developer workflow

[qAp]

I want to make a change to xarray.plot.plot.line and use that change, what is the best way of going about this? I was thinking that I could

(1) Fork xarrary on Github
(2) Make a local clone of the fork
(3) Edit the code for the function line in 'plot/plot.py'

Then? I already use xarray that is installed with conda: conda install xarray .

Sorry need to ask because never done this before.

[fmaussion]

Hi, if you want to change something in xarray and use it instead of the original one you should uninstall it, and then install it in dev mode. From the xarray root directory:

pip install -e .

However, in that case this is probably not the best thing to do. Since the logic for xarray's line plot is extremely simple, in that case you'd better make you're own function out of it and call it instead. I.e. my_plot_func(dataarray) instead of datarray.plot().

Finally, you may want to consider using an accessor for this purpose. You will then be able to write something like:

datarray.qAp.plot()

[qAp]

I used an accessor to add more options for plotting 1d arrays; main ones demonstrated here:
https://github.com/qAp/climatools/blob/master/climatools/examples/1d_plotting/plotting_1d_DataArray.ipynb

Do you think these would be useful in xarray? I guess they save time for people who are not familiar with matplotlib...

[fmaussion]

Thanks for these examples!

I cannot speak for the xarray developers, but my felling is that these options are too specific. Furthermore, doing the same things would require only one or two lines of code on top of xarray.plot().

However, I understand the argument that people who don't know matplotlib might find this useful, thought. The figsize kwarg could be a useful addition for example.

[shoyer]

A size argument would definitely be useful (see #637, which needs somebody to finish it up). I could see log scale options also being useful, though I would stick with the names used by pandas for consistency (e.g., logx=True)

[fmaussion]

Right, I forgot about #637 . I'll see what I can do.