Editing a world map

[waldronate]

If you look at the lasso cursor, you'll see that it has a little "+" when Shift is down. If you release Shift and just use Ctrl, you'll see that the cursor has a little "-" next to it. If you use both Shift and Ctrl, you get a "x" next to the cursor. + is add to selection, - is subtract from selection, and x is intersection of existing selection with the new one that you're drawing (only keeps those parts in both the old AND the new selection).

See if this discussion offers some assistance:

[Deadshade]

If you look at the lasso cursor, you'll see that it has a little "+" when Shift is down. If you release Shift and just use Ctrl, you'll see that the cursor has a little "-" next to it. If you use both Shift and Ctrl, you get a "x" next to the cursor. + is add to selection, - is subtract from selection, and x is intersection of existing selection with the new one that you're drawing (only keeps those parts in both the old AND the new selection).

Well my problem was actually more stupid than that : I didn't realize that I had to draw a selection with ctrl around an existing selection (without crossing it) and then when the "ctrl" selection was closed, both disappeared. Alles klar now.

See if this discussion offers some assistance:

Oh by Jove this helped !
Now the mountains really look like mountains, connect continuously to plains and even the plains are no more so flat ! This is actually exactly what I have been after all this time.
Of course with magics there is always a price and in this case the price was that by feathering I lost the nice fractal coastline that took me 2 days to produce and import to Wilbur. But that price is Worth the result and I will somehow find how to reestablish the fractal coast because I have a selection with it.

A few minor questions though :
- I don't get the image you show at page 10. When I choose orthographic (or any other projection for that matter) , I only see a tiny a few pixels sized spot in the middle of the black background.
- I don't know what "tweaking" you did to the shader. Indeed teh "tweaked" version looks better.
- what is the general principle behind the numbers in the sphere parameters and the number in the scaling ? I still have no intuition what these settings will do.
- why does the "calculate heights" work with the fractal settings while using "fractal noise" doesn't ?
- I know that this is pretty minor but is to possible to get the ocean so that it is deep far from the land with only very large structures (here too much fractalising is not desirable). I found out that substracting a hybrid multifractal comes near but I just selected the size and scale settings randomly so can't reproduce it.

[waldronate]

The small-sized area is most likely because you didn't use File>>New to create the map area. You can get the same effect later by using Surface>>Map Info and setting Top=90, Left=-180, Right=180, and Bottom=-90. The Map Projection window is one of the few places that actually cares about the defined extents of the plane being viewed (in this case, it's the latitude and longitudes in degrees of the map boundaries).

The tweaking of the shader was mostly turning on absolute coloring and setting the maximum altitude value in the shader a little less than the maximum altitude on the surface (which forces the highest mountains to white) and setting the minimum altitude in the shader to a bit more less than the lowest altitude on the surface. I also put the sea color list to a single grayish color and set Opacity to 1 to get rid of the shading in the ocean.

The way that Wilbur generates altitudes is to slice a space-filling noise along some curve (at every XYZ point in space, the noise function effectively generates a hashed version of the coordinate, which results in an apparently-random but repeatable value). For spherical evaluation, this curve is the set of points along the surface of an ellipsoid centered at Sphere Center X, Y, Z with Sphere Radii X, Y, and Z; the portion of the surface that is evaluated is bounded by Spherical Area Top, Left, Right and Bottom. The central axis of the ellipsoid being evaluated is given by the polar direction Az, El (azimuth, elevation). I'm not sure when I last wanted the area or axis values. The evaluation is done on uniform angle steps in longitude and latitude, with those values being placed into an XY planar raster: this sphere to plane transformation is the definition of the equirectangular projection. For non-spherical evaluation, things are a little simpler: Position X and Y determine the center of the plane being evaluated, with Z being the vertical location in space; size is the size of the plane in X and Y (size Z isn't used). Az and El define the "up" direction for the plane. Again, I'm not sure when the last time I used Az and El on that dialog is.

