Dubious explanation in documentation for why division by zero doesn't raise an error

[tyilo]

The documentation for gleam/float currently contains the following text:

Infinity and NaN

Under the JavaScript runtime, exceeding the maximum (or minimum) representable value for a floating point value will result in Infinity (or -Infinity). Should you try to divide two infinities you will get NaN as a result.

When running on BEAM, exceeding the maximum (or minimum) representable value for a floating point value will raise an error.

Division by zero

Gleam runs on the Erlang virtual machine, which does not follow the IEEE 754 standard for floating point arithmetic and does not have an Infinity value. In Erlang division by zero results in a crash, however Gleam does not have partial functions and operators in core so instead division by zero returns zero, a behaviour taken from Pony, Coq, and Lean.

This may seem unexpected at first, but it is no less mathematically valid than crashing or returning a special value. Division by zero is undefined in mathematics.

---

The justification

In Erlang division by zero results in a crash, however Gleam does not have partial functions and operators in core so instead division by zero returns zero, a behaviour taken from Pony, Coq, and Lean.

seems weird, as this doesn't explain why 1.0e308 *. 2.0 is allowed to raise an error, but 1.0 /. 0.0 is not.

I don't now what a partial function is defined as in Gleam, but I think *. could be considered partial when it can raise an error.

I think the reasoning could be reworded to explain why overflows are allowed to raise an error, when division by zero is not.

[inoas]

https://gleam.run/frequently-asked-questions/#why-does-division-by-zero-return-zero

[inoas]

and https://tutorial.ponylang.io/gotchas/divide-by-zero.html

[inoas]

Also, https://hexdocs.pm/check_maybe_div_by_zero/
This checker is probably far from perfect but you can use or be inspired by it.
Despite being the author I made peace with this decision.

If in doubt (user/uncontrolled input) I use the int/float module divide/remainder/modulo functions.
If I need speed and/or can control the input I use the literal operators.

[tyilo]

None of these explain why it is ok for overflow to raise an error.

I'm not trying to argue that division by zero should behave differently in Gleam, I just think the current explanation in the documentation doesn't make a lot of sense (at least to me).

[tyilo]

@lpil Why was this closed?

[lpil]

Because it's a duplicate of the previous discussion. Sorry, I thought it would say that when I pressed the duplicate button.

[tyilo]

Can you link that here?