def initialize_new_units(buildings, residential_units):
"""
This data maintenance step initializes units for buildings that have been
newly created, conforming to the data requirements of the
'residential_units' table.
Data expectations
-----------------
- 'buildings' table has the following columns:
- index that serves as its identifier
- 'residential_units' (int, count of units in building)
- 'residential_units' table has the following columns:
- index named 'unit_id' that serves as its identifier
- 'building_id' corresponding to the index of the 'buildings' table
Results
-------
- extends the 'residential_units' table, following the same schema as the
'initialize_residential_units' model step
"""
# Verify initial data characteristics
'''
ot.assert_orca_spec(OrcaSpec(
'',
TableSpec(
'buildings',
ColumnSpec('building_id', primary_key=True),
ColumnSpec('residential_units', min=0)),
TableSpec(
'residential_units',
ColumnSpec('unit_id', primary_key=True),
ColumnSpec('building_id', foreign_key='buildings.building_id'))))
'''
old_units = residential_units.to_frame(residential_units.local_columns)
bldgs = buildings.to_frame(['residential_units', 'deed_restricted_units', 'building_id'])
# Filter for residential buildings not currently represented in
# the units table
new_bldgs = bldgs[~bldgs.building_id.isin(old_units.building_id)]
new_bldgs = new_bldgs[new_bldgs.residential_units > 0]
if len(new_bldgs) == 0:
return
# Create new units, merge them, and update the table
new_units = _create_empty_units(new_bldgs)
all_units = dev.merge(old_units, new_units)
all_units.index.name = 'unit_id'
print "Creating %d residential units for %d new buildings" % \
(len(new_units), len(new_bldgs))
orca.add_table('residential_units', all_units)
# Verify final data characteristics
'''
def initialize_new_units(buildings, residential_units):
"""
This data maintenance step initializes units for buildings that have been
newly created, conforming to the data requirements of the
'residential_units' table.
Data expectations
-----------------
- 'buildings' table has the following columns:
- index that serves as its identifier
- 'residential_units' (int, count of units in building)
- 'residential_units' table has the following columns:
- index named 'unit_id' that serves as its identifier
- 'building_id' corresponding to the index of the 'buildings' table
Results
-------
- extends the 'residential_units' table, following the same schema as the
'initialize_residential_units' model step
"""
# Verify initial data characteristics
'''
ot.assert_orca_spec(OrcaSpec(
'',
TableSpec(
'buildings',
ColumnSpec('building_id', primary_key=True),
ColumnSpec('residential_units', min=0)),
TableSpec(
'residential_units',
ColumnSpec('unit_id', primary_key=True),
ColumnSpec('building_id', foreign_key='buildings.building_id'))))
'''
old_units = residential_units.to_frame(residential_units.local_columns)
bldgs = buildings.to_frame(['residential_units', 'deed_restricted_units'])
# Filter for residential buildings not currently represented in
# the units table
new_bldgs = bldgs[~bldgs.index.isin(old_units.building_id)]
new_bldgs = new_bldgs[new_bldgs.residential_units > 0]
# Create new units, merge them, and update the table
new_units = _create_empty_units(new_bldgs)
all_units = dev.merge(old_units, new_units)
all_units.index.name = 'unit_id'
print "Creating %d residential units for %d new buildings" % \
(len(new_units), len(new_bldgs))
orca.add_table('residential_units', all_units)
# Verify final data characteristics
'''
def add_buildings(buildings, new_buildings, remove_developed_buildings=True):
old_buildings = buildings.to_frame(buildings.local_columns)
new_buildings = new_buildings[buildings.local_columns]
if remove_developed_buildings:
unplace_agents = ["households", "jobs"]
old_buildings = \
_remove_developed_buildings(old_buildings, new_buildings,
unplace_agents)
all_buildings = dev.merge(old_buildings, new_buildings)
orca.add_table("buildings", all_buildings)
def add_buildings(buildings, new_buildings,
