the myth generator output is mostly structurally correct. there's an event listing, all references to other objects are linked, actors have a list of all events they were involved in (though not what changes were made during what events), etc. there are a few issues with ids not being recursively set.
but now it's fairly solidly in the realm of "the dataset is too dry", rather than "all the code is busted". this is a problem, because part of the reason the dataset is so dry is that it's difficult to encode any really complex interactions without immediately having a bunch of weird edge cases. and maybe that's fine. but this is the part where i generate histories and go "okay sure you can maybe interpret a narrative, but none of it really means anything". and the question is, what does mean anything mean.
also there's the part where it's pretty impossible to encode complex data. like, okay i can have gods and i can give those gods different aspects and i can have them do different things based on their aspect. what i definitely can't do is have gods gain aspects. change, yes, but not gain, because then they would go from having one of a thing to having multiple of a thing, and there isn't really any way to encode that in the data as of yet. it falls down a lot when it comes to numbering things, actually. one of the things i want to add is things like "{minor god or mythical hero/demon} was slain and {their} blood/bones/etc became {some set of objects}", whether that's "their blood became the fifty-seven oases" or "their bones became the five mountains"; there's not any way to say "generate x number of these templates and instance all of them".
there's actually some stuff that i could add in now, like e.g., when a world is generated as cohered around a bound god then something like "somebody dug up some weird magic ore" that would be parts of the binding chain, thus weakening the binding. or ditto from a world made from a dead god, being able to dig up ore that's literal god bones. the actual issue with that is... randomness? like when a character acts it doesn't mean anything because there's no preferences; the system just randomly picks an option from all options and then keeps trying until it finds one that it can successfully instance. even if i put in actions like "somebody finds weird ore" "somebody turns the ore into magic metal" "somebody makes magical artifacts from magic metal", then even if i put in character templates that can make artifacts, they'll only happen to do that when the dice rolls go their way; they have no way to plan for the future (which would be difficult) or even express goals (which would be less difficult).
so there are more than a few critical steps missing before this can really do anything i'd like it to do. like okay sure everything being random noise isn't the worst; there are plenty of simulation setups that randomly roll everything and they're ultimately fine. but right now the only real way to give actors different 'desires' would be to give them classes, basically, and to make different sets of actions for each class. which... is fine. but would probably still be pretty dry.
i mean that being said even with the code in its current state i could stand to expand my dataset a lot. its just kind of frustrating to look at it in terms of all the things that are absolutely impossible given the current state of the code.
but now it's fairly solidly in the realm of "the dataset is too dry", rather than "all the code is busted". this is a problem, because part of the reason the dataset is so dry is that it's difficult to encode any really complex interactions without immediately having a bunch of weird edge cases. and maybe that's fine. but this is the part where i generate histories and go "okay sure you can maybe interpret a narrative, but none of it really means anything". and the question is, what does mean anything mean.
also there's the part where it's pretty impossible to encode complex data. like, okay i can have gods and i can give those gods different aspects and i can have them do different things based on their aspect. what i definitely can't do is have gods gain aspects. change, yes, but not gain, because then they would go from having one of a thing to having multiple of a thing, and there isn't really any way to encode that in the data as of yet. it falls down a lot when it comes to numbering things, actually. one of the things i want to add is things like "{minor god or mythical hero/demon} was slain and {their} blood/bones/etc became {some set of objects}", whether that's "their blood became the fifty-seven oases" or "their bones became the five mountains"; there's not any way to say "generate x number of these templates and instance all of them".
there's actually some stuff that i could add in now, like e.g., when a world is generated as cohered around a bound god then something like "somebody dug up some weird magic ore" that would be parts of the binding chain, thus weakening the binding. or ditto from a world made from a dead god, being able to dig up ore that's literal god bones. the actual issue with that is... randomness? like when a character acts it doesn't mean anything because there's no preferences; the system just randomly picks an option from all options and then keeps trying until it finds one that it can successfully instance. even if i put in actions like "somebody finds weird ore" "somebody turns the ore into magic metal" "somebody makes magical artifacts from magic metal", then even if i put in character templates that can make artifacts, they'll only happen to do that when the dice rolls go their way; they have no way to plan for the future (which would be difficult) or even express goals (which would be less difficult).
so there are more than a few critical steps missing before this can really do anything i'd like it to do. like okay sure everything being random noise isn't the worst; there are plenty of simulation setups that randomly roll everything and they're ultimately fine. but right now the only real way to give actors different 'desires' would be to give them classes, basically, and to make different sets of actions for each class. which... is fine. but would probably still be pretty dry.
i mean that being said even with the code in its current state i could stand to expand my dataset a lot. its just kind of frustrating to look at it in terms of all the things that are absolutely impossible given the current state of the code.
