los.py expecting number of TAZs in land use to match length of skim index

[aletzdy]

The current beta release of ActivitySim expects the length of TAZs in the land use data to match the length of skim indices. This can cause a crash if the skims include external stations, but the land use data only contains internal zones. This is the case with SEMCOG implementation and might be with other models too.

I am brining this up here so others be aware of it, and we can discuss it if needed.

@JilanChen

[i-am-sijia]

The source code already provides a solution to this issue. See code snippet below. When the TAZs in the land use data mismatch the skims (which is reasonable as you said re: externals), Users can specify a land_use_taz table in the input_table_list in settings.yaml. The land_use_taz just need to have a TAZ column including all the TAZs in skims, increasing monotonically, see example here.

activitysim/activitysim/abm/tables/landuse.py

Lines 50 to 61 in 739c2e4

	@workflow.table
	def land_use_taz(state: workflow.State):
	try:
	df = read_input_table(state, "land_use_taz")
	except MissingInputTableDefinition:
	# if the land_use_taz table is not given explicitly in the settings,
	# we will construct our best approximation of the table by collecting
	# a sorted list of unique TAZ ids found in the land_use table of MAZs.
	# In nearly all cases this should be good enough, unless the model
	# includes TAZs without MAZs (e.g. external stations) or for some
	# reason wants TAZs in some not-sorted ordering.
	land_use = state.get_dataframe("land_use")