HatBot (dino game)
A Dino game.
---
I landed on 'HatBot' to name the entry for the Metroidvania Month (jam) 19.
Named after the main character, who runs around in a floppy mad hatter hat.
Also called: mvania19 (dino game) for the first half of the jam.
world outline
> landing site
grassy cave
tall walls that need to be climbed
wooden blocks blocking path to wall-climb power up
enemies:
- floor walker goombas
- punk-kicky turtles
powerups:
- jump
- wall-climb
required:
- wall-climb to reach snow/volcano
- wood-smash to reach wall-climb
- wood-smash to reach kingdom?
underground access to simulation and kingdom
climb the wall to reach the snow/volcano door
> simulation
spaceship/facility full of wood and metal
powerups:
- dash boots (a horizontal dash, for more quickly crossing a room)
- wood-smashing bat
enemies:
- robot enemies crawl on all walls
dash levels need to be crossed before doors close
wooden blocks need to be destroyed
later, revisit to destroy metal blocks
> kingdom
purple stone and gilded temple
powerups:
- double jump
- magnet
enemies:
- bats
- floaty witches casting spells
- coins to gather to open doors
required:
- wooden block smash
need magnet power up to get to double jump room
(gather both?)
> hot and cold
snow, volcano
a world split in two directions, choose which to tackle
enemies:
- hot boss
- cold boss
- wall-walkers
- floaty snow/lava tossers
required:
- wooden block smash
- magnet
- double jump
- dash
magnet-requiring coin door rooms
double jumps required
dash required for timer rooms
wooden blocks to destroy
postmortem
> Longest jam i've done (1 month)
before now my longest was 9 days
> ~10 minutes to play through
now that's alot of content!
> built for native platforms!
not just web this time
linux/osx/windows/html
the web build was scrappy, so now we're getting more native experience
>> codesiging /notarizing/staple was a pain
but i learned/got it done on apple
>> skipped code-signing on windows
> 5 metroidvania areas/zones
The world outline above and the tileset focus day helped get the envs going.
>> Level Zero
Very happy about this area - it was the last one added, and solved a few things at once: tutorial, a secret-end-game reward room.
It also gave space for some title and credits setup.
>>> Tutorial
>>> Credits
>>> Secret room
>> The Landing Site
Env intended as a metroid reference.
Could have used some space-ship wreckage art.
drop in from unreachable level zero elevator.
climbing gloves found here, but sword is required to get to them.
with climbing gloves, can climb to the volcano.
otherwise, can walk through enemy + candle rooms to simulation elevator.
>> Simulation
Intended as a mechanical/robot/programmed world. got some 'spaceship' naming integrated. Ended up kind of plain overall. needs more world building!
sword is accessible through a few rooms.
with sword, can collect coins to open door to candle and then boss room.
elevator to kingdom beyond boss room. technically, you can run by the boss if you want.
originally connected to the kingdom elevator right next door to the landing site elevator, but I moved it b/c it was too easy/available to skip the first boss. Otherwise there was no real reason to even go to the boss, especially with a lack of a reward.
>> The Kingdom
'golden' or 'gilded'.
tileset a bit different - blockier, less natural-looking, more man-made, in an ancient pyramids kind of vibe.
lots of coin rooms (3?) - along the gilded theme.
long fall to the double jump powerup, then double jump your way back up.
a tough enemy + coins room, with a ceiling exit - only accessible way back to level zero.
'throne room' named for the boss room. I guess the Monstroar is the King?
beyond the throne, an elevator back to under the landing site, where you have to jump and blindly double-jump and sword up to get out. a bit clumsy.
>> Volcano
You climb up to the elevator with the climbing gloves, and need them for a few of the rooms.
You know, volcano. b/c it's red!
had a red-ish canvas modulate for a while, but went to more black after a few days - felt like everything was getting washed, and the tilesets look better with the colors popping.
one coin room with a bunch of goons. probably some of the harder enemy rooms.
a long climb to the final boss, preceded by a forgiving candle room.
Then, the final boss fight.
