Ideas for expiration dates

[LarrySnyder]

I'm thinking something roughly like this:

A Product will have a new attribute called shelf_life. This indicates the number of periods the product can survive before it expires. If shelf_life = None, the product has an infinite shelf life; it never expires.

The expiration-date bookkeeping will happen in NodeStateVars. Each relevant attribute will have an extra level of nesting that indicates the number of periods until expiration.

For example: Currently inventory_level[prod] = inventory level of product prod. Under the new scheme, inventory_level[prod][exp] = inventory level of product prod that expires in period exp. If exp = None, it refers to items that will never expire.

Similarly, inbound_shipment[p][rm][exp] = shipment quantity of raw material rm that expires in period exp that is arriving at the node from predecessor node p in the current period. inbound_shipment_pipeline[p][rm][r][exp] = shipment quantity of raw material rm that expires in period exp and will arrive from predecessor node p in r periods from the current period. And so on.

[LarrySnyder]

Then, some changes need to be made in sim.py. For example, in _receive_inbound_shipments():

stockpyl/src/stockpyl/sim.py

Lines 767 to 774 in da7de61

	# Loop through raw materials at predecessor.
	for rm_index in (p.product_indices if p is not None else [node._external_supplier_dummy_product.index]):
	if rm_index in node.raw_materials_by_product('all', return_indices=True, network_BOM=True) and \
	p in node.raw_material_suppliers_by_raw_material(rm_index, network_BOM=True):
	# Determine number of items that will be received from p (if there is no disruption),
	# not including inbound disrupted items waiting to be received.
	ready_to_receive = node.state_vars_current.inbound_shipment_pipeline[p_index][rm_index][0]

ready_to_receive will have to be indexed by exp, i.e., by different expiration dates.

stockpyl/src/stockpyl/sim.py

Lines 790 to 799 in da7de61

	# Set inbound_shipment attribute.
	node.state_vars_current.inbound_shipment[p_index][rm_index] = IS
	# Remove shipment from pipeline.
	node.state_vars_current.inbound_shipment_pipeline[p_index][rm_index][0] = 0
	# Add shipment to raw material inventory.
	node.state_vars_current.raw_material_inventory[rm_index] += IS
	# Update on-order inventory.
	node.state_vars_current.on_order_by_predecessor[p_index][rm_index] -= ready_to_receive
	# Update inbound_disrupted_items.
	node.state_vars_current.inbound_disrupted_items[p_index][rm_index] += IDI

Line 791 would change to something like

for exp in node.state_vars_current.inbound_shipment[p_index][rm_index].keys():
    node.state_vars_current.inbound_shipment[p_index][rm_index][exp] = IS[exp]

And so on. In _raw_materials_to_finished_goods(),

stockpyl/src/stockpyl/sim.py

Lines 879 to 883 in da7de61

	# Process units: remove from raw material and add to finished goods.
	for rm_index in node.raw_materials_by_product(prod_index, return_indices=True, network_BOM=True):
	node.state_vars_current.raw_material_inventory[rm_index] \
	-= num_to_make * node.NBOM(product=prod_index, predecessor=None, raw_material=rm_index)
	node.state_vars_current.inventory_level[prod_index] += num_to_make

the deductions from raw_material_inventory should happen in increasing order of exp, with None coming last. In _process_outbound_shipments(),

stockpyl/src/stockpyl/sim.py

Lines 1009 to 1010 in da7de61

	# Update IL and BO.
	node.state_vars_current.inventory_level[prod_index] -= node.state_vars_current.inbound_order[s_index][prod_index]

the deductions from inventory_level should happen in increasing order of exp, etc.

[LarrySnyder]

And, at each time period, we need to examine every expiration-date-indexed pile at every level -- in raw material inventory, finished goods inventory, and in-transit inventory, at every node -- and if the expiration date is before or equal to the current period, then it expires.

Expiring means that it disappears -- we remove that entry from the nested dict. And we should keep track of expired products, through a new attribute in NodeStateVars.

[LarrySnyder]

There are a lot of details to work out -- this is just a rough sketch. Let's discuss.