remove_developed_buildings=True):
old_buildings = buildings.to_frame(buildings.local_columns)
new_buildings = new_buildings[buildings.local_columns]
if remove_developed_buildings:
unplace_agents = ["households", "jobs"]
old_buildings = \
_remove_developed_buildings(old_buildings, new_buildings,
unplace_agents)
all_buildings = dev.merge(old_buildings, new_buildings)
orca.add_table("buildings", all_buildings)
def developer_reprocess(buildings, year, years_per_iter, jobs,
parcels, summary, parcel_is_allowed_func):
# this takes new units that come out of the developer, both subsidized
# and non-subsidized and reprocesses them as required - please read
# comments to see what this means in detail
# 20% of base year buildings which are "residential" have job spaces - I
# mean, there is a ratio of job spaces to res units in residential
# buildings of 1 to 5 - this ratio should be kept for future year
# buildings
s = buildings.general_type == "Residential"
res_units = buildings.residential_units[s].sum()
job_spaces = buildings.job_spaces[s].sum()
to_add = res_units * .05 - job_spaces
if to_add > 0:
print "Adding %d job_spaces" % to_add
res_units = buildings.residential_units[s]
# bias selection of places to put job spaces based on res units
print res_units.describe()
print res_units[res_units < 0]
add_indexes = np.random.choice(res_units.index.values, size=to_add,
replace=True,
p=(res_units/res_units.sum()))
# collect same indexes
add_indexes = pd.Series(add_indexes).value_counts()
# this is sqft per job for residential bldgs
add_sizes = add_indexes * 400
print "Job spaces in res before adjustment: ", \
buildings.job_spaces[s].sum()
buildings.local.loc[add_sizes.index,
"non_residential_sqft"] += add_sizes.values
print "Job spaces in res after adjustment: ",\
buildings.job_spaces[s].sum()
# the second step here is to add retail to buildings that are greater than
# X stories tall - presumably this is a ground floor retail policy
old_buildings = buildings.to_frame(buildings.local_columns)
new_buildings = old_buildings.query(
'%d == year_built and stories >= 4' % year)
print "Attempting to add ground floor retail to %d devs" % \
len(new_buildings)
retail = parcel_is_allowed_func("retail")
new_buildings = new_buildings[retail.loc[new_buildings.parcel_id].values]
print "Disallowing dev on these parcels:"
print " %d devs left after retail disallowed" % len(new_buildings)
# this is the key point - make these new buildings' nonres sqft equal
# to one story of the new buildings
new_buildings.non_residential_sqft = new_buildings.building_sqft / \
new_buildings.stories * .8
new_buildings["residential_units"] = 0
new_buildings["residential_sqft"] = 0
new_buildings["building_sqft"] = new_buildings.non_residential_sqft
new_buildings["stories"] = 1
new_buildings["building_type"] = "RB"
# this is a fairly arbitrary rule, but we're only adding ground floor
# retail in areas that are underserved right now - this is defined as
# the location where the retail ratio (ratio of income to retail sqft)
# is greater than the median
ratio = parcels.retail_ratio.loc[new_buildings.parcel_id]
new_buildings = new_buildings[ratio.values > ratio.median()]
print "Adding %d sqft of ground floor retail in %d locations" % \
(new_buildings.non_residential_sqft.sum(), len(new_buildings))
all_buildings = dev.merge(old_buildings, new_buildings)
orca.add_table("buildings", all_buildings)
new_buildings["form"] = "retail"
# this is sqft per job for retail use - this is all rather
# ad-hoc so I'm hard-coding
new_buildings["job_spaces"] = \
(new_buildings.non_residential_sqft / 445.0).astype('int')
new_buildings["net_units"] = new_buildings.job_spaces
summary.add_parcel_output(new_buildings)
# got to get the frame again because we just added rows
buildings = orca.get_table('buildings')
buildings_df = buildings.to_frame(
['year_built', 'building_sqft', 'general_type'])
sqft_by_gtype = buildings_df.query('year_built >= %d' % year).\
groupby('general_type').building_sqft.sum()
