From the issue report, this plot works withkine="line". Why do we need all this logic forscatter but it just work forline?

You are right - I was trying to solve the issue as an edge case, on top of the piece that was causing this inconsistency of coloring between ScatterPlot and LinePlot.

The source of the issue is that matplotlib handles ploting both line and scatter plots. Matplotlib has its own way of color-cycling when no color is defined. Here is an example:

plt.plot([1,2,3], [1,2,3])plt.plot([4,5,6], [4,5,6])plt.plot([7,8,9], [7,8,9])# This results in having 3 different colors, one for each of the plots above

Similarly

plt.scatter(0,0)plt.scatter(1,1)plt.scatter(2,2)# This results in having 3 different colors, one for each of the points above

However, the reason for this inconsistency betweenScatterPlot andLinePlot behaviours is thatLinePlot doesn't pass any color inkwds toax.plot when calling
df.groupby("layer").plot(x='x', y='y', ax= plt.gca(), kind='line')
WhileScatterPlot explicitly defines thec= argument inax.scatter when calling
df.groupby("layer").plot(x='x', y='y', ax= plt.gca(), kind='scatter')

The solution for the source issue is to setc_values = None when no color is passed toScatterPlot; Whenself.c is None and self.color is None

Great! This is much nicer!