paved roads

[yobbo]

I'm curious as to how you pulled off the straight edges-- good work.

With not a pixel to spare .

I think it looks very nice, it's just a bit too wide with the way transitions are drawn in this screenshot.

This is as thin as the road transitions can get without creating a new method of drawing terrain .

I would like to see how this turned out if it where drawn like the bridge is now. This is because I find the road nice looking, but a bit wide with this trans and I think it would look better if a bit narrower.

This is exactly what I meant when I suggested that the best way to make roads would be as an overlay, earlier in this thread .

I might try to figure out how this would work in WML. I really really hope someone is kind enough to implement separate background / overlay tiles.


Woop! I can attach the tiles .


Really, is there some internal reason why I can't use 5-sided (and even 6) transitions?

EDIT: Yep, there was. Fixed in next post.

[yobbo]

Aha - there was some internal reason, in that the 5 and 6 sided transitions just hadn't been described in WML .

[Jetrel]

I would like to see how this turned out if it where drawn like the bridge is now. This is because I find the road nice looking, but a bit wide with this trans and I think it would look better if a bit narrower.

This is exactly what I meant when I suggested that the best way to make roads would be as an overlay, earlier in this thread .

I might try to figure out how this would work in WML. I really really hope someone is kind enough to implement separate background / overlay tiles.

As do I. These really look great, but would look a lot better that way.

[yobbo]

I read it badly.

However, if we are going to create an algorithm that will determine the sides to use, why wouldn't we use it for all road variations, including the "/", "", and "|"? It doesn't make sense to me to have 4 different letters for a terrain type that would only need 1. Especially since the way the editor currently works, you would need to change what terrain you were selecting every time you wanted to make a road curve.

Yeah, it should be changed like that. I don't know how hard the WML for this stuff would be though. Thats a question for Ayin I guess.

Hmm, it looks like the old "canyon" WML is perfectly suited for this.

At a quick count, to have images for every combination of road junctions there would need to be something like 61 images, not including variations.

I think I could cut this down to 12 or 18 by using the bigger tiles at non-trivial junctions. I'll see what I can come up with...

[hands]

Hmm, it looks like the old "canyon" WML is perfectly suited for this.

That seemed like the most logical solution when I first read it. The canyon WML would seem pefect if it behaved more consistently. There are points where canyons aren't connected correctly or they aren't connected in a very pretty way. Maybe someone could take the existing canyon WML and use it as a starting point. I actually think that if the code was changed to work well for roads, it would improve the way that canyons are drawn as well.

On second thought...
The problem with the canyons might simply be a lack of directional variations. If that's the case, copying the canyon code verbatim (and adding any necessary directions) might be the best solution.

[freim]

I actually think that if the code was changed to work well for roads, it would improve the way that canyons are drawn as well.

The canyon has been replaced by the new chasm ( http://www.wesnoth.org/forum/viewtopic.php?t=7936 ) drawn with the castle WML macro.

[Eleazar]

The old canyon WML code couldn't properly use the tiles that it had, (unless there were no branches) That was my motivation for changing to the "castle-like" transitions for the chasm, but i discovered that it was a better method even if the old way worked perfectly.

However, i think it would be great if the old canyon code could be fixed to line things up properly. We could, for instance, make a dirt road, with ruts that would curve around and follow the road.

[Darth Fool]

Actually, I think that the castle transitions or a slight variation, might be a better starting point for making a narrow road. And by better, I mean requires less images to be made for the same graphic quality. The problem is one of how we think about the castle transitions. Normally we
think of them as being fixed on a corner and thus we need an image for each corner(3 after reflections) times the number of different possible adjacent hex combos (4 in the naive case) = 12.

What we want, is something that transitions between adjacent hexes, not the corners which oddly enough is 4 hexes(after reflection) * 2 possibilities (road or no road in adjacent hex) = 8. The corner case can be used to cover the adjacent case, but not vice versa. So, the current castle corner transitions could be used.

Now, it should be possible to define an adjacent hex macro which would be even easier to use. It would have the additional feature that it would enable one to design castles with lots of variations a little more easily. Define towers on the corners, and walls with the adjacent hex.

I may give it a try at some point, but Ayin is the real master. Anyone else is welcome to give it a go as well.

[hands]

The old canyon WML code couldn't properly use the tiles that it had, (unless there were no branches) That was my motivation for changing to the "castle-like" transitions for the chasm, but i discovered that it was a better method even if the old way worked perfectly.

I understood that. I think your chasm solved the problem very well.

However, i think it would be great if the old canyon code could be fixed to line things up properly. We could, for instance, make a dirt road, with ruts that would curve around and follow the road.

