The former implementation was sorting the quants per location and trying
to take as much quantities as possible from the same locations, to limit
the number of operations to do. Then, only, removal order (fifo, ...)
was applied. It is more important to respect removal order than limiting
the operations, so remove this "optimization".
The rules created in demo data of stock_reserve_rule make the tests of
stock_vertical_lift (and possibly other modules) fail because the
transfers can't be made available.
Deactivate the rule in stock_reserve_rule and activate it only in its
tests. Users can still activate the rule manually to test.
When rules are configured and have been applied, we should not
have an implicit fallback on the base location, as it would kind
of cancel the benefits of the rules (as it would then take whatever
it wants anywhere in all the locations).
This reverts commit 768f186fd2.
Which is not more optimized, the optimization based on parent_path
doesn't make sense here as the ORM will read parent_path in the location
and get the parent ids by splitting the ids, it doesn't need more than
one query on stock_location which is done based on its id and can reuse
the cache, there is no lookup on parent path for parent_of.
>>> env["stock.reserve.rule"].search([("location_id", "parent_of", 3125)])
2020-05-27 05:36:59,938 1 DEBUG log_p odoo.sql_db: query: SELECT "stock_location"."id" as "id","stock_location"."name" as "name","stock_location"."complete_name" as "complete_name","stock_location"."active" as "active","stock_location"."usage" as "usage","stock_location"."location_id" as "location_id","stock_location"."comment" as "comment","stock_location"."parent_path" as "parent_path", <stripped>,"stock_location"."create_uid" as "create_uid","stock_location"."create_date" as "create_date","stock_location"."write_uid" as "write_uid","stock_location"."write_date" as "write_date" FROM "stock_location" WHERE "stock_location".id IN (3125)
2020-05-27 05:36:59,942 1 DEBUG log_p odoo.sql_db: query: SELECT "stock_reserve_rule".id FROM "stock_reserve_rule" WHERE (("stock_reserve_rule"."active" = true) AND ("stock_reserve_rule"."location_id" in (1,7,8,133,134,135,144,207,3125))) ORDER BY "stock_reserve_rule"."sequence" ,"stock_reserve_rule"."id"
It could not work properly here as we need the "fallback" to apply
even if there is no quantity at all in the stock. As we hook the
reservation rules in StockMove._update_reserved_quantity(), and
this method is called only if we have at least 1 product in qty,
the fallback was not applied with zero qty.
A new module will handle this concept: https://github.com/OCA/wms/pull/28
Example of configuration:
Rule location: Stock
Removal rule 1: Stock/Zone1
Removal rule 2: Stock/Zone2
Reservation of a stock move with Stock/Zone2 as source location.
Previously, it would reserve in Stock/Zone1.
Now, it will never be allowed to reserve in Stock/Zone1.
A warning message was added previously to warn the user about potential
issues, which is now obsolete so I removed it.
Before the change, the implementation of the fallback goes like this:
If I reserve a move of 3000 and it finds 600 units, it splits the move
to create a new move of 2400 and pretend to the caller that 3000 was
reserved so the initial move is changed to 'assigned'.
Now, if we have a move of 2400 and finds zero, it still splits the move,
and pretend to the caller that 2400 was reserved → the initial move has
no move line but is assigned. In this case, we should not split the move
but only update the source location of the move.
There are cases where auto-filling of qty_done of stock move line is not desirable.
e.g. you assign quants manually for some of the moves in a picking and not the others,
in such case you need to go over all the moves in the picking to either remove qty_done
or fill it in to proceed with the validation of the entire moves. Auto-fill behavior is
also troublesome when this function is used in a manufacturing order. i.e. having
qty_done of the component move live messes up the outcome of the production.