print "New square feet by general type in millions:\n",\
sqft_by_gtype / 1000000.0
def developer_reprocess(buildings, year, years_per_iter, jobs,
parcels, summary, parcel_is_allowed_func):
# this takes new units that come out of the developer, both subsidized
# and non-subsidized and reprocesses them as required - please read
# comments to see what this means in detail
# 20% of base year buildings which are "residential" have job spaces - I
# mean, there is a ratio of job spaces to res units in residential
# buildings of 1 to 5 - this ratio should be kept for future year
# buildings
s = buildings.general_type == "Residential"
res_units = buildings.residential_units[s].sum()
job_spaces = buildings.job_spaces[s].sum()
to_add = res_units * .05 - job_spaces
if to_add > 0:
print "Adding %d job_spaces" % to_add
res_units = buildings.residential_units[s]
# bias selection of places to put job spaces based on res units
print res_units.describe()
print res_units[res_units < 0]
add_indexes = np.random.choice(res_units.index.values, size=to_add,
replace=True,
p=(res_units / res_units.sum()))
# collect same indexes
add_indexes = pd.Series(add_indexes).value_counts()
# this is sqft per job for residential bldgs
add_sizes = add_indexes * 400
print "Job spaces in res before adjustment: ", \
buildings.job_spaces[s].sum()
buildings.local.loc[add_sizes.index,
"non_residential_sqft"] += add_sizes.values
print "Job spaces in res after adjustment: ",\
buildings.job_spaces[s].sum()
# the second step here is to add retail to buildings that are greater than
# X stories tall - presumably this is a ground floor retail policy
old_buildings = buildings.to_frame(buildings.local_columns)
new_buildings = old_buildings.query(
'%d == year_built and stories >= 4' % year)
print "Attempting to add ground floor retail to %d devs" % \
len(new_buildings)
retail = parcel_is_allowed_func("retail")
new_buildings = new_buildings[retail.loc[new_buildings.parcel_id].values]
print "Disallowing dev on these parcels:"
print " %d devs left after retail disallowed" % len(new_buildings)
# this is the key point - make these new buildings' nonres sqft equal
# to one story of the new buildings
new_buildings.non_residential_sqft = new_buildings.building_sqft / \
new_buildings.stories * .8
new_buildings["residential_units"] = 0
new_buildings["residential_sqft"] = 0
new_buildings["deed_restricted_units"] = 0
new_buildings["building_sqft"] = new_buildings.non_residential_sqft
new_buildings["stories"] = 1
new_buildings["building_type"] = "RB"
# this is a fairly arbitrary rule, but we're only adding ground floor
# retail in areas that are underserved right now - this is defined as
# the location where the retail ratio (ratio of income to retail sqft)
# is greater than the median
ratio = parcels.retail_ratio.loc[new_buildings.parcel_id]
new_buildings = new_buildings[ratio.values > ratio.median()]
print "Adding %d sqft of ground floor retail in %d locations" % \
(new_buildings.non_residential_sqft.sum(), len(new_buildings))
all_buildings = dev.merge(old_buildings, new_buildings)
orca.add_table("buildings", all_buildings)
new_buildings["form"] = "retail"
# this is sqft per job for retail use - this is all rather
# ad-hoc so I'm hard-coding
new_buildings["job_spaces"] = \
(new_buildings.non_residential_sqft / 445.0).astype('int')
new_buildings["net_units"] = new_buildings.job_spaces
summary.add_parcel_output(new_buildings)
# got to get the frame again because we just added rows
buildings = orca.get_table('buildings')
buildings_df = buildings.to_frame(
['year_built', 'building_sqft', 'general_type'])
sqft_by_gtype = buildings_df.query('year_built >= %d' % year).\
groupby('general_type').building_sqft.sum()
print "New square feet by general type in millions:\n",\
sqft_by_gtype / 1000000.0
def initialize_new_units(buildings, residential_units):
"""
This data maintenance step initializes units for buildings that have been
newly created, conforming to the data requirements of the
'residential_units' table.
