User Tools

Site Tools


Technical Information About Maps

This documented is intended to convey technical information on how Crossfire deals with the map objects and objects placed on the maps.
For the most part, I only intend document how the new code works, and not go too much into history on the older methods.
A lot of the map code was re-written in early July 2001, which changed how many things are dealt with.

Mark Wedel July 7, 2001

1. Map Header

The map header is the section at the start of the map file that describes the maps characteristics. The values are described below.

The map variables now make some sense, and are only stored in the map structure itself. I still include the old value (the 'was') so if you are looking at old maps, you know what they mean. Generally speaking, the values in the map files themselves match the same element name in the map structure.

'width','height', was 'x','y': Size of the map.

'enter_x', 'enter_y', was ('hp','sp') = (x,y) of the destination on the new map. These are only used if the exit does not have a specific location set.

'reset_timeout', was 'weight': stores the number of seconds that need to elapse before this map will be reset. Ie, if 1800, it means this map expires after 30 minutes. This value is not modified once loaded - instead reset_time is used to track this. The value 0 means to use a default timeout (MAP_DEFAULTRESET).

'swap_time', was 'value': This controls how many ticks must elapse after the map has not been used before it gets swapped out. Swapping out is different than reset, as a swapped out map will get loaded back into memory if someone re-visits it before it is due to reset.

'difficulty', was 'level' stores the map difficulty. If not set to anything, the server code will try to come up with some difficulty value.

'fixed_resettime', was 'stand_still': If nonzero, the map reset time will not be updated when someone enters/exits the map. Thus, once the map has been loaded, it will reset in 'reset time' no matter what access happen. This is useful for shops and towns, which are constantly accessed, but should be reset periodically.

'darkness', was 'invisible'. Light/darnkess of map (overall). If 0, all of map is fully bright.

'unique' - if set, this entire map is unique. Exactly unique to what will depend on how it was created (it could be a per player unique map, or maybe a common map that is just permanent for all the players.

'nosmooth' - if set, no faces in this map will be smoothed.

'outdoor' - if set, this is a hint that this is an outdoor map. If this is not set, weather and dawn/dusk will not occur on this map. It is highly advised that this be set appropriately.

tile_path_<x> - Used with map tiling. <x> is a number, 1 is north, 2 is east, 3 south, 4 west. This determines what map is tiled in that direction. See the section below for more information about map tiling.

'shopitems', 'shopgreed', 'shoprace', 'shopmin', 'shopmax' - the type of thing the shop trades in, see doc/Developers/shops for more details

'temp' - The base temperature in Fahrenheit for this map. The temperature is modified by the season and weather conditions. In a map without weather effects, this temperature will be used as the static temperature for the entire map. This can be useful to make an ice cave actually cold.

'pressure' - This should really never be set on a map. The pressure in millibars.

'humid' - Again, should rarely be set on a map. The humidity in percent.

'windspeed' - Rarely set. The windspeed in kph/h.

'winddir' - Rarely set. Direction of wind, 1-8, 1 is north, clockwise.

'sky' - The sky conditions for this map. See weather.h. Don't set this unless you really know what you are doing.

2. Map Objects

The objects within the map are saved in standard 'object save' form (same as is used for the players objects). Other files document the actual meaning, but the general form is:

arch <some name>
x <some value>
y <some value>
<other object specific values>

Note that x and y are in fact optional. If not present, the values default to zero.

Multipart objects

Multipart objects pose a tricky problem, in that they have to appear together in the map file - this makes proper handling of layers hard to deal with.

In old map code, all the single spaces objects were saved, and then all the multi part objects were saved. This effectively means that the multi part objects always ended up on top. The multipart objects were saved with all their parts. For example:

slaying shops/magicshop
hp 14
sp 14
x 1
y 13
arch store_magic_2
name Magic Shop
slaying shops/magicshop
hp 14
sp 14
x 2
y 13
<snip - there are really two more parts>

This method does not work very well with the map tiling however (how do you reasonably deal with a monster that may be straddling the two maps?) Current code now only saves the head of the object. When the map is loaded, the objects on the map are examined to see what objects need to have more objects added on. Additional parts linked in are put just above floor level when linked in, so things like shops won't hide items that someone drops on them. For monsters, this linking shouldn't be a problem - once they start moving, they will get relinked as normal (on top).

The effect of saving only the head does have the effect of not being able to customize the non head parts of the object. This generally should not be a problem (in the case of shops/building, the exit code now knows to look only at the head for valid information). The case where this may not work as well as expected is for buildings where setting the move_block to non-archetype defaults will get lost.