The Filter>>Calculate Height Field is an older and more cumbersome interface to three features: Filter>>Noise>>Fractal Noise, Filter>>Noise>>Spherical Fractal Noise, and Filter>>Mathematical>>Calculate Function. They two interfaces should behave in the same way. If they don't, it's purely programming error. The Calculate Height Field feature is the original UI for calculations in Wilbur, back when I was of the opinion that generalized interfaces were easier to use. I did learn along the way that people seem to prefer simpler tools with fewer options. I learned this as much from my experiences with the brush interface as anything else. Originally, Wilbur just had a "brush" dialog for the UI on the paintbrush. All of those things on the "Paint Tools" toolbar were just selections on the dialog. People kept getting confused when I tried to explain what seemed to me to be a perfectly obvious dialog littered with controls that did different things in different modes. I added the Paint Tools dialog and people became excited about how suddenly easy to use this thing was. I was a little sad, but I did learn.

If you select the sea (land below 0) and add a mound that runs from a minimum of 0 to a maximum of some negative number (say, -1000), then things will get deeper as they get away from land. A little precipiton erosion or blur will tend to fill in the deepest creases pretty quickly.

[waldronate]

One way to keep most of your desirable crinkly coastline but still get smoother edges is to select the terrain from the surface (Select>>From Terrain>>Height Range between 0 and 10000000), use Select>>Modify>>Expand with a value of 3, Filter>>Fill>>Mound with a value of 300 or so, Filter>>Height Clip from -1 to 200, and finally offset by -100ish. It won't be perfect and you may need to adjust the amount that you modify the selection, but it may work better.

[Deadshade]

The tweaking of the shader was mostly turning on absolute coloring and setting the maximum altitude value in the shader a little less than the maximum altitude on the surface (which forces the highest mountains to white) and setting the minimum altitude in the shader to a bit more less than the lowest altitude on the surface. I also put the sea color list to a single grayish color and set Opacity to 1 to get rid of the shading in the ocean.

"Tweakings" like that show that this is like piano. Easy to learn the basics and to understand principles. But only years of practics allow to play Bach. Thanks, I found this great.

The way that Wilbur generates altitudes is to slice a space-filling noise along some curve (at every XYZ point in space, the noise function effectively generates a hashed version of the coordinate, which results in an apparently-random but repeatable value). For spherical evaluation, this curve is the set of points along the surface of an ellipsoid centered at Sphere Center X, Y, Z with Sphere Radii X, Y, and Z; the portion of the surface that is evaluated is bounded by Spherical Area Top, Left, Right and Bottom. The central axis of the ellipsoid being evaluated is given by the polar direction Az, El (azimuth, elevation). I'm not sure when I last wanted the area or axis values. The evaluation is done on uniform angle steps in longitude and latitude, with those values being placed into an XY planar raster: this sphere to plane transformation is the definition of the equirectangular projection. For non-spherical evaluation, things are a little simpler: Position X and Y determine the center of the plane being evaluated, with Z being the vertical location in space; size is the size of the plane in X and Y (size Z isn't used). Az and El define the "up" direction for the plane. Again, I'm not sure when the last time I used Az and El on that dialog is.

OK, I see what it's doing. I don't understand why somebody would want to change the center of the ellipsoid but I can probably live without understanding that. And the scaling ? For me that means to multiply a vector OM by some k. I guess the coefficient I enter is the k but what is O and what is M ? And what is the relation (if any) between this homothecy and the sizes of the ellipsoid ?

Also while I am at it what does "deterracing" do ? I thought it would be a way to get rid of cliffs and more generally of any vertical discontinuities but when I ran it on a selection, it stayed stuck at 0% progress for some 10 minutes so I cancelled it.

The Filter>>Calculate Height Field is an older and more cumbersome interface to three features: Filter>>Noise>>Fractal Noise, Filter>>Noise>>Spherical Fractal Noise, and Filter>>Mathematical>>Calculate Function. They two interfaces should behave in the same way.

So if it is the same thing why did you use in your tutorial above Calculate Height and not Fractal Noise ?

If you select the sea (land below 0) and add a mound that runs from a minimum of 0 to a maximum of some negative number (say, -1000), then things will get deeper as they get away from land. A little precipiton erosion or blur will tend to fill in the deepest creases pretty quickly.

This is a great idea. One needs really a screwed mind to find that because - 1000 < - 1 so that - 1000 can never be a maximum of that interval .

One way to keep most of your desirable crinkly coastline but still get smoother edges is to select the terrain from the surface (Select>>From Terrain>>Height Range between 0 and 10000000), use Select>>Modify>>Expand with a value of 3, Filter>>Fill>>Mound with a value of 300 or so, Filter>>Height Clip from -1 to 200, and finally offset by -100ish. It won't be perfect and you may need to adjust the amount that you modify the selection, but it may work better.

At what stage of the process should I introduce this sequence ? Just after the operation at page 7 (the Calculate Height Field) ?

EDIT :

Clearly it is not page 7 and I don't understand the why of the last 2 commands. I tried it as substitution to the feathering in the beginning but the clip created flat plateaus that survived (more or less) untill the end. As for the offset - 100 it just pushes the coast Under water so that the "green area" is reduced.

EDIT 2 :

There is something rather unpleasant happening with the memory management. I noticed on several occasions that when one works for a longer time with many undoings, resetting by using "New", reloading selections etc, strange artefacts tend to creep in. I can't nail it to something specific but the quality becomes worse and worse. For example with your tuto above, I was playing with different settings, reloading selections, undoing etc and at some point the selection contour stayed "engraved" in the map. Even doing deselect and running a fractal noise there still stayed a narrow but clearly visible trace of the previous selection which became a part of the texture of the map.
Empirically I found out that when I quit Wilbur and reboot it, everything becomes fine again. For a certain time . So I thought that I might as well tell you about this problem.

[waldronate]

If I understand the notation correctly, the spherical evaluation starts with a unit sphere centered at a given location. The sphere is scaled around that point by the radii ( k={Sphere Radii:X,Sphere Radii:Y,Sphere Radii:Z} ) and rotated by the vector direction az,el. Then the sector of the sphere indicated by top,left,bottom,right is evaluated. One reason that it's useful to change the center of the ellipsoid is that each center generates a new world with no changes to any other parameter. Animating a slow change in center gives what appears to be a set of world maps flowing smoothly into each other (the Wilbur Feature Window>>Journey Into Texture Space allows multiple frames of a linear translation along the degrees of freedom for the fractal parts of the Calculate Height Field operation).

Window>>Journey Through Texture Space seems to be fast enough these days to provide useful real-time feedback. Start Wilbur, use Window>>Journey Through Texture Space to bring up the UI. Click RGS to Read the Global Settings from the fractal features. Note that all fractal parameters are shown in pairs: the left-hand edit control for a pair is the starting value, while the right-hand one is the ending value. Set #Frames to 100, and the ending Sphere Center XYZ values to 0 and Click OK. Watch the sphere crawl through a diagonal in space from {-10,-10,-10} to {0,0,0}. An equally interesting journey is one that dilates the sphere without moving. Bring up the texture journey window again, click RGS, and then enter 0 the left-hand value for Sphere Radii XYZ and 20 for the right-hand side; click OK and watch the features get small as the sphere inflates through the constant-feature-size texture space. Turning off Spherical Evaluation, clicking RGS, and using the Z position as the animated parameter will generate planar noise that sweeps along the Z direction.

Deterracing is a feature intended to smooth out maps generated from layers of contours. At each point, it sends out rays to find a step in altitude and then interpolates the altitudes according to how far along the ray it is. The results are usually pretty bad and it doesn't work at all unless your surface is a set of steps. If you'd like to try preferentially smoothing areas that have large disparities in altitude, try this:

1) Texture>>Gray Maps>>Relative Elevation to compute a texture that has whiter values where altitude steps are greater.

2) Texture>>Transfer>>Texture to Selection using "Grayscale Combine" to transfer the texture to the selection, with whiter parts as more selected.

3) Filter>>Blur>>Gaussian Blur with a sigma of 2 to blur the surface, with the effect being greater at the parts that are more selected.


