Game Development

but yeah it was only sth supersmall to get images onto a screen made of i think ~10k leds, built by a friend back when i studided electrical engineering before switching majors

Vaillancourt

Still, a nice challenge!

The most complex thing I wrote toward something like that was rending 3d boxes and spheres using SFML, in 2d.

DMGregory

2:06 PM

@Vaillancourt I think there's a persistent misconception that DirectX/OpenGL/Vulkan/Metal are middleware libraries that add an extra layer between you and the device, rather than API specs for how to talk to the driver.

So every so often we get someone saying "Not invented here! I'll roll my own for Maximum Efficiency™!"

dot_Sp0T

@DMGregory how can it be there's no tvtropes article on that term Oo

DMGregory

Guess it's more of a programming trope than a pop culture trope.

Vaillancourt

Maybe; although I wonder what ratio of users come up with this idea "for maximum efficiency" vs those who want to do it "to learn how it's done".

DMGregory

There was an interesting GDC talk about that I was watching recently...

It was arguing that computer graphics courses that focus on first principles are somewhat misaligned with the reality of working in graphics today, where submitting commands to a GPU via an API IS "how it's done" unless you're one of the very few people writing graphics card drivers.

dot_Sp0T

@Vaillancourt believe me, most are of the first kind

DMGregory

2:19 PM

The video is behind a paywall, but the slides are here

Vaillancourt

Okay, thanks!

@dot_Sp0T We'd need to track the questions we get here during the next year :P

dot_Sp0T

@DMGregory well you could argue that many/most universities are misaligned in that it is expected that people can instantly enter certain non-research jobs after finishing university education; while classic university education is and should be about understanding and learning about underlying basics

@Vaillancourt wanna spend some time sunday setting up a tracker?

Vaillancourt

TBH, I think wanting to "write your own" is fine once you understand how "the mainstream ones" are working. But once you know how "the mainstream ones" are working, you'll likely not want to write one of your own anymore :P

dot_Sp0T

yeah, which means almost all 'write your own' projects are started by folks that have not yet invested time into the mainstream stuff

Vaillancourt

Ahaha I don't really have time for this, unfortunately (nor the interest of really knowing the answer...)

dot_Sp0T

2:24 PM

i mean that is mostly how my gameengine project/idea started. yet it gradually shifted to the 'wanna understand this stuff' kind of work; and now I'm mostly using it because it works and i enjoy digging into stuff

DMGregory

@dot_Sp0T I'd say even for a research job, you should understand the current state of practice, so you don't reinvent the wheel or miss opportunities to build on battle-tested approaches.

Vaillancourt

Yeah, I don't think it's an issue per-se. You can write your own thing without knowing how the mainstream stuff is made, you'll just face the same issues they have faced and had to fix.

user92578

@DMGregory IMO that's exactly how universities should work, OpenGL/DX/Vulkan/Metal tricks become obsolete in 3 or so years, which is why graphics programming courses should definitely focus on the principles instead of the tech.

user92578

That whole thing of going concepts first, language/tech/implementation second is really strongly present in pretty much all the uni courses I have ever taken

dot_Sp0T

@DMGregory yes. But in my opinion you should understand / have worked with current practices before deciding to further deepen your understanding and study them. Studying used to be about deepening your knowledge from User to Creator. E.g. Studying Engineering meant you would be building the machines that you were using until then

Vaillancourt

2:31 PM

I have a hard time with professors teaching anything that is "used in the industry". Most of those professors I had were interested by research. Their code was bad. They knew the theory well (i.e. how it works), but the way they used the technology on a global level seemed to be lacking.

Maybe it depends on the University.

And where you live.

dot_Sp0T

@Vaillancourt def depends on the uni. the uni i study at requires profs to work parttime in the field / does not allow fulltime professures

DMGregory

To be fair, in the talk the speaker wasn't talking about cutting-edge industry tricks. But about the paradigm of submitting commands/programs for parallel execution by shader cores. That's been with us for decades and it looks like it's got some life in it for quite a while yet. ;)

user92578

@dot_Sp0T What? When do they do any research?

Vaillancourt

As you further your education, you should be going deeper in to the theory, and less in the practice.

dot_Sp0T

@Tyyppi_77 as part of their jobs i guess?

