bigi - BigInt library | The Gleam Programming Language | Page 1

chrome dew Feb 4, 2024, 7:36 PM

#

The Gleam Int type behaves differently in the Erlang and JavaScript targets. In the Erlang target it maps to Erlang's integer which is a "big integer", meaning it has arbitrary precision and can be as big as you need. In the JS target, Int maps to Number, which is a floating point value and has max/min values after which it is converted into Infinity, and additionally max/min safe values after which precision suffers.

To provide a consistent interface, I made the little Bigi package: https://hex.pm/packages/bigi. It works the same in both targets, using BigInt in the JS side. It has a rudimentary set of math operations implemented.

I mainly see this as a useful base for other packages that want to maintain correctness without imposing limits, and want to support both targets. If you are just targeting Erlang, you mostly don't need this, and if you are just targeting JS, you can also use the BigInt bindings in https://hex.pm/packages/plinth.

placid rover Feb 4, 2024, 7:55 PM

#

Nice. At one point I worked quite a bit the BigInt ffi in gleam, but after quite a while I figured out I actually had to work with decimals (wanting to divide two large values and get a decimal value) and then on the JS side I actually had to work with strings to implement that kind of functionality, which then became a mess :p

#

A bit offtopic, but just in case you ever consider making a BigDecimal lib :d

chrome dew Feb 5, 2024, 4:42 AM

#

It's not in the cards right now, because while this was just mostly bindings and thus easy to make, a decimal library probably isn't 😛 I'm not a great maths person

#

Could take a look at how Decimal works in Elixir though

opaque sigil Feb 5, 2024, 7:31 AM

#

I like it! If you could unify the language for floats too that would be nice. I made a proof of concepts with NIFs that try to emulate js-floats in Erlang, but there might be other options that work better.

chrome dew Feb 5, 2024, 7:43 AM

#

What do you mean with "unify the language for floats"? I understand Erlang doesn't do things like NaN while JS does, do you mean that?

opaque sigil Feb 5, 2024, 9:51 AM

#

Yes, that's exactly what I meant. 😅

chrome dew Feb 5, 2024, 9:52 AM

#

I think that's above my skillset right now 🙂 would require a lot of investigating

#

I took a look at Decimal too and while I think I could translate it to Gleam, I couldn't really fix any bugs in it or add new features because I don't know how it works. So I wouldn't be the best package owner.

#

now what I have been thinking about is a ranged integer library, to allow you to have unsigned ints and things like uint32, int8, etc etc. with optional overflow/underflow semantics

chrome dew Feb 15, 2024, 7:34 PM

#

the shame: https://gitlab.com/Nicd/bigi/-/commit/7545b2aaf8cf3e1466afa1bb8c9bc31f95ef7966

GitLab

Fix power function to return a result (7545b2aa) · Commits · Mikko ...

GitLab.com

placid rover Feb 15, 2024, 8:59 PM

#

there'll have to be a bigdecimal lib at some point ;)

chrome dew Feb 16, 2024, 6:16 AM

#

maybe I should have made the negative exponent do integer division that many times? but since it rounds on every step, the result might be far from expected

chrome dew Feb 16, 2024, 9:37 PM

#

Bigi 2.1.0 has been pushed with decoders from Dynamic and String, thanks to McNimble for the idea

chrome dew Apr 11, 2024, 3:10 PM

#

Oh wow, looks like I've swapped the order of arguments in bigi.compare when compared to int.compare and friends from stdlib

#

no wonder my blog is ordering things confusingly!

#

Bigi 3.0.0 has been released, the only change is that the order of arguments for bigi.compare has changed.

chrome dew Sep 12, 2024, 4:03 PM

#

Bigi 3.1.0 has been released. It contains added to_bytes and from_bytes functions to convert big integers to/from bit arrays (with adjustable endianness and signedness), courtesy of @primal brook. 🙂

chrome dew Sep 13, 2024, 7:58 AM

#

Oh, and added information to the docs on the Erlang big integer limits.

chrome dew Oct 3, 2024, 12:50 PM

#

@hushed ibex I noticed you used bigi as a dependency in your package but had to FFI for some shifting functions. Do you think those should be in bigi directly? I could take a look at it.

#

it's curious that I failed to add left/right shift since they're directly supported in JS

hushed ibex Oct 3, 2024, 4:32 PM

#

I think it would be useful to ensure that the bigi api matches the gleam int api. Adding the bit shifting to my code was a bit of a learning experience so I didn't think about creating an issue/pull request for bigi.

chrome dew Oct 3, 2024, 4:33 PM

#

true, there's a lot of functions I didn't implement

#

I'll take a look at those when I have time

chrome dew Oct 3, 2024, 7:52 PM

#

most of them are done: https://git.ahlcode.fi/nicd/bigi/commit/00a57cf8de4299e51dbc32f03ccae03f6ab907ed

but I forgot product, will have to add that tomorrow and publish

chrome dew Oct 4, 2024, 8:52 AM

#

Bigi 3.2.0 has been released with the following changes:

+ Added functions to match gleam/int in stdlib:
  + bitwise_and
  + bitwise_exclusive_or
  + bitwise_not
  + bitwise_or
  + bitwise_shift_left
  + bitwise_shift_right
  + clamp
  + floor_divide
  + is_odd
  + max
  + min
  + negate
  + product
  + sum
  + undigits

hushed ibex Oct 4, 2024, 9:08 AM

#

I appreciate you updating your useful library which has allowed me to remove the ffi from my library.

chrome dew Oct 4, 2024, 9:21 AM

#

Np! Don't hesitate to contact me whenever you have wishes about my libraries. 🙂

chrome dew Jun 21, 2025, 7:26 AM

#

Bigi 4.0.0 has been released. This removes the decode(dyn) function and replaces it with a decoder() function for a more idiomatic decoding API. Thanks to @high void for noticing it and @static crater for pinging me. There's no need for creating your own decoder now.

static crater Jun 21, 2025, 7:42 AM

#

Hero

chrome dew Jun 21, 2025, 7:47 AM

#

now I just need to bump all my dependent libraries to use 4.0.0…

high void Jun 21, 2025, 1:23 PM

#

chrome dew **Bigi 4.0.0 has been released.** This removes the `decode(dyn)` function and re...

Thanks

chrome dew Jun 22, 2025, 3:56 PM

#

heads up @forest acorn you might check if you can relax the bigi dep in lenient_parse to include 4.0.0

#

also heads up @hushed ibex

forest acorn Jun 22, 2025, 4:03 PM

#

chrome dew heads up <@408106253626245120> you might check if you can relax the `bigi` dep i...

I’m still using a hardcoded version, but I should swap to using yours directly now with the GitHub support in gleam packages

chrome dew Jun 22, 2025, 4:03 PM

#

GitHub..? 🤔

forest acorn Jun 22, 2025, 4:03 PM

#

Oh damn

#

It’s not on github

chrome dew Jun 22, 2025, 4:04 PM

#

why do you need a git dependency?

forest acorn Jun 22, 2025, 4:04 PM

#

I’m thinking of the wrong thing too

#

Thinking of the float package you wrote

chrome dew Jun 22, 2025, 4:05 PM

#

oh yeah (fixed not float 😏)

#

also heads up @median storm I see you depend on bigi, probably you can widen the dependency without issues (if you're not using the decode function)

forest acorn Jun 22, 2025, 4:55 PM

#

What do you suggest for relaxing the dep version?

#

Maybe I can change bigi = ">= 3.2.0 and < 4.0.0" to just bigi = ">= 3.2.0

#

How do I update my local stuff to use the bigi 4.0.0? I removed that constraint and tried updating, but 3.2.0 remains. Gleam removing and adding also only brings in 3.2.0.

#

Oh, maybe my hardcoded version of pilkku is keeping it pinned

#

I'm not sure. 🤷‍♂️

static crater Jun 22, 2025, 5:09 PM

#

forest acorn Maybe I can change `bigi = ">= 3.2.0 and < 4.0.0"` to just `bigi = ">= 3.2.0`

Possibly just change the upper bound to < 5.0.0?

#

So bigi = ">= 3.2.0 and < 5.0.0"

chrome dew Jun 22, 2025, 5:10 PM

#

yeah that one

#

I can't guarantee what break 5.0.0 would have

forest acorn Jun 22, 2025, 5:11 PM

#

I'm able to relax that, but the gleam CLI still continues to install 3.2.0

#

#

lament halo Jun 22, 2025, 5:15 PM

#

Maybe try gleam deps update?

forest acorn Jun 22, 2025, 5:18 PM

#

oh! That worked! nice. I'm still bad at dep management.

#

Thank you!

lament halo Jun 22, 2025, 5:18 PM

#

You're welcome!

forest acorn Jun 22, 2025, 5:23 PM

#

Alas, now I have compilation issues in dev deps. First with gleam_json, which I relaxed as well in the same way, then in startest, which is using the latest version. Looks like the current dep restraints are set up such that we download all the latest versions of all deps now. 😭

lament halo Jun 22, 2025, 5:24 PM

#

That's odd, might it be an issue with the stdlib?

forest acorn Jun 22, 2025, 5:25 PM

#

📎 message.txt

#

Does startest need to be updated?

lament halo Jun 22, 2025, 5:26 PM

#

I'd guess so

high void Jun 22, 2025, 6:08 PM

#

forest acorn

What editor is this?

#

Reminds me of Atom

forest acorn Jun 22, 2025, 6:10 PM

#

https://zed.dev

Zed

Zed — The editor for what's next

Zed is a high-performance, multiplayer code editor from the creators of Atom and Tree-sitter.

#

Our founders built Atom previously

high void Jun 22, 2025, 8:32 PM

#

forest acorn https://zed.dev

How well does it work with Gleam?

wind nova Jun 22, 2025, 8:52 PM

#

Works well

static crater Jun 22, 2025, 9:19 PM

#

I use Zed all the time. Gleam support is top notch

median storm Jun 23, 2025, 8:40 AM

#

chrome dew also heads up <@431378090380361739> I see you depend on `bigi`, probably you can...

No worry, it's for my Bun bindings, no problem with bigi I think ! 😄 Thanks for warning though.

#bigi - BigInt library