> 3 boss battles
>> beefstronaut
fires arrows that can be knocked back with the sword.
intended flow was knock a missile back, then jump and slash while they're stunned.
normal gravity applies, but maybe it could not?
art was a strange one - weird legs and an astronaut helmet... just went with it!
i like some of the animations - the firing one lends itself to the direction it could grow in - a strong arm that wires around. maybe digital arms, something reconfigurable for a few attacks?
>> monstroar
'swoops' at you. you kind of knock at them while trying not to get hit.
floats (no gravity except in knocked-back/stunned/dead states)
art was odd too - ended up with a head and big handlebar mustache. kind of works for a 'floating head' thing.
>> both
warp around between multiple fixed points.
knocked back -> stunned -> warp phase
'laugh' state, largely unrealized
>> fighting both at the same time
both made for the most fun fight, i think. probably these bosses belong together more than they belong apart. maybe that makes sense b/c of how they were conceived - one boss broken into two, then both added to the same final fight.
> 3 mechanics
>> sword
impled in isolation - could be attached to any node (in theory)
tying it to a beehive state could be interesting - could those states be reused
by player, boss, and enemy machine? makes me think of a mirror/copy enemy/boss
ended up with a large hitbox - i exaggerated the strike alot more in a second version of the animation.
can be aimed up/down/left/right based on player input. kind of hollow knightish.
feels chunky, not really nimble. maybe i know the code too much to really feel it? not sure what i'd do differently yet.
an interesting thing done here (and on the kick) was supporting attacks on specific frame numbers of the animation. it's tricky to support not hitting things multiple times, but also making sure more than just the first frame is used to determine what to hit.
>> double jump
Not too bad to impl, but some crossover/overlap with the coyote jump i'd originally handled. should a coyote jump support an extra/double afterwards? or does introducing the double mean dropping coyote?
i leaned towards more generous air time. not much in the game to really enforce or hold players back from reaching something they're going for.
jump/double jump still feels a bit clunky, passing double-jump allowed flags between jump/fall/etc states.
>> climb
first wall climb impl!
used a few ray casts:
- high/low on players' top/bottom
- can't remember what these do...
- to-ground ray cast
- helped prevent wall-climbing when it's not necessary, like climbing stairs
the state may have been complicated, or maybe not enough. I ended up with a fixed climbing speed, and you can just hang-forever, no worries.
> boss rewards ( dev secrets )
an interesting move - rather than unlocking/earning mechanics, you get some developer tip.
>> shift for slowmo
>> e for screenshake
>> p for debug ui
> 3 non-boss enemy types
>> soldier (turtle)
more potential here, and made 5 other alt sprites for it.
each with their own weapon. expanding on these would be good pixel art practice at the least.
I also realized i could add some of them as background art, especially b/c they have idle states already. (feeling street fighter bg art-ish).
>>> kick attack
similar impl to the sword's attack - check specific frames for hitbox collision.
fighting games probably have hitboxes per frame of the animation - i wonder if i should reach for that soon. so far i've been leaving things a bit loose.
>> goomba
doesn't do much.
borrowed art from gloomba in ghost.
added a death state towards the end.
should we remove enemy collision layers when they die? sometimes they dead enemies stack in awkward ways, b/c the collision shape is pretty large vs the death art frame. would be fine if the collision shape for the death anim was more accurate, but i tend to just leave them.
maybe animationTree style characers lend themselves to multiple collision shapes more easily.
>> shooty-crawly
sticks to whatever wall it starts near - kind of a nice impl.
also does some 'laughy' things when it hits.
their bullets can be fired back with the sword - it completely reuses the boss bullet script.
> game objects
>> coins
collect to open coin-doors... and that's it
>> doors
open from one side, not the other
energy/fancy animation
>> elevators
pretty cool, travel between areas
godot ui for selecting destination (in editor)
>> candles
provide light. can light them, then sit to heal (then they go out).
they are also checkpoints that you return to.
>> bullet (boss and enemy created)
can be fired back
pop animation
> new addons
>> Hotel
DB for game state + node lifecycle
named for it's original purpose: room management.
grew out of how weird it was to select a destination elevator in the ui. like, why don't we have this data?
