DOC: Flip the imshow plot types example to match the other examples

[scottshambaugh]

PR summary

Flips the imshow plot types example to put the origin at the lower left instead of the default upper left. This allows for direct comparison with the 6 other examples that use the same data.

Before:

After:

PR checklist

• [n/a] "closes #0000" is in the body of the PR description to link the related issue
• [n/a] new and changed code is tested
• [n/a] Plotting related features are demonstrated in an example
• [n/a] New Features and API Changes are noted with a directive and release note
• Documentation complies with general and docstring guidelines

[story645]

am back and forth b/c agree that direct comparison is better suited for choosing an appropriate function but a bit unsure about this b/c ideally the plot types gallery would use as many defaults as possible/show the default behavior

[scottshambaugh]

We do put a few non-default kwargs in other examples, like an alpha value. Just added the "after" screenshot to the top level description, does the 1:1 comparison look worthwhile after seeing it?

[story645]

We do put a few non-default kwargs in other examples, like an alpha value.

Yeah styling, not data reading :/ But yeah the flip option of inverting the first array so the images match is probably more confusing.

[story645]

Thinking about this more, on net, better for the functions to be comparable. Also honestly might be good to expose the origin keyword so it doesn't trip folks up in precisely this data arrays read are arranged differently from images way.

[jklymak]

I have exactly the opposite take. Folks will use imshow, their plots will be upside down, and they will wonder why that is when this page has the plots all the same. I doubt most people click through to the example to see the use of the origin keyword. I think this change will cause more confusion than not.

[scottshambaugh]

So my thought with making this change was:

• I think that it's easy to get confused by the imshow coordinates, and showing that kwarg in more examples helps bring attention to the flag.
• Having looked at that page maybe a dozen times, I don't think it's obvious that the plot is upside down - it just looks like blobs. I had no idea it was the same as the other examples until I was looking at the sources for another update. I think having it match the rest is a better way to key in on what the differences in the different methods are.
• My impression is that imshow is really meant to be used with imread/PIL.Image.open rather than with generated data. If we're going to further improve the example, I'd propose we use an actual image for the example rather than a blobby raster field.

[jklymak]

To me, the goal is to show what the same data looks like when plotted with different methods. I think that is best served using the defaults.

[story645]

To me, the goal is to show what the same data looks like when plotted with different methods

That's why I came around to thinking that showing the use of the origin keyword is helpful. Like the first time I plotted data using imshow it was upside down & I was so confused - and manually flipped it using something like data[::-1] b/c I didn't know about the keyword.

[timhoffm]

The plot types page is primarily to give an overview of the different kinds of visualizations, not teach any of their specifics. In that sense, it doesn't matter too much whether the data/result image is flipped or not. But generally, I'd go with using the functions mostly as is and not doing extra things to hide any specific behavior. So, I'm -0.5 on this, but don't find it important enough to go into a lengthy discussion.