Hi @FilipeFcp Thanks for this!
The interpolation approach makes some sense, but also introduces these new parameters (like number of points). I wonder if instead we should include a way to downsample the fields?

In xarray there is a convenient way to grab every n data point by using slice
https://docs.xarray.dev/en/latest/user-guide/indexing.html

>>> import xarray as xr
>>> da = xr.DataArray([1, 2, 3], [("x", [0, 1, 2])])
>>> da.isel(x=slice(None, None, 2))
<xarray.DataArray (x: 2)> Size: 16B
array([1, 3])
Coordinates:
  * x        (x) int64 16B 0 2

perhaps we could instead allow the user to input the n for slice(None, None, n) (with default of 1)? and use that to downsample in all directions?

Just a thought let me know how you think that would work or if you prefer the interpolation approach

What I had in mind was a way to add more data points, making it easier to create smoother visualizations. For example, moving from this:

To this:

I understand that the slice option is useful for quickly plotting a large dataset, but could it be used to generate a smoother image? However, if users are most likely looking to downsample datasets before plotting rather than generating more points for visualization purposes, the slice option would make more sense.

ah ok, I misunderstood. Maybe in the same sprit, it could be nice if the interpolation argument was a float = 1 that just specified the ratio of points in the interpolation coords to the number of points in the raw data? for example nterpolation=2 would mean we sample the points at 2x resolution of the supplied data. Would this make sense?

The only thing about the current approach I think is not ideal is that it requires the hardcoding of these values (200x200) whereas the data might have a different aspect ratio, so it could lead to weird results.

Are there other examples of interpolation in plotting and what do they do?

Also, could you include a changelog item under Added for the final PR? thanks

Hi @tylerflex, your float approach is actually way better.

Ideally, the user could be able to pass the shading argument for plotting, but it seems to be bugged at the moment: pydata/xarray#3002.

looks like they could also try this?

pydata/xarray#3002 (comment)

Oh... I saw this comment, but I tried with infer_interval = True not False. Yes, it works with False.

I think we can pass the shading argument through - plot_field -> plot_field_monitor_data -> plot_scalar_array -> field_data.plot. Or create a **args in plot_scalar_array and input it into field_data.plot.

Oh... I saw this comment, but I tried with infer_interval = True not False. Yes, it works with False.

Great

I think we can pass the shading argument through - plot_field -> plot_field_monitor_data -> plot_scalar_array -> field_data.plot. Or create a **args in plot_scalar_array and input it into field_data.plot.

yea I think either a shading argument or just exploring all of the **kwargs for xarray would work. I dont have a strong preference. what do you think is better? will this be used often you think?

Hi @FilipeFcp sorry I let this slide, what's the current status of it?

I would say that the current implementation looks pretty good, but we should also add one test, just to make sure calling this with a value of, say 2, will work? and also adding a CHANGLOG item mentioning the new feature

Thanks again!

Hi @tylerflex,

It was my bad. I had a lot of meetings and other tasks when Tom was in China, and I forgot about this for a while.

If you don't mind, I’ve pushed a different version that passes the shading argument. It seems simpler, though we lost the option to reduce the data for plotting. It was ready a month ago, just pending a review and push.

Let me know which implementation you think is better, and I’ll add it to the changelog.

Thank you.

It was my bad. I had a lot of meetings and other tasks when Tom was in China, and I forgot about this for a while.

No worries at all. I also forgeot.

If you don't mind, I’ve pushed a different version that passes the shading argument. It seems simpler, though we lost the option to reduce the data for plotting. It was ready a month ago, just pending a review and push.

Yea this seems great I like it better since it's more simple and adds less complication.

Let me know which implementation you think is better, and I’ll add it to the changelog.

That would be great when you get around to it, thanks!

Ok, I think is done.
Let me know if there is anything to improve.

Thank you!

Thanks @FilipeFcp !