Note that I said that the split-up versions of the Calculate Height Field feature "should" have the same effect. In practice, they aren't quite there. Filter>>Noise>>Spherical Fractal Noise, for example, has no way to scale the fractal before applying it to the surface. Oops. I have spent a large amount of time working with the Calculate Height Field dialog and am used to it. The others were my attempt to make things a little friendlier in preparation for a major version upgrade. If I ever get the chance to finish it, I'll get the other parts fixed eventually (maybe).

The mound "minimum" and "maximum" would be more correctly named "edge" and "center". However, the original concept was to allow for mountains and I never got around to fixing the UI. You may have noticed this lack of consistency in a number of places around Wilbur.

The clipped mound would be a replacement for the feathered selection. I forgot to mention the Texture>>Gray Maps>>Height Map to get the grayscale selection and Texture>>Transfer>>Texture to selection steps in there, too. It might be better to ignore that suggestion. The idea was to get a selection with sloped edges that approximated your original selection better, but it's pretty hard to get a good result.

Wilbur uses what I'll describe as an "optimistic" memory management strategy. That is, it hopes that nothing will go wrong. Memory can get fragmented and fill up quickly, especially on large surfaces and using the 32-bit version. My machine is running the 64-bit version of Windows and has 32 GB of RAM, so I forget what it's like to have those memory problems (which is why they don't get fixed). One limiting factor is the amount of undo space allocated. Under Edit>>Preferences, setting the Undo Levels value to 4 or 2 can reduce your memory congestion issues at the cost of limiting the number of undo levels.

[Deadshade]

The sphere is scaled around that point by the radii ( k={Sphere Radii:X,Sphere Radii:Y,Sphere Radii:Z} ) and rotated by the vector direction az,el. Then the sector of the sphere indicated by top,left,bottom,right is evaluated.

Most of this I don't understand. For me the x, y z values were just coordinates of the sphere radius (like in (X-X0)² + (Y-Y0)² + (Z-Z0)² = R²). Then there is another thing - the scaling button. It only asks for a "scaling factor" and my question was what that "scaling" factor scales. If it scales a spherical surface then it multiplies R² (the squared radius) by some constant. I tried to experiment with it but couldn't see what effect it was having. aZ is surely the normal to the sphere in some point A but I have no idea what eI means and even less how or why a sphere would rotate around some vector (presumaby aZ + eI ?). Of course I realize that it's hard to put in words what a line of equation would say clearly.

Wilbur uses what I'll describe as an "optimistic" memory management strategy. That is, it hopes that nothing will go wrong. Memory can get fragmented and fill up quickly, especially on large surfaces and using the 32-bit version. My machine is running the 64-bit version of Windows and has 32 GB of RAM, so I forget what it's like to have those memory problems (which is why they don't get fixed). One limiting factor is the amount of undo space allocated. Under Edit>>Preferences, setting the Undo Levels value to 4 or 2 can reduce your memory congestion issues at the cost of limiting the number of undo levels.

Whan I was talking about "memory" management, it was more an image. What I wanted to say was that when one does much of editing, undoing etc, graphical artefacts finish by appearing (like a selection contour becoming part of the texture and staying through all operations) and the quality of the image gets steadily worse. I have a 64 bits system and 10 GB of memory and don't think I have a memory problem per se. That's why I don't need to set the undo limit at less than what it is.
It rather feels like small bits of garbage slowly accumulating in time and once they manifest with too visible bad image parts, to get rid off of them one has to quit Wilbur and start anew.
Next time I get one of these anomalies, I will make a SS.

[waldronate]

Most of this I don't understand. For me the x, y z values were just coordinates of the sphere radius (like in (X-X0)² + (Y-Y0)² + (Z-Z0)² = R²). Then there is another thing - the scaling button. It only asks for a "scaling factor" and my question was what that "scaling" factor scales. If it scales a spherical surface then it multiplies R² (the squared radius) by some constant. I tried to experiment with it but couldn't see what effect it was having. aZ is surely the normal to the sphere in some point A but I have no idea what eI means and even less how or why a sphere would rotate around some vector (presumaby aZ + eI ?). Of course I realize that it's hard to put in words what a line of equation would say clearly.

