Skip to content

A simple text-rpg-engine than can run in a multitude of places

Notifications You must be signed in to change notification settings

kauffecup/text-rpg-engine

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

60 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

text rpg

Configuring

The game is entirely configurable via the JSON files under the data directory. There are example files in this repo to help you get started.

keys.json

Lists all keys. Keys are fairly simple objects that look like:

"key_door_02_03": {
  "name": "steel key",
  "description": "its handle is shaped like a book",
  "aliases": ["steel key", "the steel key"]
}

doors.json

Lists all doors. Doors look like:

"door_02_03": {
  "name": "the library door",
  "description": "the library door",
  "areas": ["area_02", "area_03"],
  "aliases": ["library door", "the library door", "library", "the library"],
  "locked": 1,
  "key": "key_door_02_03",
  "doorLockedText": "The door won’t budge, but you hear angry muttering from inside the door. Something about losing his keys. Maybe you can find them?",
  "doorOpenText": "Hooray! The archives are open for business! Let’s see what’s inside."
}

doorLockedText, doorOpenText, and key are optional. doorLockedText is custom text that prints when trying to open that locked door. doorOpenText is custom text that prints when the door is unlocked/opened.

items.json

Lists all items. Items are fairly simple objects that look like:

"item_coin": {
  "name": "Coin",
  "description": "This might come in handy later...",
  "aliases": ["coin"]
}

To add for more fun scenarios, you can specify a oneOfType. When multiple items have the same oneOfType, a player can only hold one of those at a time. When this is the case, you can also specify a oneOfTypeAliases array (that behaves the same as the regular aliases array) to allow for easier access. If there are a bunch of one-of-type swords, then a player can just say sword to get the job done.

Items can also be used as weapons. When doing this you can specify an attack property that acts as that item's attack. The syntax in the battle would be:

attack|hit|slap|punch|kick|smack {{monsterName}} with|using|on {{weapon alias}}

An item can also be used to revive someone who is dead. To do this, set revive:true in the item's json.

This can be used via:

revive|awaken|use {{item}} with|using|on {{player}}

monsters.json

Monsters describe a monster used in a battle:

"monster_scarecrow": {
  "name": "Scarecrow",
  "description": "A Scarecrow",
  "aliases": ["scarecrow"],
  "hp": 10,
  "attack": 3,
  "attackText": "{{monsterName}} slashes with its pole towards {{playerName}}!",
  "weakness": {
    "item_blue_wand": 2,
    "item_orange_wand": 2
  }
}

Note that monsters can optionally specify an attack, custom attackText, and weaknesses to given items. Thus if a player attacks with one of those items, that monster will receive more damage. This is currently implemented via a multiplier of itemAttack * monsterWeaknessToThatItem.

areas.json

Areas are where the meat happens:

"area_04": {
  "description": "a description",
  "lookText": "text thats printed when the user says look",
  "dialogue": {
    "progress": "area_04_01",
    "resetKey": "area_04_01",
    "conversation": {
      "area_04_01": {
        "description": "the first part of the conversation",
        "helpText": "help text for the first part of the conversation",
        "progression": {
          "area_04_02": ["user inputs", "that will", "progress the", "conversation"]
        }
      },
      "area_04_02": {
        "description": "the next part of the conversation",
        "helpText": "help text for the next part",
        "progression": {
          "area_04_01": ["etc"],
          "area_04_03": ["other", "things"]
        },
      },
      "area_04_3": {
        "description": "Oh noes! you're being attacked!",
        "helpText": "Try attacking the monsters",
        "battle": {
          "speed": 3,
          "monsters": {
            "monster_monster": 5
          }
        },
        "completeDrops": {
          "key_door_04_05": 1
        },
        "progression": "complete"
      }
    },
    "completeText": "text that prints when the user finishes the conversation",
    "completeHelp": "text that prints when the user types help when the conversation is done"
  },
  "inventory": {
    "items": {
      "item_coin": 10
    }
  },
  "doors": ["door_03_04", "door_04_05"]
}

Ok, so what's going on here. The description for an area is printed when the user enters that area. The look text is printed whenever the user executes any of the Commands.look commands. Areas are connected by doors and can have their own inventory.

Dialogue describes the progression of text puzzles/battles in a given area. The bulk of this description happens via the conversation object. Each key in this object points to a sub-object of the conversation. When that part of the conversation is triggered, the description text prints. If the user types help, the helpText prints for that given convo-part. Conversations can either be linear or branch - and all of this is described via progression. progression is itself an object - the keys point to conversation keys and the values are arrays of accepted commands. This allows you to describe "by typing one of x you go here and by typing one of y you go there." In turn you can have circular convos allowing for complex logic puzzles.

A sub-conversation object can also specify a battle. Battles have a speed property and an object describing what monsters are present. The progression for a battle can only be single valued - when you complete a battle this automatically puts the user in the next part of the conversation.

At any point you can specify that the user goes to the "complete" stage. In this stage the area is "done" and the completeText prints. completeHelp prints if the user types help.

All the "stock" strings used in the game are described in src/Strings.json, and all acceptable commands are defined in src/Commands.json.

Everything with an aliases array uses the regex under src/helpers/createRegex - even the commands.

Code Structure

Game

The main game entry point. Receives text and additionally functions from its adapter (described below). Maintains the player object, and current area object, handles any commands that involve moving stuff between them. For example, the user picking up an item, or the user unlocking a door, or the user traversing.

Also maintains the gameState object and updating it when the correct command is invoked.

The main function cycles through all available commands. If the user's input matches a command, it executes it. If not, it delegates the commands to the current area (which is of course an Area object).

EntityManager

The entityManager's job is to know what entities are available for doing stuff. Entities are loaded in via entityManager.load(type, largeJsonObject). It converts what's described in the data/* config files into JS objects. It also maintains a map of id -> object for fast lookup. This way things like areas, etc need only keep reference to ids of doors and ids of items.

The EntityManager is instantiated in main.js and passed by reference throughout the application instead of being a singleton. This is for easier unit testability.

Inventory

Maintains an inventory! Keeps track of what items are in it, what type they are, and has methods for adding/removing.

State

This class maintains the gameState. It keeps track of the visited areas, the current areas, the players, and all fun things like that. It uses the loadProm, save, clearSave, and entityProm methods passed in from the adapter.

Area

Areas maintain the current progress in the current dialogue, any items in the area, and an array of door ids that are in the area. All of this is described in the config file data/areas.json. The items are maintained in an Inventory class which is also used by Player.

Areas initialize a Dialogue when they are created. Dialogue maintains the logic for progressing through the conversation, and initializing a battle if there is one.

Battle

Right now the battle mechanics are rather simple. First, a speed property is specified in the areas.json. That speed dictates after however many player actions, a monster will act. Currently player actions are only attack and dodge and monster actions are target and attack. For example, if the speed is 1, the battle will look like:

player attacks monster
monster targets player
player dodges
monster attacks

Monsters attack players based on how often players have attacked. For example, if there are 3 players, and p1 has attacked 10 times, while p2 and p3 have each attacked 5 times, player 1 has a 50% chance of being targeted, while p2 and p3 have a 25% chance of being targeted.

Player

Maintains an Inventory for the given Player. Maybe one day this'll be more complicated.

Monster

Monsters are enemies that can attack players. The logic for the actual attacking is handled by Battle but each monster object maintains its HP, and attack power.

Door

A simple door object. Maintains its unlocked/closed/locked status. If it is lockable, it maintains which key id is necessary to unlock it. All doors are described via data/doors.json.

Key

A simple key object.

Adapters

This game is run via "adapters". The main class (src/main.js) exports two functions:

  • (default): takes three arguments input, userObj, respond. The input is, you guessed it, the user's input. The userObj is an object with four keys: _id, description, name, isAdmin that describes the user whose input we're sending to the game. respond is a function that takes text to output, and can be used however the adapter deems necessary.
  • initialize(save, loadProm, clearSave, entityProm): these are two functions and two promises that save, load, or clear the save data. Save takes an object that has four keys: players - an array of player data, doors - an array of door data, areas - an array of area data, and currentArea - a string representing the current area's id. loadProm and entityProm are a bit more fun - they're promises that resolves with either load data (the same object as the argument of save) or game data (an object containing areas, items, keys, and doors).

This allows adapters to be extremely customized. They must implement:

  • a way of retrieving user input and info and sending it to the game (input and userObj).
  • a function for outputting text (respond). if we're running via a terminal, this would be your basic standard out, if we're running in a chat client, it would be that client's send method.
  • a way to load in the default game data (areas, items, keys, and doors)
  • functions for saving, loading, and clearing save data

Implementing an Adapter

import main, {initialize} from 'main';
initialize(saveFunc, loadProm, clearSaveFunc, entityProm).then(() => {
 user.on('input', text => main(text, user, respondFunc));
});

Implemented Adapters

There are two adapters already implemented in this repo:

  • ./adapters/adapterLocal.js - runs locally in your command prompt
  • ./adapters/adapterSlack.js - runs as a Slack bot and is configured via environment variables.

There are two loadData-ers - ./adapters/_loadDataLocal.js and ./adapters/_loadDataDropbox.js, showing examples of how to load data into the game. The local one just reads in the JSON files from the ./data directory and the dropbox one, well, reads them from dropbox.

There are also two "savers" - ./adapters/_saveLocal and ./adapters/_saveCloudant that implement and export the three save functions. This way both the local and slack adapters can share these. ./adapters/_saveLocal creates 4 json files in a saveData directory, and ./adapters/_saveCloudant saves everything in a Cloudant database.s

Running and Testing

To run locally (after running a good ol' npm i of course):

npm start

To run on Slack:

npm run slack

To test:

npm test

To lint:

npm run lint

Environment Variables

The ./env module abstracts out the difference between running locally and running in da cloud (currently only configured for Bluemix). If you're purely running locally (using adapterLocal with _loadDataLocal and _saveLocal), then you don't need to set any environment variables.

When running locally with fun things, just make a ENV_VARS.json file in the main directory. It takes key value pairs so that will be read by the app.

If you're using the adapterSlack, you need to set:

  • SLACK_ADMINS: an array of user ids that can perform admin game functions
  • SLACK_CHANNEL_ID: the channel id you want to limit the bot to playing in
  • SLACK_TOKEN: your bot's slack token

If you're using the _loadDataDropbox, you need to set:

  • DROPBOX_ACCESS_TOKEN: an access token for a dropbox account containing the game files you need
  • DROPBOX_KEY: your bots dropbox key
  • DROPBOX_SECRET: your bots dropbox secret

If you're using _saveCloudant, you need to create a VCAP_SERVICES.json file in the main directory and just copy+pasta your vcap services from Bluemix.

Running on Bluemix

A manifest.yml is included in this repo, but note that it relies on a Cloudant service named text-rpg-cloudant to be bound to the node app.

Also note that it has no-route: true. This avoids the health check that'll crash the app. The app, when running in Bluemix (currently) only connects to Slack. Thus, it does not accept incoming connections except for Slack messages.

About

A simple text-rpg-engine than can run in a multitude of places

Resources

Stars

Watchers

Forks

Packages

No packages published