Skip to main content

Rules engine for life simulation games

Project description

LiSE is an application for developing life simulation games.

Documentation

Development blog

What is a life simulation game?

Life simulation games simulate the world in relatively high detail, but not in the sense that physics engines are concerned with -- rather, each game in the genre has a different take on the mundane events that constitute everyday life for its purpose. Logistics and RPG elements tend to feature heavily in these games, but a lot of the appeal is in the parts that are not directly under player control. The game world feels like it continues to exist when you're not playing it, because so much of it seems to operate independently of you.

Existing games that LiSE seeks to imitate include:

  • The Sims
  • Redshirt
  • Rimworld
  • Princess Maker
  • Monster Rancher
  • Dwarf Fortress
  • Democracy
  • Crusader Kings
  • The King of Dragon Pass
  • Galimulator
  • Vilmonic

Why should I use LiSE for this purpose?

LiSE is a game engine in the sense of RPG Maker or Ren'Py. It assumes that there are certain problems any designer of life simulators will have, and provides powerful tools specialized to those problems. Though you will still need to write some Python code for your game, it should only be the code that describes how your game's world works. If you don't want to worry about the data structure that represents the world, LiSE gives you one that will work. And if you don't want to write a user interface, you can play the game in the IDE.

The LiSE data model has been designed from the ground up to support debugging of complex simulations. It remembers everything that ever happened in the simulation, so that when something strange happens in a playtester's game, and they send you their save file, you can track down the cause, even if it happened long before the tester knew to look for it.

Features

Core

  • Infinite time travel, rendering traditional save files obsolete.
  • Integration with NetworkX for convenient access to various graph algorithms, particularly pathfinding.
  • Rules engine for game logic. Rules are written in plain Python. They are composable, and can be disabled or reassigned to different entities mid-game.
  • Can be run as a web server, so that you can control LiSE and query its world state from any other game engine you please.

IDE

  • Instant replay: go back to previous world states whenever you feel like it, with minimal loading.
  • Rule constructor: Build rules out of short functions representing triggers and actions.
  • Rule stepper: view the world state in the middle of a turn, just after one rule's run and before another.
  • Edit state in graph or grid view
  • Edit rule functions with syntax highlighting

Getting started

Windows

# install the Kivy app framework
python3 -m pip install kivy.deps.sdl2 kivy.deps.glew
# install LiSE and the ELiDE frontend
git clone https://github.com/LogicalDash/LiSE.git
cd LiSE
git submodule init
git submodule update
pip3 install --user LiSE\ ELiDE\

Linux

# install the Kivy app framework
sudo apt-get install cython3 python3-dev python3-pip \
libsdl2-dev libsdl2-image-dev libsdl2-mixer-dev \
libsdl2-ttf-dev
# ELiDE doesn't play movies, so disable gstreamer
USE_GSTREAMER=0 pip3 install --user kivy
# install LiSE and the ELiDE frontend
git clone https://github.com/LogicalDash/LiSE.git
cd LiSE
git submodule init
git submodule update
pip3 install --user LiSE/ ELiDE/

Mac

# if you don't have https://brew.sh installed:
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
# if you don't have python 3 installed: 
brew install python pip3
# install the Kivy app framework
brew install cython
# ELiDE doesn't play movies, so disable gstreamer
USE_GSTREAMER=0 pip3 install --user kivy
# install LiSE and the ELiDE frontend
git clone https://github.com/LogicalDash/LiSE.git
cd LiSE
git submodule init
git submodule update
pip3 install --user LiSE/ ELiDE/

You could now start the graphical frontend with python3 -mELiDE, but this might not be very useful, as you don't have any world state to edit yet. You could laboriously assemble a gameworld by hand, but instead let's generate one, Parable of the Polygons by Nicky Case.