I agree with this. Sometimes the attitude here seems to be too willing to throw out the old because only one approach seems "good" at a particular time. I still like the old canyon. Your chasm seems too deep for some scenarios. It would require some work to update, but I don't think the old canyon's a lost cause. The code (and tile-work) just wasn't sufficient to satisfy variation requirements.

I like yobbo's idea to apply the old canyon code to roads because it would allow for narrower roads with straight lines. I have made a complete list of all tiles necessary in order to accomplish this well (the list could be abbreviated by using reflections). Yobbo may not like the work necessary for all of this, but the biggest reason that the old canyons didn't look as nice as they could have is because there weren't sufficient tiles (only 30 out of 64), and it doesn't look like reflections were used. The road tiles might be a bit easier to make though because after the unidirectional tiles are made, layers could be used to create all the rest. To create the unidirectional tiles, a transparent border would just need to be created on all but one side. The new road borders could then be added to each. I'd even be glad to help make some directional variations. Of course, by using reflections, the number of necessary tiles could be reduced to as low as 24, but I'm not sure how well that would work for everything or how much the code would need to be modified.

Directions (variations) [tiles needed if reflections are utilized as much as possible]
None (1) [1]
base tile not extending to any side

One (6) [2]
n, ne, nw, s, se, sw

Two (15) [6]
n-ne, n-nw, n-s, n-se, n-sw, ne-nw, ne-s, ne-se,
ne-sw, nw-s, nw-se, nw-sw, s-se, s-sw, se-sw

Three (20) [6]
n-ne-nw, n-ne-s, n-ne-se, n-ne-sw, n-nw-s
n-nw-se, n-nw-sw, n-s-se, n-s-sw, n-se-sw
ne-nw-s, ne-nw-se, ne-nw-sw, ne-s-se, ne-s-sw
ne-se-sw, nw-s-se, nw-s-sw, nw-se-sw, s-se-sw

Four (15) [6]
n-ne-nw-s, n-ne-nw-se, n-ne-nw-sw, n-ne-s-se, n-ne-s-sw
n-ne-se-sw, n-nw-s-se, n-nw-s-sw, n-nw-se-sw, n-s-se-sw
ne-nw-s-se, ne-nw-s-sw, ne-nw-se-sw, ne-s-se-sw, nw-s-se-sw

Five (6) [2]
n-ne-nw-s-se, n-ne-nw-s-sw
n-ne-nw-se-sw, n-ne-s-se-sw
n-nw-s-se-sw, ne-nw-s-se-sw

Six (1) [1]
all

Total (64) [24]

In the end, it might turn out that the castle WML would work best in all cases where tiles join in odd ways, but to my knowledge the drawback to the castle WML is that it doesn't seem well suited to straight lines. That is the biggest reason to me why the old canyon code seems to be a good idea. Another reason is that even though the number of necessary tiles seems to be quite high, there would be little or no need for transition tiles.

[freim]

I agree with this. Sometimes the attitude here seems to be too willing to throw out the old because only one approach seems "good" at a particular time. I still like the old canyon. Your chasm seems too deep for some scenarios. It would require some work to update, but I don't think the old canyon's a lost cause. The code (and tile-work) just wasn't sufficient to satisfy variation requirements.

We experiment with a lot of stuff not just one approach, but I think it's a good policy to replace when we have an alternative that looks better and works better. If not we would end up with heaps of half-finished, half-working duplicate terrain, basicly a mess.

If someone fixes up the old canyon to be drawn like the new one and adds the missing stuff, there is nothing that says it can't be included again.

[hands]

We experiment with a lot of stuff not just one approach, but I think it's a good policy to replace when we have an alternative that looks better and works better. If not we would end up with heaps of half-finished, half-working duplicate terrain, basicly a mess.

That's understandable, and I appreciate the reasoned response.

If someone fixes up the old canyon to be drawn like the new one and adds the missing stuff, there is nothing that says it can't be included again.

What do you mean by, "drawn like the new one?" I assume that by "the missing stuff" you mean the additional directional tiles and the code to take the new tiles into account.

[freim]

What do you mean by, "drawn like the new one?" I assume that by "the missing stuff" you mean the additional directional tiles and the code to take the new tiles into account.

If it is to be fixed up I think it should be drawn like the current chasm, so it can be of any width also. Yes, that's what I refered to by "missing stuff".

[hands]

If it is to be fixed up I think it should be drawn like the current chasm, so it can be of any width also.

Maybe if I'm feeling ambitious one day, I'll ask for clarification on that.

Here is a file that contains some possible tiles for a narrower road that could be overlaid on top of existing terrain.

EDIT: I have created all possible directional variations.

[hands]

Here are the rest of the directional tiles. I hope they're useful.

[yobbo]

Using a grand total of eleven (11) new images, one (1) terrain type ("R"), and some shiny new WML: