gavmor / dunmanifestin Goto Github PK

We should come up with a more general solution for phrase templating that will let us handle plurals and agreement.

I think the best way to handle this is for phrases to be objects. A phrase's to_s method will interpolate some other to_sed phrases into a string, like:

class Phrase
  def method_missing meth, *args, &block
    instance_variable_get "@#{meth}" or
    instance_variable_set "@#{meth}", meth.capitalize.constantize.new(*args)
  end

  def initialize with_string
    @__string__ = with_string
  end

  def to_s
    @__string__.to_s
  end
end

class Denizen < Phrase
  def to_s
    "#{name}, a #{race} with a bunch of #{item.pluralize}"
  end
end

A phrase can have other methods like pluralize that may delegate to subphrases.

class Item < Phrase
  def to_s
    "#{adjective} #{baseItem} #{description}"
  end

  def pluralize
    baseItem.pluralize
  end
end

The default pluralization is just to to_s yourself and add "s"

class Phrase
  def pluralize
    Phrase.new "#{self}s"
  end

  def article
    if self.to_s =~ /^[aeiou]/i
      Phrase.new "an #{self}"
    else
      Phrase.new "a #{self}"
    end
  end
end

Users should be able to generate output like this:

charcter:
  name: Domi Erkthan
  race: dwarf
  profession: bodyguard
  hometown: Raaldeep
  str: 12
  int: 10
  wis: 11
  con: 10
  dex: 9
  cha: 8

From input like this:

name:
  - [humanNameGenerator]
  - [dwarfNameGenerator]
  - [elfNameGenerator]

race:
  - human
  - dwarf
  - elf

profession:
  - adventurer
  - farmer
  - miner
  - aristocrat
  - warrior
  - bodyguard

hometown:
  - Almodel
  - Korvo
  - Raaldeep

stats:
  - str:
    mean: 10
    stddev: 2
    min: 3
    max: 18
  - int:
    mean: 10
    stddev: 2
    min: 3
    max: 18
  - wis:
    mean: 10
    stddev: 2
    min: 3
    max: 18
  - con:
    mean: 10
    stddev: 2
    min: 3
    max: 18
  - dex:
    mean: 10
    stddev: 2
    min: 3
    max: 18
  - cha:
    mean: 10
    stddev: 2
    min: 3
    max: 18

dwarf -> mean str +1, mean con +1, stddev con -1
elf -> mean wis +1, mean int +1, mean dex +1, mean con -1

dwarf -> profession: miner +1, farmer -1
elf -> profession: aristocrat +1

dwarf -> hometown: Raaldeep +1
hometown: Raaldeep -> race: dwarf +1

elf -> hometown: Almodel +1
hometown: Almodel -> race: elf +1

human -> hometown: Korvo +1
hometown: Korvo -> race: human +1

The DSL needs some work... if the user wants to, for, example, change the name of a town, they also have to change all correlation rules that reference the name. There's also no obvious way to express two-way dependencies, e.g. dwarves are more likely to be miners, and miners are more likely to be dwarves. We basically need a GUI to manage this stuff... or a smarter DSL.

Why

We're currently feeling a need to have phrases recur more often than they do by random chance. Recurrence would be useful because it would let us generate worlds with characters and places that reappear over and over, rather than being mentioned only once.

Syntax

I propose the following two-part syntax to allow users to define the "population" of a palette.

Population number in palette definition:

|character*20
[name.capitalize] the [epithet.capitalize]

Used by itself, this syntax would have no effect on the generated output. But it would govern the probability with which characters would recur, when referenced with the syntax below....

Recurrence markers in variables

|gossip
[recur:character] likes [recur:character]!

Here, recur: says explicitly that we want characters we've seen before to be reused. It's necessary to have this syntax because if we apply recurrence blindly, we'll end up generating identical characters when we want to have unique ones.

What *20 means in the character*20 definition is that when we generate a recur:character, the probability of using someone we've seen before will be POP / (POP + 20), where POP is the population of characters that currently exist. If we'd chosen a number larger than 20, the world would seem bigger (because characters would recur less often), and if we'd chosen a smaller number, the world would seem smaller.

@gavmor what do you think? Does this seem okay, at least in outline?