Any interest in a RMG?

[llamabeast]

Well whether or not most people learn how to use them, a few of us can make some really good maps using it and get them widely used. Wraparound maps in particular are in very short supply.

Actually I think I might be making a similar point to Gandalf.

[Agrajag]

Re: Gandalf Parker
That's a good idea, and it might be a motivating thought that even if no one downloads my potential RMG, people will still enjoy maps made by it.
And there's no reason why it couldn't work without a GUI. I'm tending towards a config file more than command line, because it is much easier to assume that the config file is correct (and much easier to mess around with).
One thing I'm not sure about, though, is if it will be possible to generate many maps every midnight, because going through the entire process of making a fully playable map might take a lot of time. (Especially that part where borders are calculated and drawn, because the algorithm I have in mind might be quite slow [it would go through each pixel on the map at least 4 or 5 times])
I'm also not very familiar with threading, so no taking advantage of multiple cores
And of course, sadly, you can't automate anything yet seeing as how the server is dead
Re: Llamabeast
Are you sure you are making a point?
And do you really think you could transform a random map into a really good map? Would you even need all the stages involved? (Or would you only need a base image to prettify and work on)

[Gandalf Parker]

Servers are perfect for such things. Projects which can take a full day of computing would be fine. Im not tying up my desktop or slowing my usual activities by having it do some major project in the background all day. Ive often wished that more projects actually took that into account. Generators often give up excellent actions because it makes the generation time too long for anyone to put up with. PbEM style games (like Dominions) often cut out some options because the devs feel that most players would not put up with the time it takes (such as extensive AI thinking). I cant really argue their logic but often I wish it stayed in at least as an option. After all its not like the community here NEEDS 30 maps daily in each of tiny, small, medium, large, huge, mega sizes. Even if I only generated ONE map a day it would be a huge boost.

One example is that the DomMap generator was forever without a blur routine even though it made the maps much easier on the eyes. I wrote a script to take the generated maps, save the white pixels, blur the map, restore the white pixels, and resave. It was very time intensive but added a nice new optional group to the maps.

Oh and only one of my servers went down. The main one. The old one (which originally housed Dom2minions) is still up and still running games. I could shift everything to it but I dont expect the main server to be down that much longer.

[lch]

Zentar said:
I think there is already a RMG out.

The RMG that went the furthest so far is the one from paradoxharbinger, who hasn't posted about it since a long time, half a year at least. So ongoing development is unclear.

Gandalf Parker said:
Excellent. I find that RMGs tend to die.

Which is why I'm not really interested in anything where the source isn't being released. Projects die because people commit their time to other things, that's normal. But it would be good if somebody who wants to tackle the problem can have a look at how it was done before, even if just as a reference, instead of re-inventing the wheel yet again. This seems to happen rather often with things here, though.

What I'm putting my money on right now is the announced map generator from Snoddasmannen. As I wrote before, I have some code bits and pieces myself, but I lack the time and commitment to come up with a finished end product, so I'd rather code dive into finished work from somebody else.

[Gandalf Parker]

lch said:

The RMG that went the furthest so far is the one from paradoxharbinger, who hasn't posted about it since a long time, half a year at least. So ongoing development is unclear.

Yes he was just getting to the point of it being extremely useful. He had licked the biggest complaints. He had just made colors fully selectable, and added the ability to scatter icons as the Dom3 generator does. A user-created icons file would be a big addition.

He also had incorporated requests of mine to include data in the .map output that the map generator "knew" which map modifiers had trouble digging up later. Like this..
-- #37, {1642,82}
-- neighbor 32, 49, 65, 879, 897
-- medium
-- swamp
-- forest
and we were working on adding recognition of penninsula, lake, island, etc. Also a breakdown of the bit-code to make it easier for math-incapable people like me to make changes. Altho it was beginning to look like extensive documentation of a map file might be better done as a seperate info file or maybe even a seperate program entirely.

Which is why I'm not really interested in anything where the source isn't being released. Projects die because people commit their time to other things, that's normal. But it would be good if somebody who wants to tackle the problem can have a look at how it was done before, even if just as a reference, instead of re-inventing the wheel yet again. This seems to happen rather often with things here, though.

Yes, the DomMap for Dom2 went thru 5 major revisions at least by different people since it was readily available source code (and still is). Even I was able to make some changes in it without knowing C.

I tried to get someone to put it up as an official SourceForge project and it probably would have but Johan's announcement that Dom3 would have a generator kindof killed it. Too bad since it was moving in directions that were diffrent than the Dom3 one ended up doing.

What I'm putting my money on right now is the announced map generator from Snoddasmannen.

Hmm I have to catch up. I havent looked into that one yet.

