Question/Design: Map generation:

[Gilgamensch]

Arctic > Antarctic

I don't get this one?!?!?!?

Adn the difference between sphere and globe is a rather minor one, in my eyes. But it ain't a doughnut. Then they shall have better taken spheroid.

[Flinx]

Take a towel (paper does not bend too well)
Decide which edge is 'North'
Join the 'East' and 'West' sides together
- you now have a cylinder
Join the 'North' and 'South' sides together
- You now have a torus aka a doughnut

Now if your towel had actually been a map of the world, traveling 'North' on your map you would take the path:
France > Benelux > Norway > Arctic > Antarctic > South Africa > Algeria > France
whereas such a trip in the real world (which is approximately spherical) takes the path:
France > Benelux > Norway > Arctic > Siberia > China > Indonesia > Australia > Antarctic > South Africa > Algeria > France .

[Flinx]

Question:

Why is a North-South (but not east-west) connected map called a 'Neptune' world?

If it is because someone thinks the planet Neptune in our solar system is tipped over on its side, um…, I hate to break the news but it is Uranus that is tipped over.

[Martin Gühmann]

Originally posted by Gilgamensch
Adn the difference between sphere and globe is a rather minor one, in my eyes. But it ain't a doughnut. Then they shall have better taken spheroid.

The difference between a sphere and a globe and also a ball is a minor one, but a map that connects the Sourth with the North and the West with the East is everything else than a ball it is a ring or doughnut.

Originally posted by Flinx
If it is because someone thinks the planet Neptune in our solar system is tipped over on its side, um�, I hate to break the news but it is Uranus that is tipped over.

I always had the assumption that I was wrong. So I have to fix the source code.

-Martin

[Gilgamensch]

Originally posted by Flinx
Take a towel (paper does not bend too well)
Decide which edge is 'North'
Join the 'East' and 'West' sides together
- you now have a cylinder
Join the 'North' and 'South' sides together
- You now have a torus aka a doughnut

If you see it this way, Yes.

But North/South (same for West/East) is defined a bit different. So this explanantion isn't really logical by itself.

I still would prefer the expression: Globe, that's what it is. The problem with CTP2-map: Globe is that there is no North/South-pole. If you only have a flat map (East/West-connected), yes the North is North and the South is South. But that was one reason why I suggested the above formula, to really have two poles.

[Flinx]

Originally posted by Gilgamensch
If you see it this way, Yes.

The map in CTP2 (CTP, Civ, Civ2, Civ3...) is a square grid. This can only ever be a flat, page like, map. We can try to make it more like the real world by joining the right and left (east and west) sides. However having a wall at the top and bottom of the map seems unrealistic to some people too, so the easiest way to get rid of this effect is to connect the top and bottom (North and South) in the same way as the right and left were connected.

But North/South (same for West/East) is defined a bit different. So this explanantion isn't really logical by itself.

I still would prefer the expression: Globe, that's what it is. The problem with CTP2-map: Globe is that there is no North/South-pole. If you only have a flat map (East/West-connected), yes the North is North and the South is South. But that was one reason why I suggested the above formula, to really have two poles.

You could prefer to call it a banana, but that does not change the fact that a square grid can never accurately model a sphere/spheroid/globe. The only way to model a sphere is a geodesic (i.e. pentagons and hexagons or triangles arranged as pentagons and hexagons).


OK I have helped to take this thread off on a tangent, but the intent of this thread is to discuss improving the model for map generation, and I do not see how we can do that when the implications of the most fundamental aspect of the model we are developing i.e. the square grid, seem to be in some doubt.

Ignoring any considerations of terrain type and unit movement limited by terrain type, the square grid is the easiest model to use and is what we have got.

The standard east-west joined “Earth World” map is obtained by projecting the spherical earth onto a cylinder thus:

The poles are beyond the top and bottom of the map and are not shown. This is not a problem as the poles are (or are assumed to be) ice covered wasteland and impracticable to traverse. Thus the top and bottom edges of the map are impenetrable walls.

Now this choice of orientation of the cylinder relative to the “poles” while natural in many ways is not the only one, and a situation described in the first post (with the pole at a random (Xp,Yp) would be obtained from this type of projection:

The impenetrable walls at the non-wrapping edges do present a problem (for me) in this case however. I suppose the suggestion of pretending in your mind that the torus/doughnut shape is really a globe can work, but I do not see how this in any way improves the model. Starting positions are random so this neutralises any changes introduced by the random pole proposal (in my point of view) with the exception that the top and bottom of the minimap are not always the cold areas.

An alternative way of dealing with the top and bottom edges may be to connect them “across the pole” in this manner:
On a 50x100 map
Moving up from 0,0 puts you on 0,50
Moving up from 0,1 puts you on 0,51
…
Moving up from 0,49 puts you on 0,99
Similarly
Moving down from 50,0 puts you on 50,50
Moving down from 50,1 puts you on 50,51
…
Moving down from 50,49 puts you on 50,99
I do not know how hard this would be to implement in the code, and I do not know how strangely it might affect gameplay that two squares are next to each other in one direction and 50 squares apart in another.
[edit]Fixed a small typo.

[Gilgamensch]

1.) Flinx and the rest: I opened this threat for exactly this purpose, so that we can discuss this problems and hopefully find a solution.

The map in CTP2 (CTP, Civ, Civ2, Civ3...) is a square grid. This can only ever be a flat, page like, map. We can try to make it more like the real world by joining the right and left (east and west) sides. However having a wall at the top and bottom of the map seems unrealistic to some people too, so the easiest way to get rid of this effect is to connect the top and bottom (North and South) in the same way as the right and left were connected.

