Proper way to declare a const int variable?

[sam0x17]

I need to declare a const MY_NUM: I512 but I can't figure out syntax that would allow me to set the value in a const context. Is there not a way to do this yet? Would be happy to write a macro for this purpose if there is any possible way to initialize one of these in a const context.

[isaacholt100]

Hi, thanks for opening this issue! Writing a proc macro is on my todo list for this crate, but I haven't got round to it yet. Is the value that you want to initialise a single literal or is it the result of a computation? I think in theory it should be possible to create any value using the const methods defined on I512 but a macro for compile time parsing of literals would definitely be beneficial.

Currently, the from_str method in the FromStr trait implementation is not const but I could make a const equivalent, which would at least allow const parsing of decimal literals, and would be fairly quick to release as a temporary workaround.

[sam0x17]

That would be awesome. In the mean time is there an example of initializing one with one of the const methods?

…

On Fri, Feb 24, 2023 at 1:19 PM Isaac Holt ***@***.***> wrote: Hi, thanks for opening this issue! Writing a proc macro is on my todo list for this crate, but I haven't got round to it yet. Is the value that you want to initialise a single literal or is it the result of a computation? I think in theory it should be possible to create any value using the const methods defined on I512 but a macro for compile time parsing of literals would definitely be beneficial. Currently, the from_str method in the FromStr trait implementation is not const but I could make a const equivalent, which would at least allow const parsing of decimal literals, and would be fairly quick to release as a temporary workaround. — Reply to this email directly, view it on GitHub <#17 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAOE4LIRDAMC4GLHBHIWY53WZD3UDANCNFSM6AAAAAAVFDJVXE> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[isaacholt100]

I think the simplest example would be using the from_digits method on U512 to convert an array of digits to a U512 and then using the from_bits method on I512 to convert that U512 to an I512. So for example (I'm using I256 here for readability), if the hexadecimal literal was 0x5501ece471b59c0ace87fdad70306960eca1623f20207afa6e316e83cb2e3 and you wanted this represented as an I256, you could use:

use bnum::types::{U256, I256};

const MY_NUM: I256 = {
    let d1 = 0xafa6e316e83cb2e3u64;
    let d2 = 0x960eca1623f20207u64;
    let d3 = 0xc0ace87fdad70306u64;
    let d4 = 0x5501ece471b59u64;
    I256::from_bits(U256::from_digits([d1, d2, d3, d4]))
};

The from_digits method takes the least significant digit first so the lower order part of the literal first is the first item in the array. Hopefully this makes sense!

[isaacholt100]

And then you could use methods like wrapping_shl, rotate_right, wrapping_add, etc. which are const to manipulate this further if needed.

[sam0x17]

Yes thank you!

[isaacholt100]

No worries! Good news, I have just made the from_str_radix on unsigned integers const, so once I have tested it fully I will publish a new version with this change. I have also created another issue specifically for writing a macro for the purpose you described: #18.

[isaacholt100]

I have just published version 0.6.0 which provides a parse_str_radix radix method for compile time parsing of integers from string literals, so I will close this issue now.