3. Map Tiling

Map tiling is a feature that lets multiple maps be connected, effectively forming a much larger single map. This is most useful for the outdoor world maps, where it is not practical to have on massive map, but the old style tiling method (copying areas of the adjoining map to the next one) are not very efficient.

The transfer of objects from one map to another tiled map are automatic. Presuming the proper macros are used (out_of_map, get_map_..), minimal extra work is necessary for everything to work right.


Tiled maps must be the same width/height along the side they are tiled with. If map1 has a height of 15, and you want to tile along one of the sides, the map(s) it gets tiled with along that side should also be 15. Given the following diagram (not to scale):

|         |         |
| map1    |  map2   y2
y1        |         |
|         |         |
x1 is the width of map1, y1 is its height.
x2 is the width of map2, y2 is its height.
map1 will tile map2 as indicated in the above diagram.

Given that, the following must be true:

y1 must equal y2
x1 must be greater than 12
x2 must be greater than 12
x1 and x2 do not need to be equal

The value is derived as being half the maximum viewable area. The reason for this is that the line of sight code (and likely some other code) will only look one map away from a source coordinate. While the values can be less than 12, they should be at least 12 if the map tiles with another one in that direction. If the map is an 'end' map (ie, no further tiling in a specific direction), then having a value less than 12 should work just fine.

Note that tiles maps do not have to be symmetric - several maps could tile to a common map. That common map can only tile back to one of those. And example of where this might be used is for a courtyard of a multi floor house - that courtyard should be visible (and be the same) from all the levels, but you can only go from the courtyard to first floor rooms off the courtyard. This may not be ideal (ie, if flying, you should be able to go to any floor), but this tiling for elevation is just an example that can be used.

4. Stacking of Objects on the Map

The stacking of objects on a map is somewhat confusing. This is because there are really two stacks a designer may care about - the actual stacking of objects on the map (as op→above points to) and the stacking of objects as they appear to the client.

It is usually the appearance in the client the designers care most about. However, actual stacking in the code can be important - some objects only look for objects above them - if you put a teleporter above a sword, that is different than if you put the teleporter below the sword.

When loading a map from disk, the server will keep the same stacking order as set in the map. Note however that multipart (big) objects are handled specially - because only the head of the object is saved, the stacking for the non head objects is always right above the floor. So if you put a troll down, with its head above a sword, when the rest of it is linked in, those other parts will be below any items on the ground.

Additional objects put on the space during play will generally be in stack order - newest object is on top. The one real exception is that flying objects (spells typically) will always be put on the top of the space, and then followed by non flying objects. So if a sword is dropped on a space with a bunch of active spell objects, within the server stacking, the sword will be below the spell objects, but above other objects on the space.


This ordering of objects in the server has little relation to how things appear in the client. With the client (and map2 protocol command), there are 10 visible layers, displayed in following order (from bottom up):

Floor (by definition, nothing beneath a floor is visible)
no_pick (2) - things like signs, savebeds, runes
item (3) - swords, armor, scrolls, arrows, etc
living (2) - monsters
fly (2) - spells, arrows, some monsters

The number in parentheses are the number of layers for those type of objects. So while there are 10 visible layers, in most cases, at least a few will not be used. If there are 50 items (swords, bows) on a space, the client will still only be sent information on 3 of them. It will not use the other layers, as only items will be sent in the items layer.

In most all cases, the server automatically figures out what layer an object belongs to based on various flags (FLAG_IS_FLOOR, FLAG_NO_PICK, etc).

An object/map maker can override these values with the 'map_layer' object parameter. For example, if a map maker wanted to make what appears to be a flying statue, they would just put in 'map_layer fly' in the object definition. The names listed above are the correct names used for the loading of objects.

There can be other cases where map maker want to change layering. A current case is levitation boots - because they seem to fly, they would normally always be drawn on top of other objects. The archetype sets the map_layer flag to item, so that they will be drawn with the rest of the pickable objects.

In cases of there being more objects of a type than can be displayed on a space, the first fallback is to look at the visibility of the defined face. Visibility is a face, not object, attribute, but is used when certain objects should take viewing precedence over others. An example of its use is buildings and exits - you pretty much always want those visible, even if there are other non-pickable objects on the space (or a player casts some runes).


If the visibility for the face is the same, the last fallback is server (map) stacking. In this case, the object on the top of the stack will take precedence of objects further down. Note however that visibility is always used to put important objects higher on the space as far as the clients sees them. For example, suppose we have the following space with objects with the hypothetical visibilities:

mace (50)
sword (50)
bow (100)
shield (50)
floor (50) (this is included just to make stacking order clear).

One might look at that and think that since shield is at the bottom and doesn't have high visibility, order would be mace, sword, bow, floor (shield getting dropped).