Make a new Python script, let's say 'polygons.py', and write the following in it (or use the example version):

from LiSE import Engine
import networkx as nx

with Engine(clear=True) as eng:
    phys = eng.new_character('physical', nx.grid_2d_graph(20, 20))
    tri = eng.new_character('triangle')
    sq = eng.new_character('square')

This starts a new game with its world state stored in the file 'world.db'. Because of clear being True, it will delete any existing world state and game code each time it's run, which is often useful when you're getting started. It creates three characters, one of which, named 'physical', has a 20x20 grid in it. The others are empty, and in fact we don't intend to put any graph in them; they're just for keeping track of things in physical. Add the following inside the with block of polygons.py:

    empty = list(phys.place.values())
    eng.shuffle(empty)
    # distribute 30 of each shape randomly among the empty places
    for i in range(1, 31):
        place = empty.pop()
        square = place.new_thing('square%i' % i, _image_paths=['atlas://polygons/meh_square'])
        sq.add_avatar(square)
    for i in range(1, 31):
        place = empty.pop()
        triangle = place.new_thing('triangle%i' % i, _image_paths=['atlas://polygons/meh_triangle'])
        tri.add_avatar(triangle)

Now there are thirty each of squares and triangles in the world. They are things, rather than places, which just means they have locations -- each square and triangle is located in a place in the graph.

The new_thing method of a place object creates a new thing and puts it there. You have to give the thing a name as its first argument. You can supply further keyword arguments to customize the thing's stats; in this case, I've given the things graphics representing what shape they are. If you wanted, you could set the _image_paths to a list of paths to whatever graphics. The 'atlas://' in the front is only necessary if you're using graphics packed in the way that the default ones are; read about atlases if you like, or just use some .png files you have lying around.

The add_avatar method of a character object marks a thing or place so that it's considered part of a character whose graph it is not in. This doesn't do anything yet, but we'll be using it to write our rules in a little while.

Now we have our world, but nothing ever happens in it. Let's add the rules of the simulation:

    @eng.function
    def cmp_neighbor_shapes(poly, cmp, stat):
        """Compare the proportion of neighboring polys with the same shape as this one

        Count the neighboring polys that are the same shape as this one, and return how that compares with
        some stat on the poly's user.

        """
        home = poly.location
        similar = 0
        n = 0
        # iterate over portals leading outward from home
        for neighbor_portal in home.portal.values():
            n += 1
            neighbor_home = neighbor_portal.destination
            # there's really only 1 polygon per home right now, but this will still work if there are more
            for neighbor in neighbor_home.contents():
                if neighbor.user is poly.user:
                    similar += 1
        return cmp(poly.user.stat[stat], similar / n)

    @phys.thing.rule
    def relocate(poly):
        """Move to a random unoccupied place"""
        unoccupied = [place for place in poly.character.place.values() if not place.content]
        poly.location = poly.engine.choice(unoccupied)

    @relocate.trigger
    def similar_neighbors(poly):
        """Trigger when my neighborhood fails to be enough like me"""
        from operator import ge
        return poly.engine.function.cmp_neighbor_shapes(poly, ge, 'min_sameness')

    @relocate.trigger
    def dissimilar_neighbors(poly):
        """Trigger when my neighborhood gets too much like me"""
        from operator import lt
        return poly.engine.function.cmp_neighbor_shapes(poly, lt, 'max_sameness')

The core of this ruleset is the cmp_neighbor_shapes function, which is a plain Python function that I've chosen to store in the engine because that makes it easier for the rules to get at. Functions decorated with @engine.function become accessible as attributes of engine.function. Every LiSE entity has an attribute engine that you can use to get at that function store and lots of other utilities.

If you didn't want to use the function store, you could just import cmp_neighbor_shapes in every rule that uses it, like I've done with the operators ge and lt here.

cmp_neighbor_shapes looks over the places that are directly connected to the one a given shape is in, counts the number that contain the same shape, and compares the result to a stat of the user--the character of which this thing is an avatar. When called in similar_neighbors and dissimilar_neighbors, the stats in question are 'min_sameness' and 'max_sameness' respectively, so let's set those:

    sq.stat['min_sameness'] = 0.1
    sq.stat['max_sameness'] = 0.9
    tri.stat['min_sameness'] = 0.2
    tri.stat['max_sameness'] = 0.8

Here we diverge from the original simulation a bit by setting these values differently for the different shapes, demonstrating an advantage of avatars.

Run python3 polygons.py to generate the simulation. To view it, run python3 -m ELiDE in the same directory. Just click the big > button and watch it for a little while. There's a control panel on the bottom of the screen that lets you go back in time, if you wish, and you can use that to browse different runs of the simulation with different starting conditions, or even stats and rules arbitrarily changing in the middle of a run.

If you'd prefer to run the simulation without ELiDE, though, you can add this to your script:

    for i in range(10):
        eng.next_turn()

Every change to the world will be saved in the database so that you can browse it in ELiDE at your leisure. If you want to travel through time programmatically, set the properties eng.branch (to a string), eng.turn, and eng.tick (to integers).

