在 matplotlib 轴上设置科学限制后调整指数文本

目前，如果我将 matplotlib y 轴刻度标签设置为科学模式，它会在表单 y 轴顶部给出一个指数1e-5

我想调整这个来阅读r'$\mathregular{10^{-5}}$'这样就可以很好地打印出来。

这是我的示例代码：

# Create a figure and axis
fig, ax = plt.subplots()

# Plot 100 random points 
# the y values of which are very small
ax.scatter(np.random.rand(100), np.random.rand(100)/100000.0)

# Set the y limits appropriately
ax.set_ylim(0, 1/100000.0)

# Change the y ticklabel format to scientific format
ax.ticklabel_format(axis='y', style='sci', scilimits=(-2, 2))

# Get the offset value
offset = ax.yaxis.get_offset_text()

# Print it out
print '1st offset printout: {}'.format(offset)

# Run plt.tight_layout()
plt.tight_layout()

# Print out offset again - you can see the value now!
print '2nd offset printout: {}'.format(offset)

# Change it to latex format
offset.set_text(r'$\mathregular{10^{-5}}$')

# Print it out
print '3rd offset printout: {}'.format(offset)

# Add some text to the middle of the figure just to 
# check that it isn't the latex format that's the problem
ax.text(0.5, 0.5/100000.0, r'$\mathregular{10^{-2}}$')

# And show the figure
plt.show()

我的输出如下所示：

1st offset printout: Text(0,0.5,u'')
2nd offset printout: Text(0,636.933,u'1e\u22125')
3rd offset printout: Text(0,636.933,u'$\\mathregular{10^{-5}}$')

可以找到代码和输出图here http://nbviewer.ipython.org/gist/HappyPenguin/9718c2031413c227200d.

有两个奇怪的地方：一个是我无法覆盖 y 轴顶部的 1e-5 （这是目标），另一个是我必须运行plt.tight_layout()以便将 unicode 值视为偏移量。

谁能告诉我哪里错了？

谢谢

编辑：最初的问题没有明确表示我想自动检测当前计算的指数ticklabel_format。因此，它不应该将设置的字符串传递给偏移文本，而是应该自动检测该值并相应地调整乳胶字符串。

基于@edsmith的答案，一种可能的解决方法是获取偏移文本，将其转换为乳胶字符串，关闭偏移并在轴的顶部添加该字符串。

def format_exponent(ax, axis='y'):

    # Change the ticklabel format to scientific format
    ax.ticklabel_format(axis=axis, style='sci', scilimits=(-2, 2))

    # Get the appropriate axis
    if axis == 'y':
        ax_axis = ax.yaxis
        x_pos = 0.0
        y_pos = 1.0
        horizontalalignment='left'
        verticalalignment='bottom'
    else:
        ax_axis = ax.xaxis
        x_pos = 1.0
        y_pos = -0.05
        horizontalalignment='right'
        verticalalignment='top'

    # Run plt.tight_layout() because otherwise the offset text doesn't update
    plt.tight_layout()
    ##### THIS IS A BUG 
    ##### Well, at least it's sub-optimal because you might not
    ##### want to use tight_layout(). If anyone has a better way of 
    ##### ensuring the offset text is updated appropriately
    ##### please comment!

    # Get the offset value
    offset = ax_axis.get_offset_text().get_text()

    if len(offset) > 0:
        # Get that exponent value and change it into latex format
        minus_sign = u'\u2212'
        expo = np.float(offset.replace(minus_sign, '-').split('e')[-1])
        offset_text = r'x$\mathregular{10^{%d}}$' %expo

        # Turn off the offset text that's calculated automatically
        ax_axis.offsetText.set_visible(False)

        # Add in a text box at the top of the y axis
        ax.text(x_pos, y_pos, offset_text, transform=ax.transAxes,
               horizontalalignment=horizontalalignment,
               verticalalignment=verticalalignment)
    return ax

请注意，您应该能够通过调用来使用偏移文本的位置pos = ax_axis.get_offset_text().get_position()但这些值不是轴单位（它们可能是像素单位 - 感谢@EdSmith - 因此不是很有帮助）。因此我刚刚设置了x_pos and y_pos根据我们正在查看的轴的值。

我还编写了一个小函数来自动检测适当的 x 和 y 限制（尽管我知道 matplotlib 有很多奇特的方法可以做到这一点）。

def get_min_max(x, pad=0.05):
    '''
    Find min and max values such that
    all the data lies within 90% of
    of the axis range
    '''
    r = np.max(x) - np.min(x)
    x_min = np.min(x) - pad * r
    x_max = np.max(x) + pad * r
    return x_min, x_max

因此，要更新问题中的示例（稍作更改以使两个轴都需要指数）：

import matplotlib.pylab as plt
import numpy as np

# Create a figure and axis
fig, ax = plt.subplots()

# Plot 100 random points that are very small
x = np.random.rand(100)/100000.0
y = np.random.rand(100)/100000.0
ax.scatter(x, y)

# Set the x and y limits
x_min, x_max = get_min_max(x)
ax.set_xlim(x_min, x_max)
y_min, y_max = get_min_max(y)    
ax.set_ylim(y_min, y_max)

# Format the exponents nicely
ax = format_exponent(ax, axis='x')
ax = format_exponent(ax, axis='y')

# And show the figure
plt.show()

可以使用带有显示代码输出的 ipython 笔记本的要点here http://nbviewer.ipython.org/gist/HappyPenguin/85d4483bfc686bdbd52b.

我希望这有帮助！