Vaillancourt

2:34 PM

And a college student should be doing more "practical" stuff and less "theory" stuff.

user92578

So X hours in the industry, Y hours teaching and Z hours research? Sounds like a busy life

DMGregory

I teach in an undergraduate degree program at a college, and probably go deeper into theory than I should 'cause I geek out about it. 😅

dot_Sp0T

The uni has a bunch of aligned institutes where most profs work, but these are required to get work from the 'real world', meaning they need to find investors that pay them to do jobs; i believe they use profits they run to do research

DMGregory

@Tyyppi_77 I do ~6 hours teaching, ~32 hours at Ubi, and ~12+ hours in lesson prep/grading... basically I don't get weekends during the school semester. 😭

user92578

Considering a research project? :D

dot_Sp0T

2:37 PM

see

user92578

We have our highest level of education split between universities and "polytechnical schools", where universities provide bachelors (3yrs), masters (2yrs) and doctorates (?yr), and the polytechnical schools provide a degree (4yrs) that is more directed towards getting a profession

DMGregory

...I am thinking about taking a year or two to get a master's degree. 😅

user92578

Oh neat!

Vaillancourt

@DMGregory Any topic in mind?

DMGregory

That's what I'm struggling with.

Vaillancourt

2:39 PM

@DMGregory But.. do you really do ~32 hours at Ubi? :P

dot_Sp0T

Here if you pursue a masters it is almost always just an extension of your bachelors

@Vaillancourt that is 4 days fulltime?

sounds right

DMGregory

@Vaillancourt I've been getting better about not putting in too much overtime. 😉

dot_Sp0T

although our good @DMGregory seems to go above and beyond regarding classprep

don't they give you a TA for grading and such?

DMGregory

I'm just reaaaallly slow.

Not for a part-timer like me. I was able to request one when I was teaching 3 sections, because without that I wouldn't have been able to juggle it all.

Vaillancourt

@DMGregory Nice ;)

DMGregory

2:42 PM

I've been looking at the masters as a means to an end, to be able to teach full-time. I'm a pretty curious person, so I'd be happy studying pretty much anything. But that makes it hard to pick ONE thing.

dot_Sp0T

at my uni all classes get assigned assistants; not even the profs themselves specifically. to be fair most profs use the same assistant in different classes they teach, but still

DMGregory

Might also be a college vs university thing. Sheridan's officially a college, even though they've started adding degree programs

dot_Sp0T

@DMGregory I am def toying with the thought of going into UX-Design. I find it a super interesting field especially because it is so diverse and deep and comparatively unexplored. So much design is about testing and finding out what works, instead of deriving proper principles that can be applied broadly

Vaillancourt

@DMGregory I guy I went to school with did his masters on dynamic difficulty in games.

dot_Sp0T

UX is all about understanding your users. which is awesome. (basically what I always understood psychohistory to be all about, if you read Asimov)

DMGregory

2:44 PM

@dot_Sp0T UX design is a huge area of development in my studio atm. 😀

dot_Sp0T

@DMGregory see :D

DMGregory

@Vaillancourt Ooooh cool! I've been thinking a bit about procedural techniques, or game design for accessibility. There's an Inclusive Technology program at a nearby school where that one could be a fit.

Vaillancourt

We also had a teacher at university for which it was his job: teaching. He was not a professor. Really passionate about all the stuff he was teaching. Really a good teacher. But he was "just" a teacher. He had to eat the crumbs that the professors left for him. To have a "full time" job, he needed four classes per semester, but he often only had 2 or 3.

dot_Sp0T

@Vaillancourt that's what language teachs at our uni have to do :/ although one of the english teachers is awesome. He's all about: I do not understand what you folks learn here, but I love hearing you explain little bits to me, that's why i teach here

DMGregory

Interesting! I almost went to teacher's college after my undergrad, but landed a job at a game studio instead. 😁

Vaillancourt

2:53 PM

@DMGregory Maybe for now you have the best of both worlds.

It looks like being a teacher is a kind of precarious/stressful situations with all the unions and the politics, while becoming a professor looks impossible because there are not a lot of openings, and a lot of people wanting the job. At least that appears to be the case here.

Teaching is fun :)

DMGregory

I think I've got a good shot at the next opening at Sheridan. It's just the lack of the master's that's holding me back atm.

Vaillancourt

If you can get away with a master's degree, then go for it, if that's what you want ;)

Gotta have the conditions to stay focused, though.

