为何seaborn.barplot无需传入plt.figure对象却可应用其尺寸设置？

Great question—this is such a common "wait, how did that happen?" moment when you're just starting out with matplotlib and seaborn, so let's break it down in plain terms:

• Matplotlib's "Default Canvas" System
  When you run fig = plt.figure(figsize=(7,7)), matplotlib doesn't just create a figure object—it sets this new figure as the active/current figure in its internal state. Think of it like laying down a canvas and saying, "Okay, any future drawing I do goes on this one unless I say otherwise."

• Seaborn Plays Nice With Matplotlib's Defaults
  Seaborn is built directly on top of matplotlib, so all its plotting functions (like sns.barplot) follow matplotlib's rules. When you call sns.barplot() without specifying an ax (axes) parameter, seaborn checks:

  1. Is there an active axes (subplot) in the current figure?
  2. If not, it creates a new axes right inside that current figure.
     Either way, it uses the figure you already defined, which is why your bar plot ends up with the (7,7) size you set earlier.

• How to Test or Override This Behavior
  If you want to confirm this is what's happening, run print(plt.gcf()) (short for "get current figure") after creating your fig—you'll see it matches the object you assigned. If you wanted to use a totally different figure, you'd explicitly pass an axes object to seaborn, like this:

  # Create a separate figure and axes
  fig2 = plt.figure(figsize=(10,5))
  ax2 = fig2.add_subplot()
  # Tell seaborn to use this specific axes instead of the default
  sns.barplot(x='quality', y='fixed acidity', data=wineData, ax=ax2)
  

This default behavior is really handy for quick, straightforward plots, but understanding the "current figure" system will save you confusion when you start building more complex multi-panel visualizations!

内容的提问来源于stack exchange，提问作者bhargav_v