Recording camera (and mouse?) position in replays

[Dragon45]

Request: Record position of player's camera (and maybe mouse) position, mode, rotation angle, etc, so that if i have say, Player 1 selected, and I press some key, it basically shows me everything he sees and everything he was looking at and doing until i press the key again and toggle it off. In other words, it would hijack your camera and bind to whatever that player was doing.

Maybe even something for keypresses, that outputs something showing what the player was pressing at any given point in time?


http://spring.clan-sy.com/phpbb/viewtop ... 82&start=0

[KDR_11k]

Would be hard since it would require transferring that data to the other computers in first place.

[rattle]

That doesn't matter when it's only getting recorded in a local replay. Camera, cursor, etc information could be stored in a seperate file for each replay.

[KDR_11k]

Yeah but that wouldn't let you see the movements of a player that isn't the one who uploaded the replay. You'd probably want to use that to see how your opponent is playing, not how you did. Also it'd probably need a mention of the running lua widgets since otherwise it'd look pretty weird when the player clicks on empty areas.

[Tim Blokdijk]

Still it's a fun idea, the Black and White games had this to.
Maybe it could be implemented (if anybody ever thinks it's important enough to implement) by treating the camera as a unit, then the camera position could be updated like any first person controlled unit. Might even allow the camera to be targeted (and killed)?

[rattle]

Yeah but that wouldn't let you see the movements of a player that isn't the one who uploaded the replay. You'd probably want to use that to see how your opponent is playing, not how you did. Also it'd probably need a mention of the running lua widgets since otherwise it'd look pretty weird when the player clicks on empty areas.

Hmm right. That sounds like a huge increase of traffic.

[ironized]

could it be done so you can at least see the players build que?

[KDR_11k]

Um, you can see all queued orders of all units the same way you would if you're playing yourself (hold shift with the unit selected).

[theHive]

Request: Record position of player's camera (and maybe mouse) position, mode, rotation angle, etc,

You only need to note the points where the corners of the screen intersect the ground, like the box drawn in for where your screen is

and the mouse would just be the point the mouse intersects with the ground

all of it is simple raytracing, and the code is already there, just needs to be put to use

[KDR_11k]

That's more data and less accurate than just taking the camera position and direction. Also the mouse will not always point at the ground and dealing with 2d coordinates is pretty easy anyway.

The problem is that it has to be updated very often since it's unsynced code and will run more often than 30 times per second (which is the rate the synced code updates at).

[theHive]

That's more data and less accurate than just taking the camera position and direction. Also the mouse will not always point at the ground and dealing with 2d coordinates is pretty easy anyway.

The problem is that it has to be updated very often since it's unsynced code and will run more often than 30 times per second (which is the rate the synced code updates at).

Yeah, for the mouse i meant 2d coordinates, where the ray for the mouse intersects the ground.

You wouldn't need it to be every single update in the replay, for the screen you could have it update every second or other arbitrary time period (I'd make it more often for the mouse though).

[KDR_11k]

What makes you think it's a good idea to use the coordinate where the mouse intersects the ground when you can just store the screen coordinate? The mouse won't always intersect with the ground but it will always be on the screen (excluding windowed mode but you can handle that, too). Saves you one ray cast, gives you a smaller value range and causes less issues.

Updating the screen and mouse asynchronously is pointless because then you'll see the mouse click on things the camera doesn't show which makes it just as useless. Besides, sending a set of 2d coordinates and adding a camera matrix isn't that much of a difference.

[theHive]

Alright, whatever. No doubt you have more experience in this than me

[rattle]

I don't see the logic behind this either.