argriffing

I see this is related to https://github.com/akheron/jansson/pull/158. I'd use it for scientific computing instead of for bitcoin or financial data, but that PR would work for me.

Maybe also related to this bignum issue https://github.com/akheron/jansson/issues/69

Just to summarize, there are two issues. One is that `np.log(m)` has different behavior than `np.ma.log(m)` when `m` is a masked array. A separate issue is that `np.ma.log` has trouble...

Here's a PR with the tests https://github.com/fredrik-johansson/arb/pull/64. I've decided to just put the convenience functions in the test file. Maybe there does not need to be a combinatorial explosion of...

I wasn't sure what a Taylor model is, so I found the reviews [Taylor forms – use and limits](https://pdfs.semanticscholar.org/8873/628063101cb8a2a04704a61c432278d6e07b.pdf) by Arnold Neumaier and [From Taylor Series to Taylor Models](http://bt.pa.msu.edu/pub/papers/taylorsb/taylorsb.pdf) by...

I have a change that removes this warning, which I'll push when github has dealt with whatever IT problems it's having https://status.github.com/

@WeatherGod I think that using a numpydoc.py file in the sphinxext directory, as sampledoc currently does, is an obsolete and unsupported way to use numpydoc. If there are three options...

I like the idea of implementing a `fmpz_mat_snf_transform` by copying the Kannan-Bachem code and adding some bookkeeping. I'm not interested in a modular variant, and I think Flint is OK...

I ended up using only `fmpz_mat_hnf_transform` instead. But I'm explicitly inverting the `U` matrix afterwards which is slow and makes me feel stupid. I notice that GAP has options to...

> That's interesting, which gap function is it? I did not try using it, but I found a function named `HermiteNormalFormIntegerMatInverseTransforms` mentioned in the docs at http://www.math.rwth-aachen.de/~Greg.Gamble/gap4r3/doc/htm/ref/CHAP025.htm#SSEC001.7