But that is not the case - the high visibility of the bow causes it to be virtually placed above the other objects in the map view, so actual view would be bow, mace, sword, floor.

Because visibility is not an object attribute, there is generally little that can be done to control this, other than to set up the appropriate stacking.

5. Smoothing

Author Tchize Date: Seventh of July 2003

Introduction to Smoothing

The crossfire graphical interface and internal map handling relies on a map made of square. No object can lies between squares. A typical square has the size of a standing player or other humanoïd sized creature (Goblins, Orcs, Gnolls, Dwarfs, Elves, …). This lead to an awful interface concerning terrains transitions (Sea shores, road borders, mountains, a.s.o)

There are 2 ways to get rid of this problem:

  • Suppress the square by square behavior of map handling. This means rework half of Crossfire code and redraw all maps …
  • Use some magic illusion.

What is Smoothing

Large discussions in the cf-devel mailing list was around a document on terrain transitions available on []. It explains a way of handling terrain transition in cases like the one of Crossfire.

Consider the smoothing in Crossfire as the ability for an object to draw itself partially above the surrounding squares. For example, grass could overlap just a bit the road next to it, a mountain would not end abruptly but instead would have some hills above the neighbor sand.

How to Smooth?

Next of this document is divided in 2 parts.
Drawers should be interested in the first part “Basic Smoothing” ,
while developers may be interested also in Smoothing in the Code .

Basic Smoothing

Basically, there need to be some order for smoothing. If everything draws itself above every surrounding square, we would end up with some awful bunch of crappy disgusting mixture of color (a Picasso or alike).

So, we come to the…

First Rule of Smoothing

An object O may draw itself above a neighbor object V if, and only if, O has a «smoothlevel» higher than V. See the things like that:

smoothlevel.eps FIXME: Find image

O in the example picture has a higher priority than V so O will overlap V a bit but V will not overlap O.

The smoothlevel is an integer value ranging from 0 (overlap nothing) to 255 (overlap above everything except other objects having also smoothlevel of 255). This is important to consider when you are going to give the smoothing ability to new objects. Let's say, for example, you want to add this ability to mountains. Let's imagine has a smoothlevel of 10 and trees have a smoothlevel 100 (just imagine, at time writing this doc, nothing fixed yet). Now let's say you want the mountains above the grass but, for graphical reasons, you want the trees above the mountains. You Choose a value between 10 and 100, not included. But mountains is a thing that goes above a lot of thing (grass, hills, swamps, brush, rocks, desert,…). So to keep the place for adding new elements in the future, you rather choose a high value for mountains (90 seems a good value according to these conditions).

So now you have chosen the smoothlevel for the mountain. What are you supposed to draw? You know what overlap what is decided by smoothlevel. But the graphics used during the smoothing is the…

Second Rule of Smoothing

The picture used to draw a given object depend on the face of the object. All objects using the grass face, for example, if they have a smoothlevel higher than 0, will be smooth using the same pictures. So you will bind with the picture grass.111 a set of pictures telling the client what to draw when smoothing an object having the face grass.111.

Take a look at the canvas below:

canvas_smooth.eps FIXME : Find image

Each square have a size of 32×32, which is the size used by the current image sets. This canvas handles all cases of corner, borders, L shaped, U shaped and O shaped transitions. The picture is made of 16 elements by 2. The first line is all border related transition. The second line is corner related transitions. Here is an example picture for grass (yeah really need someone to rework it):

sgrass.base.111.eps FIXME : Find image

The picture is named sgrass.base.111.png and located in the ground/smooth/grasslike directory of archetype. Ok, now you have a picture for smoothing grass, you have a smoothlevel for grass. There is only one last thing to do:

Final Rule of Smoothing

Tell the server which picture to send the client. There is, in the server SVN directory a file named lib/smooth. This file is where you set this information. Each line in the file beginning with # is a comment. Other lines should be made of 2 elements:

The first is the picture you want to smooth (here it is grass.111), and the second is the 512×64 picture drawn at rule 2 and used when smoothing.

  • Note Please note we suppressed the .png and the .base. elements of grass.base.111.png filename. Since what the server needs is the internal names. Take care of this. If the new picture you drew didn't appear on the GUI, this is perhaps due to an error in this file. And don't forget to copy this file, afterwards, to the install directory of server, using a make~install.

And example line might be as follows:

grass.111 sgrass.111

The entry default smoothed.111 is used when the server can not find an appropriate entry for an object.
This correspond to the black canvas above so mapmakers or drawers can see immediately,
that there is a problem.

Now, drawers know everything. Coders might want to read following part.

Smoothing in the Code

Back to How to Smooth?

Guess what…


maps/dev/map-technical.txt · Last modified: 2018/04/03 12:20 by karl