DMGregory

Yeah. I'd likely take a break from gamedev and teaching to just power through it. I know some folks have managed a master's part-time while juggling other stuff but I think that would kill me.

Vaillancourt

An option is to start part time, while you take the required classes, and once you're done with those, you stop everything and focus on the research.

/shrug

user92578

I have a co-worker who is doing his master's thesis as a part-time employee, I think he's doing the programming/development work on the thesis on the work days, and writing the thesis on the non-work days

Vaillancourt

3:06 PM

That's an option, if you can do something for your work

DMGregory

That could be good advice. Thanks for the tip!

Vaillancourt

My failed attempt at getting my Master's involved returning living with my parents while living 6 hours away from the University, not having frequent communication with the professor, not having a topic I loved, and Star Trek.

genaray

4:26 PM

Is it common to use Utility classes inside an ECS environment instead of component methods ? Are there any advantages ( Speed ? ) ? ^^

dot_Sp0T

@genaray to do what? as compared to what?

genaray

Instead of component methods :P im trying to evaluate which approach to use. I am working on an inventory mechanic, when an item gets removed, added or updated i need to mark that entity and the inventory entity itself. I need this quite often so im not sure if i simply should duplicate that code, put it into a utility class, component method or create a ecs event for this operation

user92578

100% don't duplicate it, component method vs free function is just a choice between how "purist" you want to go, and if the operation really is something of a component instead of maybe something that touches multiple components

user92578

Some people want their components to be pure data, but I'd definitely consider it case by case if a method would make sense and follows the traditional SRP properly

dot_Sp0T

the functionality you seem to need seems to be 100% related to the component; so why not have it coupled to the component? -> i.e. i do not see where you would get code duplication from

in my ECS implementation i use markers to tell if data in a component changed - so i don't have to check however-many fields depending on what is done with that data later. that code is just another field on that component though, and whatever system processes that component checks the mark instead (+ a system running at the end of each cycle that resets marks)

if i were to do an Inventory somehow, I'd have a System that is responsible for maintaining the inventory that runs towards the end of each cycle, so marked inventoryitems can be processed(?)

@Tyyppi_77 what is SRP?

genaray

4:39 PM

@Tyyppi_77 @dot_Sp0T Thanks, than i think im gonna implement this as a free function ^^ Im approaching something similar... i actually wanna be able to run over all added, removed and modified items in an inventory. Thats the main goal. So my current thought was to mark the inventory items with an "added", "updated", "removed" component where i put in entity ids that changed.

user92578

Single responsibility principle

DMGregory

Single-responsibility principle

The single-responsibility principle (SRP) is a computer-programming principle that states that every module, class or function in a computer program should have responsibility over a single part of that program's functionality, which it should encapsulate. All of that module, class or function's services should be narrowly aligned with that responsibility.Robert C. Martin, the originator of the term, expresses the principle as, "A class should have only one reason to change," although, because of confusion around the word "reason" he more recently stated "This principle is about people." ��2�...

user92578

@genaray Components as pure data is definitely a reasonable approach

DMGregory

Sounds like a dirty flag, @genaray? In that case, I'd say setting the flag in a setter method is a robust approach, so you can't accidentally forget to set it somewhere. The interface enforces the invariant that all relevant changes set the flag.

genaray

@Tyyppi_77 Thanks ! :) Nevertheless im still not that happy with my solution... I actually want to iterate over inventorys which include new items like : "(inventory, addedItems) => {}"... But that component should only stay one frame, like a one time flag.... so i end up with a bunch of conditions to check if the inventory entity already contains "addedItems" :/

@DMGregory Ah that makes sense... i heard of dirty flagging. I actually need to read the site first, but i guess you simply attach some sort of bool to the component and when its true we know that it was modified ? ^^

DMGregory

4:46 PM

Yep. Or, if you want to iterate over it, you could consider adding it to a "recently added" list, then you can have a system that processes that list, removes the items from it, and puts them into their proper inventory. That may be more efficient than scanning over all inventory items and skipping the ones without the flag set. (Depending on your use patterns)

Similarly with a "to-be-removed" list. Probably doesn't make sense for an update-in-place operation though.

dot_Sp0T

the question really is: what do the updates mean? so what is getting changed that the inventory needs to know about it?

makes sense to have methods/lists for adding/removing items from the inventory

whooooowie finally got arc'd text to work properly ( i hate trig )

genaray

@DMGregory That actually sounds like a much better way of doing it ^^ Thanks, it would be much cleaner.

