Version: 1.35.1 (user setup)
Commit: c7d83e57cd18f18026a8162d042843bda1bcf21f
Date: 2019-06-12T14:30:02.622Z
Electron: 3.1.8
Chrome: 66.0.3359.181
Node.js: 10.2.0
V8: 6.6.346.32
OS: Windows_NT x64 10.0.17763
Python Extension: 2019.6.22090
The new release includes a new feature to interact with the plots:

This implement has caused the following issues:
import matplotlib.pyplot as plt
import numpy as np
import geopandas as gpd
# load dataset
path = gpd.datasets.get_path('naturalearth_lowres')
gdf = gpd.read_file(path)
# generate a random column
gdf['random_col'] = np.random.normal(100, 10, len(gdf))
# plot quantiles map
fig, ax = plt.subplots(figsize=(10, 10))
gdf.plot(column='random_col', scheme='quantiles', k=5, cmap='Blues',
legend=True, legend_kwds=dict(loc=6), ax=ax)

This new "expand image" feature completely change the behavior of the vs code. I understand this feature might be convenient for some users, but I'd like to use code to explicitly set the dpi and save plots.
Please consider adding a toggle for it or at least disable the double-click and restore the "drag to desktop" behavior.
Btw, please please tell me there is a way to disable this "expand image" function.
Not sure what you mean by 'expand image'. Did you mean the little icon that appears on top of the plot?
Do you have a repro for generating the image? It works with the plots I've created.
Oh 'expand image' means the button. Didn't realize that's what the tooltip said.
Why do you want to disable the 'expand image' button? Is it getting in the way somehow?
This might be mac only issue. Could be our dpi conversion isn't working correctly.
The 'expand image' button should only appear when you mouse over the plot.
@rchiodo The thing is this "expand image" completely change the behavior of the plot. Right now if I double click, it will jump to a new window and also I could no longer drag any plots to my desktop directly.
I really really like the way it was... This overcomplicated my workflow a little bit.
Is there any way to disable all of those new features? I checked the settings and didn't find it.
Sorry, there is no _easy_ way to disable this feature. We reworked how Jupyter returns images. Didn't even realize you could drag the images to your desktop. That's pretty cool.
There is a way to disable it in your code/interactive window though.
Run this line of code after your start the interactive window:
%config InlineBackend.figure_format = 'png'
We'll look into adding an official setting.
Sorry that last comment was too quick. Should have said format = 'png'
@rchiodo Thx a lot. The code works.
The png image saving works for me on Windows. Need to test on mac/linux to see if it's a dpi issue or not.
I just tried with another plot. It has the same issue:

And you're running on mac, right?
No... I'm on windows 10. I think the environment data shows that right?
Note to self:
This might work out better than trying to convert the size of the svg ourselves:
https://spin.atomicobject.com/2014/01/21/convert-svg-to-png/
@StevenLi-DS I just submitted a fix for this. Our insider's build will have the fix in about an hour.
There's a new setting: 'enablePlotViewer' that turns on/off the expand image button. With it off you should be able to drag pngs out of the interactive window again.
I also fixed the png scaling when exporting to png if you continue to use the plot viewer.
If this feature will indeed be the default setting in the future, it might be better to set the plot algin to the left than the center in case of the following:

I just saw a new release. Have this 'enablePlotViewer' implemented?
No that was a release for some specific bug fixes. It doesn't have the enablePlotViewer bits in it. That's only in the insider's build.
You can see what was fixed in the changelog for the release branch:
https://github.com/microsoft/vscode-python/blob/release/CHANGELOG.md
Looks like a fix for a/b testing.
There's a new setting: 'enablePlotViewer' that turns on/off the expand image button. With it off you should be able to drag pngs out of the interactive window again.
@rchiodo Hi, man. Do you have an approximation time for this feature implementation? Meanwhile, could you guys consider reopen this issue https://github.com/microsoft/vscode-python/issues/6473?
This has already been implemented in the insider's build. We generally add the 'sign-off' label when something moves to insider's but hasn't shipped in the general release yet.
Issues with the 'data-science parking lot' or 'parking-lot' are in limbo.
Issues with the 'backlog' label are stuff we're working on for the next release.
We close issues when the ship in the general release.
I also reopened the other issue.