What next? If you wanted, you could set rules to be followed by only some of the shapes, like so:

    # this needs to replace any existing rule code you've written,
    # it won't work so well together with eg. @phys.thing.rule
    @tri.avatar.rule
    def tri_relocate(poly):
        """Move to a random unoccupied place"""
        unoccupied = [place for place in poly.character.place.values() if not place.content]
        poly.location = poly.engine.choice(unoccupied)

    @tri_relocate.trigger
    def similar_neighbors(poly):
        """Trigger when my neighborhood fails to be enough like me"""
        from operator import ge
        return poly.engine.function.cmp_neighbor_shapes(poly, ge, 'min_sameness')

    @sq.avatar.rule
    def sq_relocate(poly):
        """Move to a random unoccupied place"""
        unoccupied = [place for place in poly.character.place.values() if not place.content]
        poly.location = poly.engine.choice(unoccupied)
        
    @sq_relocate.trigger
    def dissimilar_neighbors(poly):
        """Trigger when my neighborhood gets too much like me"""
        from operator import lt
        return poly.engine.function.cmp_neighbor_shapes(poly, lt, 'max_sameness')

Now the triangles only relocate whenever their neighborhood looks too much like them, whereas squares only relocate when they have too many triangle neighbors.

You should make sure your rules have unique names. This requirement is necessary for assigning rules by name rather than decorator; you could make triangles move in response to dissimilar neighbors like so:

    tri.avatar.rulebook.append('sq_relocate')

In this case you didn't really have to use the name of the rule, since you still have the rule object in scope, but maybe you won't always. And, anyway, rules are evaluated in an order similar to alphabetical order, so having two rules with the same name would be unacceptably ambiguous.

When you have a set of rules that needs to apply to many different entities, and you can't just make them all avatars, you can have the entities share a rulebook. This works:

    sq.avatar.rulebook = tri.avatar.rulebook

And would result in pretty much the same simulation as in the first place, with all the shapes following the same rules, but now you could have other things in phys, and they wouldn't necessarily follow those rules.

Or you could build a rulebook ahead-of-time and assign it to many entities:

    # this needs to replace any existing rule code you've written,
    # it won't work so well together with eg. @phys.thing.rule
    @eng.rule
    def relocate(poly):
        """Move to a random unoccupied place"""
        unoccupied = [place for place in poly.character.place.values() if not place.content]
        poly.location = poly.engine.choice(unoccupied)

    @relocate.trigger
    def similar_neighbors(poly):
        """Trigger when my neighborhood fails to be enough like me"""
        from operator import ge
        return poly.engine.function.cmp_neighbor_shapes(poly, ge, 'min_sameness')

    @relocate.trigger
    def dissimilar_neighbors(poly):
        """Trigger when my neighborhood gets too much like me"""
        from operator import lt
        return poly.engine.function.cmp_neighbor_shapes(poly, lt, 'max_sameness')

    # rulebooks need names too, so you have to make it like this
    eng.rulebook['parable'] = [relocate]
    sq.rulebook = tri.rulebook = 'parable'

There are a variety of graph generators accessible on character objects, and convenience methods for common game actions like travelling along a path (Thing.travel_to) accessible on thing and place objects. The API for these isn't really solid yet, but tell me how you like them.

License Information

ELiDE uses third-party graphics sets:

reify.py is derived from the Pyramid project and carries its BSD-like license.

collide.py is ported from Kivy's garden.collider module and carries the MIT license.

The allegedb, LiSE, and ELiDE source files are licensed under the terms of the GNU Affero Public License version 3 (and no later). If you make a game with it, you have to release any modifications you make to LiSE itself under the AGPL, but this doesn't apply to your game code. Game code is that which is loaded into the engine at launch time from modules specified by the following parameters to the LiSE engine:

  • trigger
  • prereq
  • action
  • function
  • method

If you write another application (not using any allegedb, LiSE, or ELiDE code) that accesses a LiSE server via HTTP(S), it is separate from LiSE and not subject to its license.

In case of my death, I, Zachary Spector, wish for every allegedb, LiSE, texturestack, and ELiDE source file to be relicensed under CC0.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for LiSE, version 0.12.1
Filename, size File type Python version Upload date Hashes
Filename, size LiSE-0.12.1.tar.gz (159.8 kB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Huawei Huawei PSF Sponsor Microsoft Microsoft PSF Sponsor NVIDIA NVIDIA PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page