>>> api
- book(path or packed scene) static data
- register(node) in _ready() (calls check_out then check_in)
- check_out(data) stored data setup a node based on whatever data was stored
- check_in(node, optional_data) updates to the data health loss, open/close of door, on/off candle, etc
- hotel_data() - impled on a node to build data dict
>>> query(opts)
a nice query-dictionary for fetching things in a group, with a growing list of
other convenient filters, and an anon func option like:
Hotel.query({"filter": func(ent): return
is_thing(ent) and is_other_thing(ent)})
>>> Hotel UI
interactive filtering/toggling to see what the db data is
>>> HUD elems read from hotel updates now
nicer than tying HUD to player somewhere
and easy to add more HUD elems - everything just comes from a
hotel.entry_updated signal
>>> boilerplate is a little much still
two functions, plus a call in 'register'
but that's a mixin, i suppose
>>> when to book? probably in ready/register, if it hasn't been booked already.
>>> learned to read data from packed scenes!
>> Debug
Logging with pretty-printing!
a huge win, really. Debugging/scanning the output so much easier
plus it made for really easy re-use in HotelUI
>>> learned to read the stack
unfortunately it doesn't work in prod
> extended addons
>> Hood
>>> Actions Hint component
got a ton of use out of this.
includes an 'action' to 'key' map, which is nice for not hard-coding keys everywhere
>>> Actions api/abstraction
ActionDetector on actor, ActionArea on interactable object, Action as a class
>>> Jumbotron component
presenting an interstitial to the player
>> Camera
>>> in-game Debugging UI
>>> POAs added for 'anchoring' points
> mvania19 dir
>> MvaniaGame, MvaniaArea, MvaniaRoom
Three classes that support a metroidvania -like game.
areas are sometimes called zones
some leakage of the MvaniaGame.player into other classes (bosses, enemies).
>> maps > demoland
Bunch of areas where we tried out some tiling techniques (solid blocks,
autotiling tilemaps)
also got enemies, and elevators working, and refactored hotel to support
persisting mvania data
>> maps > hatbot
the areas themselves, plus some quests that were reused across them
> reused entities
>> DungeonCrawler Coin
With a pretty large addition to the script
>> Ghosts Gloomba
> potential refactors/abstractions
>> 'Player' autoload/api?
Feels like we could use a shared 'Player' abstraction or autoload somewhere. There's dependence on Hood to find it, but Camera needs it as well, plus a few other things.
probably useful for managing upgrades/spawn points, etc
>> 'Room' abstraction?
runnerRoom, mvaniaRoom - there were a few times i needed to connect to the current thing's room to get a trigger/event. plus the whole thing depended on the room for some life-cycle stuff (pause/unpause).
>> quests system
quests were created adhoc, but it feels like they could be codified
- jumbotron intro
- jumbotron on complete
- jumbotron for progress? via pause menu
- other progress/completion components/code helpers
>>> CoinDoorQuest
>>> BossDefeatQuest
>> jumbotron/action-hint based dialog/menu abstraction
something data-driven that presents text, arbitrary components, and a set of options for interacting with the whole thing?
>> Mvania Area/Room editor tools
level editor? maybe.
might be low lift to add features right on top of godot's editor.
I have in mind a room regenerator that can scale/shrink rooms without breaking door constraints. maybe crazy, but it's what i want. easier in the editor, or in a full 'game' level editor?
>> sprite animation management - aseprite import helper?
adding a frame to an aesprite tag pushes all the frames down, which updates animations in place in godot's animatedsprite2d. feels like we could be creating the godot sprite_frames from the aesprite file directly....
i have some concern about losing control in this process, or needing to edit/tweak things. we'll still need to set animation speeds and looping/not-looping, for example.
> takeaway/cleanup todos
>> Remove/reduce other game autoloads
They're noisy in the logs/errors when the game runs
we should be able to work around them, or at least the preload errors they thow
>> handle licenses and game assets properly
currently symlinking in fonts, which means the repo is broken
>> build for platforms in the cloud, not on my machine
be nice to offload the build process, now that we're building for 4 platforms
>> add tileset alts, apply color-swap shader
tilesets could be providing more variety - alternate tilesets make things much
nicer
>> could learn/reimpl with animation tree-based state machines
i've seen this recommended, but can't tell if it's better for less-code-y folks
or if it's actually cleaner/better for other reasons
> wish i could have
>> env/aesthetic particle effects
I dream of celeste-level feel. I really want this!