DMGregory

Well done, dot_Sp0T!

In some circumstances, this kind of "waiting room" queue for new objects entering the system or destroyed objects leaving the system can help avoid order-of-update bugs or concurrent access problems, making sure you're not blowing away an item that another part of your code is still using until you've gotten to a safe time at the end of the update tick when only the clean-up remains.

dot_Sp0T

5:24 PM

spoke too early

still thanks :D

DMGregory

Uh-oh. Is it giving you trouble we can help with?

dot_Sp0T

nah, i know what's wrong. just need to calculate through different ideas until one works out

dot_Sp0T

5:55 PM

@DMGregory but if you want to, remember the image I posted a few days ago? still trying to properly figure a formula for x3 and x4 so it works for every point around a circle

^^

user92578

Don't you just take the midpoint between the two points on the circle, which with the center of the circle gives you the direction that x3 and x4 are from x1 and x2?

user92578

Then you add that multiplied by the side length to the points on the circle?

DMGregory

Got a link to the image? I don't remember it off the top of my head.

user92578

dot_Sp0T

you need to adjust for the angles. (i just got back from dinner) currently coding out a solution that does that, also takes into consideration the ratio between inner and outer circle

user92578

6:03 PM

What outer circle?

dot_Sp0T

so point between x1 and x2 on the circle. dist between x1 and xC multiplied with ratio gives the points for x3 and x4

x3 and x4 are on an imaginary outer circle

otherwise you need go crazy with angles, and that didn't work for me very well the last days i tried^^

so i switched to shifting points

user92578

Oh, that's not obvious at all from the picture that x3 and x4 land on a circle

dot_Sp0T

actually it is, as the distance between x1 and x3 is the same as x2 and x4

thus it follows that you can draw an outer circle where x3 and x4 are on

maybe though notations between Switzerland ~15 years ago, and the US are different :)

DMGregory

Do I understand correctly that you want to draw a rectangle such that the corners of one face sit on an inner circle, subtended by an angle alpha, and the opposite face sits on an outer circle (subtended by a smaller angle)?

dot_Sp0T

@DMGregory i do not know the word subtended (yet) but it sounds right

love it when the internet doesn't seem able to translate a very specific word to german

Funny thing i just saw: Top and Bot (when you shorten Bottom) are almost phonetically opposites

DMGregory

6:10 PM

I think I've used it incorrectly here. I mean the arc along the circle corresponds to a particular angle measured from the center of the circle.

user92578

I don't understand what is supposed to happen when the angle gets large

DMGregory

Yeah, and "down" is sometimes abbreviated as "dn" which is just "up" spun 180 degrees.

user92578

This isn't correct then? desmos.com/calculator/uqcv92jwig

dot_Sp0T

@DMGregory that is awesoem!

DMGregory

A game I made with a gravity-changing mechanic was originally called "up & dn" before our publisher made us change it. ;)

dot_Sp0T

6:12 PM

@DMGregory A W E S O M E - what's its name now?

@Tyyppi_77 i am having trouble reading that, are two of the 4 points meant to be in the space between circles in your example? if yes than no

user92578

Yeah the outer circle is positioned wrong since there's a normalization missing

DMGregory

So if we moved the blue point along the line through the blue and green points until it touches the outer circle, and did the same moving the purple point, keeping its direction from the black point the same, would that be your intended output?

dot_Sp0T

basically the proportions of the rectangle need to stay the same no matter how it's rotated in respect to the origin of the inner circle

i could do it with quaternions and matrices, but i'm trying to get it with normal trig

DMGregory

Let's back up a bit. What problem are you solving, or what game feature does this serve?

dot_Sp0T

in this case writing text along an arc.

user92578

6:22 PM

When b is appropriately sized for a character, moving a seems to create what you want, no? desmos.com/calculator/vwiidxmn73

user92578

But this is just an overly complex way of positioning points of equal angles on two circles of different radius?

dot_Sp0T

@Tyyppi_77 could be, i have no idea

@Tyyppi_77 sort of, but if i move b in your example it quickly breaks

user92578

So the correct behaviour would be what there is but the two points are always on the outer circle?

