Hello folks!

Hey.

I figured out what the difference in our corona curves is. One is reported new infections, the other is currently infected people. The number of reported new infections can drop to 0 suddenly, the number of currently infected will lag behind by 2-ish weeks.

I suppose looking at the new infections is more useful when estimating the effectiveness of measures.

Yes.

I guess the "protocol" of "what is a new infection" is quite clear, but, here at least, it's not so clear of what happens to decide when someone is no longer infected.

@Pikalek thanks

@TheMattbat999 You're welcome. I'm not sure if an improperly worded copyright diminishes the strength, but it's one of those things that if it looks out of spec, I tend to question how legit the site/source is. And if you're concerned about retaining the rights to your work, you owe it to yourself to make sure you're actually protecting them. I wish you luck w/ your game.

@Vaillancourt Here they made up a formula for that. No hospital stay means 14 days after getting infected or reporting infection. Hospital stay with known release means release date + 7 days. Hospital stay without known release means 28 days after infection or report.

It's probably wildly inaccurate for estimations, but at least the method is known.

Yeah; better than nothing I guess?

When dealing with large numbers, consistency is probably more important than accuracy with each individual case. The errors will tend to even-out in the big picture, as long as the methodology is the same.

At some point, here, we had "oh, you were with person X which got infected? Then you're probably infected too. Stay home and wait Y days after the symptoms are gone"

I came to realize that we can't compare here with the US.

Here, it's the government that pays for healthcare, so they do care about infections. As I understand it, in the US, government don't pay for healthcare, so why would they care if people are infected?

^ Well, it might play into getting re-elected. Or tax revenue. And according to legend, some members of the government are human beings.

[Citation Needed]

Yeah, there are probably some of that too.

And as I see around here, it's probably easier to deal with those who want increased restriction by just ignoring them than it is to deal with those who don't care about the rules and completely ignore them.

But sarcasm & cynicism aside, there are enough differences that I don't think the US & Canada make good 1 to 1 comparisons for the pandemic. We have trouble even comparing internally between states.

Yes.

From our side of the border, we see what happens in the US and we're kind of... WTF guys...

But the situation is much different.

Well, most of us. Ford's looking south like "Hey, there's a model I want to follow!"

Yes, we had Maxime Bernier that tried to go like Trump, but it did not work :P

@DMGregory But our good PM Trudeau will not let him have it!

Add in the other serious life & death stuff that we can't seem to get a handle on, and it feels like a hot mess.

Yep

Is there a clean way to get a gradient vector from a flat shaded triangle? I came up with something, but I wonder if I overlooked an obvious, less expensive solution.

/shrug

;)

Fair enough - I need to see if what I'm using it for checks out before I go down the path of "make it fast" anyway. I just had this nagging feeling that there's a better way than bludgeoning the problem with a geometry hammer.

In what context? In a fragment shader, or at a higher level?

Higher level. I'm messing w/ a straight skeleton lib. Given a 2d poly it returns a series of 2d polys & corresponding 'heights' which I'm visualizing into a normalized grey scale image. Recovering the shading gradient is one option I'm exploring for converting it into something that the graphics API can use to render it. I have a solution, but I feel like I missed a better one.

There's a bunch of other stuff that could be a weak link though - for instance I'm not sure if the skeleton lib is performant with a full sized problem.

I'm going to mess around with it some. It might be that the gradient idea is needed, so I'm not asking for any problem solving - I just wondered if there was an off the shelf known approach.

Hmm.. the gradient should be related to the unit normal vector of the polygon, right? Something vaguely like normal.xz/normal.y?

Ah, but negated. So if you have three points on the triangle ABC, you can compute the vectors AB and AC, then form the normal n = AB x AC (or the opposite order, depending on your coordinate conventions/winding). Then your gradient vector is (-nx/ny, -nz/ny)

(Assuming height is along the y - swap as necessary)

Intuitively, the more the normal leans in a particular direction, the more steeply the surface is rising in the opposite direction.

Ah, actually winding doesn't matter, since the signs cancel out. So using either +n or -n should give you the same gradient vector in the end. That's convenient! 😀