Michael Shoemate

I believe the pyo3 dependency adds a linker dependency on Python itself that was until now completely compiled out. Since this PR implements a third-party Polars plugin that exercises this...

Another source of panics are overflow/underflow. For example, the following, which underflow panics instead of rounding down: ```rust FBig::::try_from(-6871947604880523000.)?.exp() ```

It's also not possible to anticipate settings where the calculation might overflow. Take for example, the following, which panics: ```rust let lhs = f64::MAX; let rhs = f64::MIN_POSITIVE; FBig::::try_from(lhs)? +...

All of these options strike me as viable solutions. Comments on each approach, respectively: 1. I don't have any issue with handling results, because all my call sites already return...

Transitioning from [this conversation](https://github.com/opendp/opendp/issues/457#issuecomment-1454974377). I was hoping to get some guidance on sampling Gumbel RVs, and now have some questions/observations about specific algorithms on your site (from the linked issue)....

Ah, it's basically the same as Cannone et al, but with a Poisson sampler instead. I was hesitant to choose those parameters because I was concerned about finding the PSRN...

@peteroupc Thanks for checking in. I used inverse transform sampling to address the original reason why I was trying to sample Gumbel variates: https://github.com/opendp/opendp/pull/653 I find this approach pretty appealing,...

In my case, I only encounter this issue if I have `from __future__ import annotations` in the file. I avoided the issue by just using `Union[A, B]` instead of the...

Since this has been quiet for a while, I'm closing in favor of #628. Please send me a message if you find the time to pursue this further! Thanks.

Hi Jenny, thanks for the response! This part here located at the top of the log: ``` Registered S3 method overwritten by 'opendp': method from print.hashtab utils ✖ mod.R:390: S3...