In reply to@_slack_fatiando_U022GF9FVLL:matrix.org
Unsure if this is a Verde Q, or better for xarray. But if I wanted to calculate 2D gradients of a surface, Is that something you would do inside of Verde or no?

In reply to@_slack_fatiando_UMSRSPEMA:matrix.org
We don't have anything specific for that yet, sorry. I'm assuming you already have a grid, right? If so, a finite difference derivative with pure xarray should be doable. Otherwise, you could fit a Spline to points and do a finite difference calculation with predict. I have a PR open for something that does exactly that (haven't had time to finish it, though).

In reply to@_slack_fatiando_UMSRSPEMA:matrix.org
Actually, if you have a grid already it's pretty easy: http://xarray.pydata.org/en/stable/generated/xarray.DataArray.differentiate.html

In reply to@_slack_fatiando_U022GF9FVLL:matrix.org
No worries! I am happy to do it in xarray. I just am not trying to chase my tail 🙂 Thanks Leo!

In reply to@_slack_fatiando_U022GF9FVLL:matrix.org
I am going to go from points -> Verde -> xarray -> derivatives

In reply to@_slack_fatiando_UMSRSPEMA:matrix.org
👍 Glad to help! I'm hoping to find some time to work on https://github.com/fatiando/verde/pull/245 over the summer break, It would help skip a step in that process.

In reply to@_slack_fatiando_U0156QCM6AH:matrix.org
Sounds good

In reply to@_slack_fatiando_U022GF9FVLL:matrix.org
Any thoughts on xarray vs vaex? https://vaex.io/docs/index.html#

In reply to@_slack_fatiando_UMSRSPEMA:matrix.org
Never really used vaex but it seems made for data frames / tables rather than grid and multidimensional arrays. But it looks really promising and I’d like to see how they do things under the hood

In reply to@_slack_fatiando_U022GF9FVLL:matrix.org
At least for datasets of this size (1-5GB's) it seemed the same speed as pandas? I didn't do extensive testing, but at first sniff I could not see a big reason to move to vaex

In reply to@_slack_fatiando_U0156QCM6AH:matrix.org
No, would need to have a case where it was better or have something different I think

In reply to@_slack_fatiando_UMSRSPEMA:matrix.org
Maybe when the data is too big for RAM? But then you also have Dask for those cases 🤷‍♂️

In reply to@_slack_fatiando_U022GF9FVLL:matrix.org
Yeah I saw someone else was making the same point. Unsure of the exact use case.

In reply to@_slack_fatiando_U0156QCM6AH:matrix.org
Yes, it has lazy evaluation...Vaex I think did well in a comparison I read last year with respect to computation speed