Color, zorder, and position of Bar plot

[mattharrison]

I'm experimenting with the new interface and want to recreate this plot:

econ_df = pd.DataFrame({'Type': 'Map,Scatter,Line,Bar,Area,Cluster,Histogram,Joy Plot,Radar,Sankey,Venn'.split(','),
                       'Count': [14, 11, 10, 5, 3, 2, 2, 1, 1, 1, 1]})
econ_df2 = pd.DataFrame({'Type': ['Map']*14 + ['Scatter']*11 + ['Line']*10 + ['Bar']*5 + ['Area']*3 
                        + ['Cluster']* 2 + ['Histogram']*2 + ['Joy Plot']*1 + ['Radar']*1 
                        + ['Sankey']*1 + ['Venn']* 1})

font = 'Roboto'
grey = '#cbd5dc'
dark_grey = '#545454'
fig, ax = plt.subplots(figsize=(4,4), dpi=300)

df = (econ_df
 .sort_values('Count')
)

sns.countplot(data=econ_df2, y='Type', ax=ax, color='#006ba1', zorder=10)
#df.plot.barh(x='Type', y='Count', ax=ax, color='#006ba1', zorder=10, legend=False,
#            width=.5)

# ALSO ADD
ax.set_xlabel('')

# Tick locations
ax.set_xticks(range(0, 15, 2))
ax.set_xticklabels(range(0, 15, 2), fontname=font, color='k')

# X labels
ax.xaxis.set_tick_params(labeltop=True, labelbottom=False, pad=-1.2, labelsize=10,
                        bottom=False, # Get rid of ticks
                        )
ax.set_ylabel('')

# Y labels
ax.yaxis.set_tick_params(left=False, pad=60)
ax.set_yticklabels(labels=df.Type, ha='left', fontname='Roboto', color='k', fontsize=12, fontweight=300)
                  
# Grid
ax.grid(which='major', axis='x', color=grey, linewidth=1.4)

# Spines
for side in 'top,right,bottom'.split(','):
    ax.spines[side].set_visible(False)
    
# Make left spine thicker    
ax.spines['left'].set_linewidth(1.4)

# Move left spine above bars
ax.spines['left'].set_zorder(10)


# Red Lines
left_edge = -0.1
top_edge = 1.11

ax.add_patch(plt.Rectangle((left_edge, top_edge),       # Set location of rectangle by lower left corder
                           0.10,                       # Width of rectangle
                           -0.03,                      # Height of rectangle. Negative so it goes down.
                           facecolor='#E3120B', 
                           transform=fig.transFigure, 
                           clip_on=False, 
                           linewidth=0))

ax.add_patch(plt.Rectangle((left_edge, top_edge),                # Set location of rectangle by lower left corder
                           .98,                       # Width of rectangle
                           -0.005,                      # Height of rectangle. Negative so it goes down.
                           facecolor='#E3120B', 
                           transform=fig.transFigure, 
                           clip_on=False, 
                           linewidth=0))

# Add in title and subtitle
ax.text(x=left_edge, y=1.03, s="Chart chart", transform=fig.transFigure, ha='left', 
        font=font, fontsize=13, weight=1000)

ax.text(x=left_edge, y=.98, s=r"Main chart type in first 51 Graphic detail articles", 
        font=font, transform=fig.transFigure, ha='left', fontsize=11, weight=300)

# Set source text
_ = ax.text(x=left_edge, y=.08, s=r"""Source: $\it{The Economist}$ created by @__mharrison__""", transform=fig.transFigure, 
        font=font, ha='left', fontsize=8, color=dark_grey, weight=300)

My plot looks like this:

When I try to use the new interface:

• The order is changed (should I just sort in Pandas?)
• Not clear how to set the bar color
• The position of the axes on the figure seems to have shifted
• Not sure how I specify zorder to bump the bars above the grid

font = 'Roboto'
grey = '#cbd5dc'
dark_grey = '#545454'
fig, ax = plt.subplots(figsize=(4,4), dpi=300)

(so.Plot(data=econ_df, x='Count', y='Type')
 .add(so.Bar())
 .on(ax)
 .plot()
)

