## About The Pull Request
Fixes issues with var typing and proc arguments, discovered using
OpenDream's WIP TypeMaker feature (using improvements I haven't PR'd
upstream yet).
## Why It's Good For The Game
Codebase maintenance.
We shouldn't really be invoking the proc itself because then we don't
pass along the failing line/file to our consumers, let's use the macro
in all instances that really need it.
I noticed people were invoking the `Fail()` proc directly rather than
using `TEST_FAIL` instead, so they weren't getting those neat
annotations on their failing code because we never passed along the
failing line/file to actually apply those annotations. That's silly. We
don't even return on `TEST_FAIL` either, so there's no reason to not do
this (only upsides wahoo).
## About The Pull Request
Now comes with a Hackmd: https://hackmd.io/ImTe5FLeTgmI7spTWKBaFQ?view
In-game screenshots:
Removes the old Mining vendor console and all its bad code. Instead, the
chef produce console was generalized and a NEW mining vendor is a
subtype of it. If they try to Express this console, it will be 1.5x the
mining points (compared to 2x for the Chef produce console atm), so it
is technically possible, but it is still better to order it through
Cargo.
Different to the Kitchen crate, this one is a private order by the Shaft
Miner, using mining points instead of Credits. Cargo CAN emitter it
open, but I think that's an acceptable risk with all crates. As shown in
the screenshot, Cargo will immediately know who ordered the items so
knows who to call to pick it up when needed.
This also means Shaft Miner's vendor is now categorized somewhat. I
tried my best to make sense out of it but some items really don't make
sense (laser pointer, soap...)
I split the different sections of orderable items into different files
for better management, and de-hardcoded it and its TGUI to make it
easier for anyone who wants to add more to it.
I also made the produce console use paths and added ways 'categories' to
produce consoles, which indicates which sections you should and
shouldn't be allowed to see.
https://github.com/tgstation/tgstation/pull/71007 already did part of
this but it isn't merged yet so :/
Free golems are mostly unaffected by this. Their console only works in
express mode and doesn't increase the prices for it. The only downside
is the cooldown.
I still have some things to finish on this PR so it'll be left as draft
until at least tomorrow.
## Why It's Good For The Game
https://hackmd.io/ImTe5FLeTgmI7spTWKBaFQ?view
1. A large problem currently with Miners is that they don't interact
with the station, this will at least help integrate them more into their
own department, by making the cost of their equipment cheaper if they
bother to actually go through Cargo for their gear.
2. It also means that a non functional Cargo would affect Shaft Miners
too, and as they have access to the shuttle, maybe we can expect some
Miners to pick up the slack if needed.
3. The old mining vendor was the ONLY vendor in the game that had
infinite stock. It doesn't need a refill or anything like any other
vendor, and every other vendor uses credits, mining points is just shaft
miner credits. Why are they an exception? At least being ordered through
the shuttle makes sense.
4. It opens the QM being able to see easier what Miners are doing, and
prevents miners from hiding on Lavaland to do nothing but hunt fauna if
they were meant to be demoted or something by the QM. Basically, gives
the QM more control over the people working in their department.
## Changelog
🆑
add: Shaft Miner's equipment vendor now orders their equipment through
the Cargo shuttle, though you can spend 1.5x the points to express it,
making it a Mining version of the Chef's produce console, with a
weakened express tax.
/🆑
Starts shaving off a lot of less than 0.1s performance killers by, in
nearly every case, just writing better code.
Numbers are amount saved.
- /obj/machinery/bluespace_vendor/LateInitialize -> 29.4ms
Changes a loop over all machines to a specialized list.
- /obj/structure/table/glass/Initialize -> 42.53ms
Stops every table from initializing glass shards and table frames before
any destruction.
- /obj/structure/chair/Initialize -> 24.64ms
Removes an unnecessary addtimer that existed for chairs that weren't
anchored in emergency shuttles. Didn't do anything.
- /datum/orderable_item/New -> 44.3ms
Instead of initializing every item to get its desc, just uses initial.
Added a unit test to make sure none are dynamic.
- /obj/machinery/computer/slot_machine/Initialize -> 26.19ms
Currently goes through every coin subtype, creates it, calls a proc,
then qdels it. Changes that to only run once. Could be optimized further
by making the coin info on a datum to avoid creating the object, but it
currently sits at 7.82ms, far below worth caring about for now.
- /obj/machinery/door_buttons/airlock_controller/findObjsByTag -> 3.51ms
Loops over just doors instead of typechecking airlock in machines.
- /obj/structure/closet/Initialize -> 60.57ms
Moves the code for taking everything on the tile from a next-tick timer
to LateInitialize.
- /obj/machinery/rnd/experimentor/Initialize -> 36.92ms
Changes a list that is generated by going through every item in the game
and getting information from a large amount of them to only run when
needed.
- /obj/structure/tank_dispenser/Initialize -> 20.81ms
No longer initializes every tank in it right away, only when needed.
- /obj/machinery/telecomms/LateInitialize -> 16.63ms
Removes `urange` to instead just loop over telecomms machines and check
distance. There's not that many of them.
- /mob/living/simple_animal/hostile/carp/cayenne/Initialize -> 3.17ms
Defers a GAGS overlay creation until its needed. BTW GAGS is
*horrendous* on init costs, and is the root cause for a lot of pretty
terrible performance. I investigated precompiling but the gains weren't
crazy, but likely could be the more stuff is GAGS'd.
- /turf/open/floor/engine/cult/Initialize -> 14.64ms
Temporary visual effect that is created is no longer done on mapload,
since nobody will see it.
- /datum/techweb/specialized/autounlocking/proc/autounlock -> 5.55ms
Changes some loops to shorter checks. This whole proc is pretty bad and
it's still 14.21ms for 17 calls.
- /matrix/New -> 13.41ms
- /matrix/proc/Translate -> 42.06ms
~~Changed the mineral matrice to only generate once, then take it from a
static.~~ An extra ~0.05s taken off by avoiding setting icon and
transform every Initialize.
Penelope Haze