[lch]

Gandalf Parker said:
Yes, the DomMap for Dom2 went thru 5 major revisions at least by different people since it was readily available source code (and still is). Even I was able to make some changes in it without knowing C.

Yeah, I've seen that code. It is horribly inefficient at some places. Since it definitely looks like what the Dom3 map generator produces, I fear that some of the inefficient things went verbatim into the Dom3 one. But I haven't really checked on that. Just a hunch that it might be right. Try creating a 10 provinces map in resolutions 100x100 and 2000x2000, for example, and then time how long the different steps during map creation take to complete. In theory, there shouldn't be that much of a difference, but there is.

Gandalf Parker said:
I tried to get someone to put it up as an official SourceForge project and it probably would have but Johan's announcement that Dom3 would have a generator kindof killed it. Too bad since it was moving in directions that were diffrent than the Dom3 one ended up doing.

Not so much killed it, but rather absorbed it. And since the Dom3 map gen did generate better / prettier output, people dropped it. It's still there, people can still pick it up.

Gandalf Parker said:

Hmm I have to catch up. I havent looked into that one yet.

As I said, announced. Nothing visible yet. Let's wait.

[Snoddasmannen]

lch said:

As I said, announced. Nothing visible yet. Let's wait.

Don't panic, I have been working on it. One of my modus operandi is getting stuck in tiny details which aren't very important. Right now I'm struggling with province borders around coastal areas. One of the things I'm sure somebody like lch would fix up in five minutes if I released the source

Tell you what, I'll try to focus on getting something which compiles, runs and generates a map. Then all of us who feel like it can work on the fun stuff together.

I'm away from home for the weekend visiting with my gf's family, but I will see if I can post a thread with a screenshot when I get back. I don't want to hijack another thread

[Agrajag]

Oh please dont make me guilty of blowing you away. The other rmg appears to have stopped in development and you were on such a roll. But on the other hand I would be the last person to recommend that anyone not pursue the parts they feel are fun.

So far I'm still having fun with it, and I'd dare say I'm not at all far from getting it to a usable beta state.
It can currently generate maps, place provinces and draw borders. It still needs some improvement on the last two until they are correct. After that there's actually generating the .map files (which shouldn't be hard, as I already store information of all neighbours and the terrain bitmask.).
After that, all that's left is making everything customizable rather than having almost all values hard-coded, and making the maps look graphically prettier.
Then I can release version 1.00 and start improving it based on suggestions

Oh, and if you feel like blowing new wind into my sails, then you are doing so with your inclination to host some maps generated with this RMG on your site.
This is because I'm about to add a few things which will be quite computationally expensive

Well it works for me anyway you want to do it. As for the colors.. Old Dom 1, and DomMap, both went with speckling the province with color. So one possible choice would be assigning a color to each basic terrain, and a setting for variation. Such as brown to mountain, and yellow to waste, then have the program speckle the two colors together for provinces which were mountain-waste. If the variation variable is set to 3 then the game could randomly use -3/+3 of the brown for each mountain pixel and -3/+3 of yellow for each waste. That would give a nicely blurred effect. Not so stark. And provide some effect provinces which are entirely mountain or entirely waste.

I saw it and I didn't really like it. I'm more inclined to have a single color per basic terrain and have the resulting color just be an average of all colors that apply (that's what it's like right now, except all values are hard-coded). The real variation should come by also assigning an image (or images) to each terrain type (like little mountains and trees) and have the program spread all applicable images on the terrain.
(Of course, there's always the option of providing more than one option )

I don't want to hijack another thread

I'm actually quite interested in what you have, since in a way you are a competitor

[lch]

Snoddasmannen said:
Right now I'm struggling with province borders around coastal areas. One of the things I'm sure somebody like lch would fix up in five minutes if I released the source

Oh, don't get your hopes up too high - I'm usually suffering from the "too many things at the same time" syndrome. But yes, somebody might help and improve the code for you. At least everybody would have the chance to do so, then.

[lch]

Gandalf Parker said:
He also had incorporated requests of mine to include data in the .map output that the map generator "knew" which map modifiers had trouble digging up later. Like this..
-- #37, {1642,82}
-- neighbor 32, 49, 65, 879, 897
-- medium
-- swamp
-- forest
and we were working on adding recognition of penninsula, lake, island, etc. Also a breakdown of the bit-code to make it easier for math-incapable people like me to make changes. Altho it was beginning to look like extensive documentation of a map file might be better done as a seperate info file or maybe even a seperate program entirely.

I'd really separate those things into a good, fast, efficient map generator and a .map interpreter / commenter. No need for yet another complex unhandy bloat program.