dot_Sp0T

yeah, just imagine shifting the rectangle from my drawing along the circle always keeping x1 and x2 on the inner circle, keeping Lw and Lh the same all time

or imagine taping a post-it to the steering-wheel of your car. when you turn the wheel the post it gets rotate too, but all corners rotate accordingly

lemme get a screenshot of the current outcome (that is wrong but almost right)

DMGregory

So can we rephrase your problem as, "How can I rotate and translate a character rectangle with a given width and height so that the bottom sits on a circle with a given center and radius, and compute the corners of the resulting rectangle?"

dot_Sp0T

6:29 PM

@DMGregory invert the last argument: "How can I rotate and translate a rectangle with a given width and height, where the top corners sit on a circle with radius a and the bottom corners sit on a circle with radius B (A + ?) where the circles share the same center?" maybe

DMGregory

That could entail scaling the rectangle as well, if the distance between the two circles does not match the height of the rectangle. Is that acceptable?

dot_Sp0T

no, the second circle is determined by the height of the rectangle, resp the resulting ratios

DMGregory

It looks like you're also distorting the rectangle into a trapezoid. Is that desired? Or would you like to keep it a rectangle?

dot_Sp0T

that is sort of the point of trying to derive a formula

@DMGregory the distortion is what is currently still wrong

DMGregory

Then it's already specified by the height and you do not need to add it as an additional given.

dot_Sp0T

6:32 PM

@DMGregory i was not sure about that, as I did not yet come around to thinking of doing math to prove that^^

if it were a trapezoid then x/y could be simply 'shifted' by adding the height to the inner radius:

r2 = r1 + height)
x1 = r1 * cos(theta)
x2 = r1 * cos(theta + alpha)
x3 = r2 * cos(theta)
x4 = r2 * cos(theta + alpha)

And the same for y using sin

user92578

I think the only thing wrong I have is the rendering of the large circle, but the rectangle behaves as you want? Or then I'm still just not understanding at all what you want

DMGregory

As I sketch this out, I think you might be parametrizing this wrong for your intended use of font rendering.

dot_Sp0T

@Tyyppi_77 seems so, i probably just have a hardtime understanding it bc I never seen or used that website before in my life

DMGregory

I don't think it's the corners of the rectangle you want to sit on the circle at all. I think you want the baseline of the character to sit tangent to a circle.

user92578

^

DMGregory

6:42 PM

The baseline can in general sit somewhere inside the rectangle - it typically won't be flush with one of its sides.

The corners might end up sitting on different circles, but it's the baseline the player will be using to visually judge whether the characters are aligned. They don't get to see the corners of your rendering quads.

dot_Sp0T

@DMGregory you're talking characters now. the rectangle is the definition from the highest ascender to the lowest descender of the given font. THat part works as expected ;)

that is also the reason i never ever mentioned characters. because then people start thinking in one of a miriad of possible ways to render characters and fonts

DMGregory

Right, but I don't think you can get all the corners aligned as you want unless you distort the characters into wedges. You can align heights along the center lines of the character, but that might then place some corners inside/outside the "mean circle", depending on the width of the character.

dot_Sp0T

@DMGregory the height works universally in my fontrenderer

believe me there

DMGregory

I'm talking about facts of geometry here. If you want, say, the visible bottom-center point of a T and an I to sit on the same circle, then the corners of their quads will not sit on the same circle, because the T is wider than the I.

dot_Sp0T

yes but the difference between that width can be described using an angle and cos/sine functions

DMGregory

6:51 PM

I can describe any arbitrary shape with cos and sine functions. That doesn't make this particular issue disappear.

dot_Sp0T

so i am trying to find the function that describes the shift of the points according to my desired result. Because I can draw this whole thing using a compass and a tirangle ruler, i know that i can also derive formulae to describe these ratios

user92578

Draw a wider rectangle into the same picture you shared a few days ago

dot_Sp0T

anyway, i rly gotta go, it 2100 here

thanks for all the help :)

@Tyyppi_77 and yeah I know what you mean, that will be an issue of middling it in somehow

user92578

It will 100% show you the issue DMGregory identified

dot_Sp0T

then the language barrier might make me misunderstand the issue he identified

user92578

7:02 PM

Nice, the picture will help you then :)

Transcript for