daninet / gmp-wasm Goto Github PK

I've been trying on and off to get gmp-wasm to work for me, and I'm getting a variety of weird behaviors. I suspect one of my bugs can be blamed on Kotlin/JS and another might be related to gmp-wasm. I've created a minimal repository that allows the bugs to be reproduced:

https://github.com/danwallach/gmp-wasm-kotlin

The README.md documents how to run the code and has the outputs. In short, I built a Kotlin BigInteger class that speaks directly to the mpz functions, and tries to manage gmp-wasm memory using FinalizationRegistry (so when the JavaScript wrapper object is garbage collected, it will tell gmp-wasm to free the corresponding mpz object).

I'd appreciate if you could take a look at this.

I'm looking at the way you're managing memory, by keeping mpz_t_arr for all the temporary values and then blowing all of them away when the surrounding context finishes. I'm curious if you've looked at the possibility of leveraging finalizers, to allow the JavaScript garbage collector call back to you when it's time to ultimately call mpz_t_free.

This would allow higher-level logic to write arithmetic expressions as normal, without having to worry about object lifetimes.

Hi! I am using the gmp-wasm in typescript, the link to typescript API on README seems not working.(https://paka.dev/npm/gmp-wasm/v/0.9.4)

For what it's worth, I want to call this library from Kotlin, which apparently would be easier if there was a declaration file (*.d.ts). Not that I have any idea what that is, but I'm curious if it's on your agenda to make such a thing.

First, thank you for this library. It's exactly what I need.

GnuMP has a very useful feature for selecting bit ranges from an xmpz. For my particular application, this allows for an optimization worth about 9x (precompute to optimize modular exponentiation for known bases).

Python binding: https://gmpy2.readthedocs.io/en/latest/advmpz.html

I'd love something analogous to the "bit slice" read feature. I don't actually need anything else, but I imagine others might want the full suite of xmpz mutability features.

Also, in terms of general purpose features, it would be handy to have input and output as byte arrays. Java BigInteger defines this to be big endian, twos complement, which seems as good as anything else you might do.