you might need to use IFloatingPointIEEE754 instead of IFloatingPoint btw
you might need to use IFloatingPointIEEE754 instead of IFloatingPoint btw
Kone of those inherits from IRootFunctions and the other doesnt, i dont remember which (and the one that does inherits from the other)
M
Mhmm interesting
Ayep you're right only the ieee754 implements IRootFunctions
Kyeah theres a whole hierarchy of the interfaces with varying granularity
Kfor instance ITrigonometricFunctions, IHyperbolicFunctions, etc
Kok cool i was remembering right
Kthe root functions etc are only guaranteed for binary floats where it can be done fast
KKand then theres IBinaryFloatingPointIeee754 which adds bitwise, ispow2, allbitsset, and log2
Pwho's working on what atm
Pthis channel seems exciting all of a sudden
Kidk
Ka long discussion is happening about generic math things
Kin which petre grifon has seemingly solved our issue with length
Ai can recommend not requiring the
IFloatingPointXYZ<> interface directly, instead only requiring IRootFunctions<T> or ITrigonometricFunctions<T> etcAit will make your life easier
Kmhm
Kso theres one last concern and thats the problem of converting
Kwhich i dont think we have an easy generic constraint for
Aas in casting from one
VectorX<T> type to another VectorX<T> type?Awith a different type for
T?Pwe need to manually specialize those anyway seeing as you can't have variable Ts on implicit operators
Kwait hang on
Ptbh i'm really quite keen on the extension method approach, it probably makes sense to only require on the type itself the interface containing +-*/
Aproblem solved
Kyeah i had forgotten about CreateChecked
ACreateChecked is your best friendAi forgot stuff like
float.CreateChecked existed in my older generic math implementationsAso i used
Convert.ChangeType or whatever the similar function is calledKand by forgotten i mean i didnt even know it existed until just now
Abut
CreateChecked is much better, i use it when providing cast operators between system.numerics and my generic typesMi'm using CreateTruncating for speed
Mbut i can use checked if desired
Apersonally i'd recommend
CreateChecked simply because it reduces the risk of undefined behaviourAso if someone tries to cast a
VectorX<int> with an X value of 30000 into a VectorX<byte> it will error, just like how that would normally happenPyou do realise we have checked operators now
Pwe can do both
Aoh cool
Moh yeah cool
MGenericVector3.cs34.15KB
Minitial impl, no extension methods
Mrough around the edges
Mlemme polish it...
Aremember that
T.Zero and T.One existsMyeah that is what i meant by polishing
Aah ok
Mhow tf you meant to implement ISpanFormattable
Mhacked this together
