add-recipe single inserts a duplicate when the civ already knows the recipe #6022

add-recipe single <TYPE>:<ID> appends to the civ's resource vector without checking whether the subtype is already in there. Run it twice for the same item and you get two copies in historical_entity.resources.<x>_type.

Confirmed in game (DF + DFHack, fortress mode). Running

add-recipe single SHOES:ITEM_SHOES_BOOTS

twice on an item the civ already knew gave me a visible duplicate entry.

The single path in add-recipe.lua is upfront about skipping the check:

local addedItem = nil
--assume the user knows what they're doing, so no need for sanity checks
for _, item in ipairs(all) do
  if (item.id == itemId) then
    known:insert('#', item.subtype)
    addedItem = item
    break
  end
end

The bulk paths are fine. addItems guards with not checkKnown(known, subtype) before inserting, so this is specific to single.

The duplicates are mostly cosmetic, but they stick around in the save and there's no in-game way to see them or get rid of them.

Adding a checkKnown(known, item.subtype) call in that loop would make it behave like all/native.

Two smaller things in the same file, from reading it rather than playing. Happy to split either out into its own issue.

1. single doesn't route diggers. It looks the category up in the categories table, and there's no digger entry there, so WEAPON:ITEM_WEAPON_PICK ends up in resources.weapon_type. addItems deliberately sends mining weapons to resources.digger_type instead, with the comment "prevents picks from being duplicated, and puts great picks in correct category". Something does read that field: internal/quickfort/stockflow.lua:306 uses resources.digger_type to build MakeWeapon orders for diggers. So the two vectors aren't interchangeable.

Partial in-game note: add-recipe single WEAPON:ITEM_WEAPON_PICK on a civ that already knew picks did not produce a duplicate among the known diggers, which is what you'd expect if the subtype went into weapon_type instead. I did not separately confirm that it appeared under weapons, so treat that half as unverified.

2. Two nil checks that can't fire. Lines 4-12:

local resources = df.historical_entity.find(df.global.plotinfo.civ_id).resources
local civ = df.historical_entity.find(df.global.plotinfo.civ_id).entity_raw
...
if (resources == nil) then
  qerror("Could not find entity resources")
elseif (civ == nil) then

If find returns nil, line 4 has already thrown attempt to index a nil value. So running add-recipe with no fortress loaded gets you a Lua traceback instead of either of the qerror messages that were meant to come out.

I checked whether this is a general pattern before bringing it up: of the 439 scripts in DFHack/scripts, 70 sites dereference a find() result in the same expression, and add-recipe.lua is the only one that then nil-tests the result. The rest either guard first or have an argument that can't be nil. So it's one file, not a sweep.

Context: this came out of review on DFHack/scripts#1603, where @chdoc asked for it to be filed rather than worked around in the GUI.

Disclosure: I used AI assistance (Claude) for the code reading and for drafting this write-up. I verified the main bug in game myself, and I'm the one filing it.