Closing this gap is progress for me - i want something that looks fuller, more alive.
I think it's just checking more of the boxes.
I didn't hit:
- mid-ground particle effects
- props (of any kind)
- bg props
- foreground art
- multiple tilesets - i did this in elevator rooms, but not across areas
- alt tiles - i made sure this was possible, but mostly i need to setup a better tile aseprite file in general
- player movement dust
>> destructible wood animation
I tried this a few times, hoping to do something simple based on the tileset as source.
one surprise, i had trouble applying a shader to a single cell in a tileset. I could create a separate piece of art to slice and animate, but that's annoying.... but if necessary, i can do it. just was surprising that it wasn't quick to do.
>> proper third boss in volcano
the first two bosses exhausted me to work on. could have been the inflection point of my burn out, and was also the first time i'd approached something like this. boss battles seem very 'custom', but ultimately i landed on a shared state machine, and each 'boss attack' (swoop, fire, laugh, warp) is impled as a separate state in that machine, so not too bad, really.
I'd say it's not reusable either, but apparently it is, when you can just throw them into a room together later.
Still - what should this game's boss have been? overall there's not much throughline/world cohesion happening.
>> foreground, background props/art items
trees, lava, snow, etc
touched on this with celeste-feel boxes (or hollow knight-feel, name pending) note
>> bring the rooms/envs to life more
birds, critters, leaves, foliage, etc
touched on this with celeste-feel boxes (or hollow knight-feel, name pending) note
>> space ship wreakage art in landing site
make a space ship, then wreak it. could be better to make a whole animation?
>> boss battle background art
anything to make these big rooms more interesting
>> volcano level lava pits
i looked into a water tutorial, and it seems doable, tho a bit of a lift. would be good to reach for something fancy like that, it's a great feature for interesting environments.
>> more interesting title screen
maybe fog of war and move the mouse around to uncover it
>> MUSIC!
>> laugh
I like this idea and want it! Probably works with a player 'stunned' state, that waits for input to get back up again.
Probably helps to prevent repeatedly kicking the player while they're trying to figure out what's going on.
>>> 'laugh' sound, for bosses and enemies
>>> better/exaggerated 'laugh' animation, for bosses and enemies
>> use the extra turtle art/characters i made
cuz duh
> future work/unused ideas
>> fast travel elevators
>> light flooding into rooms through unused doors
>> SPIKES?
lava? water?
untraversable tiles of some kind
>> beefstronaut more animations
give those arms/legs/whatever more abilities - punches, stomps, knocking arrows back again?
>> boss rush mode
a new area?
or a fancy bit of code to drop you in boss rooms? would need to close the doors, so players don't just run right off during the fights.
>> extra turtle art + characters
created idle states for ~6 turtles, but only used one with a kick... and called it a soldier? didn't like that much.
More chars might be less work here, especially if we're sharing state machines.
>> enemy goon groups - packs of enemies
feels like ai should be more enemies moving in packs, like they should know about each other and even travel-around together.
I was close this idea in protomoon - like, a level 3 em with level 2s nearby, those with level 1s near them.
> happy about
>> level zero and secret room
>> hotel db
>> camera focus improvements
>> room-oriented apis
>> candles as save and healing points
>> elevators and selecting a destination
>> bunch of tilesets created
more world building components
>> tileset process learned/improved
>> 'laugh' state in enemies/bosses
not impled well, but a nice idea
Backlinks
I love rooms (game design element) as a unit of game design and iteration. A rooms api should support that.
Dino has seen a few different rooms api approaches already. The Runner Room code managed cycling rooms ahead of the player so they could be revisited. In Tower Jet, the Reptile Room was built up to generate (and regenerate) the tiles based on noise images. Then, Mvania / Hatbot featured a fuller room and area management, including room data lifecycle via hotel and pausing/unpausing the rooms based on the player location.
A month-long game jam for making a metroidvania!
Something about that game that only the devs know - maybe it's a tool they use to test things that was left in just for fun.
hatbot used dev secrets as rewards for winning the boss battles.