# ALSO ADD
ax.set_xlabel('')

# Tick locations
ax.set_xticks(range(0, 15, 2))
ax.set_xticklabels(range(0, 15, 2), fontname=font, color='k')

# X labels
ax.xaxis.set_tick_params(labeltop=True, labelbottom=False, pad=-1.2, labelsize=10,
                        bottom=False, # Get rid of ticks
                        )
ax.set_ylabel('')

# Y labels
ax.yaxis.set_tick_params(left=False, pad=60)
ax.set_yticklabels(labels=econ_df.Type, ha='left', fontname='Roboto', color='k', fontsize=12, fontweight=300)
                  
# Grid
ax.grid(which='major', axis='x', color=grey, linewidth=1.4)

# Spines
for side in 'top,right,bottom'.split(','):
    ax.spines[side].set_visible(False)
    
# Make left spine thicker    
ax.spines['left'].set_linewidth(1.4)

# Move left spine above bars
ax.spines['left'].set_zorder(10)


# Red Lines
left_edge = -0.1
top_edge = 1.11

ax.add_patch(plt.Rectangle((left_edge, top_edge),       # Set location of rectangle by lower left corder
                           0.10,                       # Width of rectangle
                           -0.03,                      # Height of rectangle. Negative so it goes down.
                           facecolor='#E3120B', 
                           transform=fig.transFigure, 
                           clip_on=False, 
                           linewidth=0))

ax.add_patch(plt.Rectangle((left_edge, top_edge),      
                           .98,                       
                           -0.005,                    
                           facecolor='#E3120B', 
                           transform=fig.transFigure, 
                           clip_on=False, 
                           linewidth=0))

# Add in title and subtitle
ax.text(x=left_edge, y=1.03, s="Chart chart", transform=fig.transFigure, ha='left', 
        font=font, fontsize=13, weight=1000)

ax.text(x=left_edge, y=.98, s=r"Main chart type in first 51 Graphic detail articles", 
        font=font, transform=fig.transFigure, ha='left', fontsize=11, weight=300)

# Set source text
_ = ax.text(x=left_edge, y=.08, s=r"""Source: $\it{The Economist}$ created by @__mharrison__""", transform=fig.transFigure, 
        font=font, ha='left', fontsize=8, color=dark_grey, weight=300)

[mwaskom]

• The order is changed (should I just sort in Pandas?)

I am pretty sure that your first example is labeling the y axis ticks incorrectly here. But if you just mean that the bars go from small to large when descending the y axis, you could do ax.yaxis.set_inverted(True) which is what the seaborn categorical functions do by default (and probably what a Nominal scale will do at some point soon).

• Not clear how to set the bar color

so.Bar(color=...)

• The position of the axes on the figure seems to have shifted

I think you want to add .layout(algo=None); perhaps the default should be to not apply a tight layout when passed an existing matplotlib object.

• Not sure how I specify zorder to bump the bars above the grid

In general,

so.Bar(artist_kws={"zorder": ...})

You could also pass zorder when you call ax.grid to draw the grid over the bars rather than drawing the bars below the grid.

FWIW a lot of the other tweaks you're applying with messy matplotlib code can also be achieved through the Plot interface, not sure how much you care, e.g.

p = (so.Plot(data=econ_df, x='Count', y='Type')
 .add(so.Bar({"zorder": .5}, color="#006ba1"))
 .on(f:=plt.figure())
 .layout(algo=None)
 .label(x="", y="")
 .scale(x=so.Continuous().tick(every=2))
 .theme({
     "axes.facecolor": "w",
     "axes.linewidth": 1.4,
     "axes.edgecolor": ".2",
     "axes.grid.axis": "x",
     "grid.color": "#cbd5dc",
     "grid.linewidth": 1.4,
     **{f"axes.spines.{side}": False for side in ["bottom", "top", "right"]},
 })
 .plot()
)

ax = f.axes[0]
ax.yaxis.set_inverted(True)
ax.xaxis.set_tick_params(labeltop=True, labelbottom=False, pad=-1.2, labelsize=10)
display(p)

I think you might be able to get the x ticks where / how you want them through rc params but I couldn't figure it out.