, allocating equally to the new sector length codes
rl_m3['FlowAmount'] = rl_m3['FlowAmount'] / rl_m3['sector_count']
rl_m3 = rl_m3.drop(columns=['sector_count'])
# append to df
if len(rl) != 0:
log.warning('Data found at ' + str(i) +
' digit NAICS not represented in current '
'data subset: {}'.format(' '.join(map(str, rl_list))))
rows_lost = rows_lost.append(rl_m3, ignore_index=True, sort=True)
if len(rows_lost) == 0:
log.info(
'No data loss from subsetting the dataframe by specified sector length'
)
else:
log.info('Allocating FlowAmounts equally to each ' +
target_sector_level +
' associated with the sectors previously being dropped')
# add rows of missing data to the fbs sector subset
df_w_lost_data = pd.concat([df_subset, rows_lost],
ignore_index=True,
sort=True)
df_w_lost_data = df_w_lost_data.replace({'': None})
return df_w_lost_data