That is my idea as well. And to make it more realistic I proposed the random pole and ist anti-pole. About the definition North/South, that is were the magnetic pole is and as we don't have to bother with this we just have to handle the temperatur, so we could choose the position randomly, therefore allowing the algorithm (which doesn't exist yet) to 'know' where are the coldest areas and could start to calculate/randomise from there.

The poles are beyond the top and bottom of the map and are not shown. This is not a problem as the poles are (or are assumed to be) ice covered wasteland and impracticable to traverse. Thus the top and bottom edges of the map are impenetrable walls.

No I don't agree at all with it. It might have been like this in the old days, but nowadays we could even live there. It is only a question of energy. And you are completly denying the existance of i.e. Inuks(sp?)/Lapps(sp?)/Iceland and so on. They lived there already for quite a while. Even before us 'civilized' people intruded into this area.

The only way to model a sphere is a geodesic (i.e. pentagons and hexagons or triangles arranged as pentagons and hexagons).

Agreed, but me thinks out of range so we have to take what we have got and try to make the best out of it.

By the way, nice graphics

The impenetrable walls at the non-wrapping edges do present a problem (for me) in this case however. I suppose the suggestion of pretending in your mind that the torus/doughnut shape is really a globe can work, but I do not see how this in any way improves the model. Starting positions are random so this neutralises any changes introduced by the random pole proposal (in my point of view) with the exception that the top and bottom of the minimap are not always the cold areas.

For the random starting position: With a proper model in place, it would allow (after playtesting) the computer to calculate better the relative value of the starting area and could give somebody living near the pole like 1 or 2 more settlers to offset the slow population growth. In the moment the computer can't really evaluate it at all. To give it a better meaning: If you are -+10 tiles away from Xp,Yp (or the anti-pole) then you will get 3 settlers. If -+20 tiles 2 settlers and the rest as usual.

An alternative way of dealing with the top and bottom edges may be to connect them “across the pole” in this manner:

I know some games where they did it like this and it is rather confusing. One problem for this is: We know that a map shall be bent to reflect the real shape of the world (spheroid), but for any reason the human brain can easier adapt to a situation with a map having wrong shape, meaning the map not being geodesic formed.

Flinx, sorry if my reply was a bit lengthy

[The Big Mc]

Well I want an ice continent it keeps the game interesting and it’s a late game challenge to build a city in the ice. I would also like to suggest that people should be prevented from altering ice covered tiles (or too far away from there cities if this is to much). To offset this we would add new tile imps (post genetic) to help people colonize these places.

[Gilgamensch]

Actually with a better biome-model in use, it could be possible. With the slider temperature an Agorithm could determine how far the poles would stretch, to say it easy

[The Big Mc]

what an ice world

[Ekmek]

I think primemover is looking on this topic

[drulius]

It seems to me that y'all are trying to mix apples and oranges. Gameplay is tied to the square grid map question of how to wrap the edges, but the map generation model doesn't have to be. The concept of a polar region is integral to real world "earth-like" objects that evolve land mass configurations that subsequently require mapping. The map generation engine should think in terms of planets that have plate tectonics, climates, poles, and that evolve over long periods of time. Then, only when the planet is ready for people should the landmasses be projected to the gameworld as an arrangement of tiles. When this projection is made it might be interesting to expose the "polar locator" so that someone could chose to arrange the tiles for the game as if the north pole was at any grid location instead of the default "off the top edge" assumption the tile layout makes now.

In terms of wrapping the map, It would require 8 maps like the current gameworld, connected as the faces of a cube, in order to simulate the way a globe wraps a projection of planetary landmass using a square grid. But the arrangement of landmasses can also be adequately projected onto a single face of that cube(like the current game does), so the wrapping question becomes a purely gameplay decision. In that sense, it might be fun to have a "going off the west side of the top edge heading north" / "coming back on the east side of the top edge heading south" option in addition to the standard four wraps.... but it might also turn out to be alot of work for no added fun.

On a slightly different topic. Has anyone else noticed an east/west wrap bias to the maps generated for a north/south wrap game? It seems like inevitably the east/west no-wrap border cuts right through a big continent and most of the north/south wrap area is water. Since water currently hampers the AIs abilities, choosing a north/south wrap option would seem to work like a human exploit. I wonder how difficult it would be to turn the map generator's orientation 90 degrees when generating for a north/south wrapped world?

[J Bytheway]

Originally posted by drulius
In terms of wrapping the map, It would require 8 maps like the current gameworld, connected as the faces of a cube

A cube has 6 faces, not 8...

[Flinx]

Originally posted by drulius
In terms of wrapping the map, It would require 8 maps like the current gameworld, connected as the faces of a cube, in order to simulate the way a globe wraps a projection of planetary landmass using a square grid.

CTP2 models a cylindrical projection producing the equivalent of a world maps in an atlas.

[drulius]

Originally posted by J Bytheway
A cube has 6 faces, not 8...

D'oh ...stupid brain

Originally posted by Flinx

Originally posted by drulius
In terms of wrapping the map, It would require 8 maps like the current gameworld, connected as the faces of a cube, in order to simulate the way a globe wraps a projection of planetary landmass using a square grid.

CTP2 models a cylindrical projection producing the equivalent of a world maps in an atlas.

I was only speaking in terms of edge wrap behavior. Of course the projected tile arrangement of the game could not be duplicated by adding FIVE more fields to the playing surface. The game already places tiles in an adequate representation using only two dimensions.