Data expectations
-----------------
- 'buildings' table has the following columns:
- index that serves as its identifier
- 'residential_units' (int, count of units in building)
- 'residential_units' table has the following columns:
- index named 'unit_id' that serves as its identifier
- 'building_id' corresponding to the index of the 'buildings' table
Results
-------
- extends the 'residential_units' table, following the same schema as the
'initialize_residential_units' model step
"""
# Verify initial data characteristics
'''
ot.assert_orca_spec(OrcaSpec(
'',
TableSpec(
'buildings',
ColumnSpec('building_id', primary_key=True),
ColumnSpec('residential_units', min=0)),
TableSpec(
'residential_units',
ColumnSpec('unit_id', primary_key=True),
ColumnSpec('building_id', foreign_key='buildings.building_id'))))
'''
old_units = residential_units.to_frame(residential_units.local_columns)
bldgs = buildings.to_frame(['residential_units', 'deed_restricted_units'])
# Filter for residential buildings not currently represented in
# the units table, and create new units
new_bldgs = bldgs[~bldgs.index.isin(old_units.building_id)]
new_bldgs = new_bldgs[new_bldgs.residential_units > 0]
new_units = _create_empty_units(new_bldgs)
# Filter for residential buildings where ADUs were added and
# create new units
old_units_by_bldg = old_units.groupby(['building_id']).agg({
'unit_num':
max,
'num_units':
'count'
}).reset_index()
old_units_by_bldg.rename(columns={
'num_units': 'num_units_old',
'unit_num': 'max_num_old'
},
inplace=True)
old_bldgs = bldgs[bldgs.index.isin(old_units.building_id)]
old_bldgs = old_bldgs[old_bldgs.residential_units > 0]
adu_bldgs = pd.merge(old_bldgs,
old_units_by_bldg,
left_index=True,
right_on='building_id')
adu_bldgs['adu_count'] = \
adu_bldgs.residential_units - adu_bldgs.num_units_old
adu_bldgs = adu_bldgs[adu_bldgs.adu_count > 0]
if len(adu_bldgs) > 0:
new_adus = pd.DataFrame({
'unit_residential_price':
0.0,
'unit_residential_rent':
0.0,
'num_units':
1,
'building_id':
np.repeat(adu_bldgs.building_id,
adu_bldgs.adu_count.values.astype(int)),
# counter of the AUDs in a building
'unit_num_adu':
np.concatenate([
np.arange(num_adus)
for num_adus in adu_bldgs.adu_count.values.astype(int)
]),
# ADUs are not deed restricted
'deed_restricted':
0.0,
'adu_count_start':
np.repeat(adu_bldgs.max_num_old + 1,
adu_bldgs.adu_count.values.astype(int))
}).sort_values(by=['building_id', 'unit_num_adu']).reset_index(
drop=True)
# update unit_num of ADUs to continue with the previous
# unit_num of non-ADUs in the same buildings
new_adus.unit_num = \
new_adus.unit_num_adu + new_adus.adu_count_start
new_adus.drop(columns=['adu_count_start', 'unit_num_adu'],
inplace=True)
new_adus.index.name = 'unit_id'
new_units = dev.merge(new_units, new_adus)
else:
print('No ADUs were built.')
# Merge new units with old units and update the table
all_units = dev.merge(old_units, new_units)
all_units.index.name = 'unit_id'
print("Creating %d residential units for %d new buildings" %
(len(new_units), len(new_bldgs)))
orca.add_table('residential_units', all_units)
# Verify final data characteristics
'''