Ah, what does the Scaling button do. That's much simpler and has nothing to do with where the fractal function is evaluated. After the fractal function is evaluated, it performs the operation indicated by the Scaling dialog. The default "Single Value" scaling operations multiplies the scalar altitude by the scalar value given in "Scale Factor" before blending it with the existing surface value. You wouldn't normally need to use other the other kinds of operations, but "To Range" should scale the fractal function to the highest and lowest values indicated in the dialog, and Broken Range scales by high if the result is greater than or equal to Break, and by Low if the result is less than break. The "Broken Range" thing dates way back to the dawn of time when Wilbur didn't have selections.

Whan I was talking about "memory" management, it was more an image. What I wanted to say was that when one does much of editing, undoing etc, graphical artefacts finish by appearing (like a selection contour becoming part of the texture and staying through all operations) and the quality of the image gets steadily worse. I have a 64 bits system and 10 GB of memory and don't think I have a memory problem per se. That's why I don't need to set the undo limit at less than what it is.
It rather feels like small bits of garbage slowly accumulating in time and once they manifest with too visible bad image parts, to get rid off of them one has to quit Wilbur and start anew.
Next time I get one of these anomalies, I will make a SS.

I'm glad to hear that it's not outright crashes. I hate it when that happens.
I don't think that I've ever seen the sort of visual corruption that you describe. There may well be some areas of the code that aren't saving everything for each undo operation and an undo might then pull things from an unintended buffer.

[Deadshade]

I'm glad to hear that it's not outright crashes. I hate it when that happens.
.

No from this point of view it is remarkably stable. I had not a single crash since 1 month when I first downloaded and installed Wilbur.

I also realized that I was wrong to tend to ask the kind of questions "How does this or that work ?".
As the natural and easiest language for that is the analytical geometry, everything would be simply and crisply expressed with only functions f(x,y,z), their partial derivatives and translation, homothecy and rotation operators acting on them.
Words are just too messy and lead to misunderstandings more often than not.
Besides it is not necessary anyway.

I noticed that I made the biggest progresses when I understood what this or that command visually did and I didn't need to understand how the inner wheels worked.
So when one doesn't intend to work on the mathematical models themselves, it is relatively irrelevant how they are constructed. And that would be the case of almost all users even if they had the mathematical training allowing to understand the inner workings. So the right kind of questions that I should ask is "What changes in the image when I change this or that parameter ?"

Some rather detailed small questions.
- I have been looking for map elements like cartouches, map borders and scale indicators. There are a few in Terraformer that I got with FT3 but in the tutorial is an announce of update (e.g more of such elements). As this was already some years ago, I wanted to find Bill Roach but no amount of Googling provided any adress or web site. Do you know if he is still in the loop ? I have also edited one of his cartouches (which are rather messy apparently scanned hand painted things) and wanted to send it to him.
- Singularily I found no place where I could find (free) scale indicators in fantasy style. Tons of sites for fonts but none for scales. The thread here that gives links for map symbols leads either nowhere or to irrelevant icons. Would you happen to have a link ?
- The same question for cartouches. I have at least the few provided with Terraformer which are in the "right" style but I would like to have a larger choice preferrably without the imperfections due to scanning hand painted images.

[waldronate]

The "Journey Through Texture Space" feature is an excellent way to answer questions of the form "What changes in the image when I change this or that parameter ?" for the fractal functions. You can set a starting and ending value for some parameter and let the system show you what happens as that parameter changes.

I'm not sure how Bill is doing these days. He reported having some health issues the last few times that he posted over on the ProFantasy forums and may just have other concerns at the moment. I don't have any contact information, sorry.

I don't know of good sources for the map elements that you're describing. There may be some here at the guild, but I don't really pay that much attention to such things because that part of mapmaking falls under "artistic" and is well outside of my skill set. You can search for classic mapmakers such as Blaeu or Mercator and look at their maps for elements that are probably in the style of interest. They will still have the problem of being scans, though.