It's not clear from the article, but "niche optimization" does not mean "optimization that is only useful in a very specific circumstance".
It is a specific optimization based on the idea of storing one type inside of another type by finding a "niche" of unused bit pattern(s) inside the second type.
It has far more useful application than a tower of Option 254 deep.
True | False | FileNotFound was a meme about 2 decades ago, and even that was a reference to MSDOS from another 2 decades earlier. I guess things never change, only the language.
Even now, I still find myself using true/false/null on occasions, but I'm usually smart enough to replace it with an enum at that point. The only time I don't is when it's an optional parameter to a function to override some default/existing value, at which point it then makes sense to keep it as an optional bool.
I'm surprised that trinary logic has not become a standard part of standard libraries yet. Almost every project I have worked on ends up with some form of a yes/no/maybe abstraction.
Yes/No/Maybe is a good fit for an enum because “Maybe” carries some specific information.
For more common situations where the yes/no bool is not available yet or should not be considered, constructs like Rust’s Option<bool> are a very good fit. Layering the bool inside of an Option creates intentional handling about the presence or lack of value first before you can work with it.
Neat. Even knowing about niche optimization I would have guessed that you could fit 7 Options - one bit for each. But the developers were smart enough to take advantage of the fact that you can't have a Some nested below a None, so you only need to represent how many Somes there are before you reach None (or the data), allowing 254 possibilities.
I doubt they were thinking about Option<bool> when making niches work like this.
Option<NonZeroU32> seems like a much more reasonable to justify this with. Also, enums can easily have invalid bit patterns that are unused without there being any specific bit that is always available. All you need is a single variant of the enum to have a free bit, and you have a niche to shove None into.
public static void main (String[] args) {
Optional<Boolean> n = Optional.ofNullable(null);
Optional<Boolean> e = Optional.empty();
System.out.println(n.equals(e));
}
true
Optional<Boolean> n = null;
Optional<Boolean> e = Optional.empty();
Optional<Boolean> t = Optional.of(Boolean.TRUE);
Optional<Boolean> f = Optional.of(Boolean.FALSE);
This question always reminds me that we often compress far more nuance into binary decisions than reality allows.
In practice most systems end up inventing “soft booleans” (flags, states, priorities) to deal with that.
> looking at Rust … it turns out that `Option<bool>` takes up exactly one byte of memory, the same as bool! The same is true for `Option<Option<bool>>`, all the way up to 254 nested options.
Ah how many of those options fit into that boolean. Word games!
It's not clear from the article, but "niche optimization" does not mean "optimization that is only useful in a very specific circumstance".
It is a specific optimization based on the idea of storing one type inside of another type by finding a "niche" of unused bit pattern(s) inside the second type.
It has far more useful application than a tower of Option 254 deep.
True | False | FileNotFound was a meme about 2 decades ago, and even that was a reference to MSDOS from another 2 decades earlier. I guess things never change, only the language.
Even now, I still find myself using true/false/null on occasions, but I'm usually smart enough to replace it with an enum at that point. The only time I don't is when it's an optional parameter to a function to override some default/existing value, at which point it then makes sense to keep it as an optional bool.
I did a govt contract early on and learned that yes/no/unanswered/unasked was a common quad. I see that in disclosures when applying for jobs as well.
I'm surprised that trinary logic has not become a standard part of standard libraries yet. Almost every project I have worked on ends up with some form of a yes/no/maybe abstraction.
Yes/No/Maybe is a good fit for an enum because “Maybe” carries some specific information.
For more common situations where the yes/no bool is not available yet or should not be considered, constructs like Rust’s Option<bool> are a very good fit. Layering the bool inside of an Option creates intentional handling about the presence or lack of value first before you can work with it.
With privacy coming back into vogue, it’s useful to distinguish “we didn’t ask” from “they wouldn’t answer”
For some vector logic the distinction could matter.
Neat. Even knowing about niche optimization I would have guessed that you could fit 7 Options - one bit for each. But the developers were smart enough to take advantage of the fact that you can't have a Some nested below a None, so you only need to represent how many Somes there are before you reach None (or the data), allowing 254 possibilities.
I doubt they were thinking about Option<bool> when making niches work like this.
Option<NonZeroU32> seems like a much more reasonable to justify this with. Also, enums can easily have invalid bit patterns that are unused without there being any specific bit that is always available. All you need is a single variant of the enum to have a free bit, and you have a niche to shove None into.
The scoop: a boolean can't be smaller than a byte. Full 254 level of nested Option<bool> fit into it. (C++ needs much more for even a single level.)
For Java developers... you can use Optional<Boolean> to store the elusive four possible booleans.
5, no? Null, Optional::empty, Optional(null), Optional(true), Optional(false)
A null in an Optional is empty. So you've got:
This question always reminds me that we often compress far more nuance into binary decisions than reality allows. In practice most systems end up inventing “soft booleans” (flags, states, priorities) to deal with that.
The deeper you go into memory layout, the more you realize that even "simple" types aren't that simple.
>and that it takes up one byte of memory
You can make them smaller using bitfields in C.
The object it's inside will still take up at least one byte.
Amortization.
If one Boolean must be a byte then 8 must be eight bytes. Which is not true. A boolean can be 1/8th of a byte which is a meaningful distinction.
Um, no. Please show me how you can fit 255 possible states in something smaller than a byte by using bitfields.
I was quoting the first paragraph, where it says a single normal bool takes a byte.
> looking at Rust … it turns out that `Option<bool>` takes up exactly one byte of memory, the same as bool! The same is true for `Option<Option<bool>>`, all the way up to 254 nested options.
Ah how many of those options fit into that boolean. Word games!