如何在 matplotlib 中以毫米为单位设置轴尺寸和位置？

Question

我需要一个轴宽 80 毫米、高 60 毫米的图形，而标签尺寸为 12 磅。图形大小必须严格裁剪，边界框外没有空间。我该怎么做呢？

另外，我想要两个间距为 10 毫米的轴堆叠。我该怎么做？

---

点是每英寸 72 点或每毫米 72/25.4 点的标准度量。 边界框是包含图形所有“墨水”的最小框。

标签大小很容易，因为它们以磅为单位定义。修复依赖于内容大小的图形大小很难。

提出这个问题的原因是需要创建多个看起来完全相同的出版质量图。此外，通过乳胶渲染，字符的大小和形状可以与正文完全相同。这也适用于PowerPoint演示文稿。一切都归结为每个字体大小的图形大小，必须固定为标准单位。

https://en.wikipedia.org/wiki/Point_(typography)：

1 分（排版）=

SI 单位 352.78×10−6 m 352.778 μm

美国习惯单位（英制单位）

1.1574×10−3 英尺 13.889×10−3 英寸

Answer 1

你（我）不想那样。最好的方法是以毫米为单位定义图形宽度和以毫米为单位的轴宽，并将轴+标签居中。高度无关紧要，应该为空白修剪。

这只是一个开始，缺少居中。下一步是清理代码并使其成为函数（并在菜单中添加“修剪高度”和“修剪宽度”按钮）。

基本上，这将是一个两步过程。第一步，定义轴的大小并用标签绘制图形。在第 2 步中，调用了tightbox 函数并以英寸为单位计算图形的宽度。然后重新调整图形并确定新的轴位置。

import matplotlib
import pylab

matplotlib.rc('text', usetex=True)
matplotlib.rc('figure', dpi=72)
font = {'family' : 'normal',
        'size'   : 10}
matplotlib.rc('font', **font)
matplotlib.rcParams['text.latex.preamble'] = [
       r'\usepackage{lmodern}'    # latin modern, recommended to replace computer modern sans serif
       r'\usepackage{siunitx}',   # i need upright \micro symbols, but you need...
       r'\sisetup{detect-all}',   # ...this to force siunitx to actually use your fonts
       r'\usepackage{helvet}',    # set the normal font here
       r'\usepackage{sansmath}',  # load up the sansmath so that math -> helvet
       r'\sansmath']  # <- tricky! -- gotta actually tell tex to use! 

matplotlib.rcParams['xtick.major.pad'] = 3 # ticklabel spacing between axis and text
matplotlib.rcParams['ytick.major.pad'] = 2 # 

# APS, PRL, Two Columns
column_width_mm     = 86.4581 #mm
column_height_mm    = 2.5*column_width_mm

fig_hspace_mm       = 10 # mm
fig_wspace_mm       = 10 # mm

axis_width_mm       = 69.16648
axis_height_mm      = 43.09763

column_width_inch   = column_width_mm/25.4
column_height_inch  = column_height_mm/25.4

trim_height_below_mm = 0.0
trim_height_above_mm = 0.0

pylab.figure(num = 1, figsize=(column_width_inch, column_height_inch))

ax_0   = pylab.axes([fig_wspace_mm/column_width_mm, (1.0*fig_hspace_mm + 0.0*axis_height_mm)/column_height_mm, axis_width_mm/column_width_mm, axis_height_mm/column_height_mm])
ax_1   = pylab.axes([fig_wspace_mm/column_width_mm, (2.0*fig_hspace_mm + 1.0*axis_height_mm)/column_height_mm, axis_width_mm/column_width_mm, axis_height_mm/column_height_mm])
ax_2   = pylab.axes([fig_wspace_mm/column_width_mm, (3.0*fig_hspace_mm + 2.0*axis_height_mm)/column_height_mm, axis_width_mm/column_width_mm, axis_height_mm/column_height_mm])
ax_3   = pylab.axes([fig_wspace_mm/column_width_mm, (4.0*fig_hspace_mm + 3.0*axis_height_mm)/column_height_mm, axis_width_mm/column_width_mm, axis_height_mm/column_height_mm])

ax_0.set_position([fig_wspace_mm/column_width_mm, (1.0*fig_hspace_mm + 0.0*axis_height_mm - trim_height_below_mm)/column_height_mm, axis_width_mm/column_width_mm, axis_height_mm/column_height_mm])
ax_1.set_position([fig_wspace_mm/column_width_mm, (2.0*fig_hspace_mm + 1.0*axis_height_mm - trim_height_below_mm)/column_height_mm, axis_width_mm/column_width_mm, axis_height_mm/column_height_mm])
ax_2.set_position([fig_wspace_mm/column_width_mm, (3.0*fig_hspace_mm + 2.0*axis_height_mm - trim_height_below_mm)/column_height_mm, axis_width_mm/column_width_mm, axis_height_mm/column_height_mm])
ax_3.set_position([fig_wspace_mm/column_width_mm, (4.0*fig_hspace_mm + 3.0*axis_height_mm - trim_height_below_mm)/column_height_mm, axis_width_mm/column_width_mm, axis_height_mm/column_height_mm])

###########################
# plot figure here
###########################
pylab.sca(ax_0)
pylab.xlabel(r'$\mathrm{10pt~Test~Label~with~huge~symbols:}~\int~~\mathrm{[\frac{m}{s}]}$')
###########################

fig = pylab.gcf()

old_size        = fig.get_size_inches()
old_width_mm    = old_size[0]*25.4
old_height_mm   = old_size[1]*25.4 

bbox3 = ax_3.get_tightbbox(pylab.gcf().canvas.get_renderer())
bbox0 = ax_0.get_tightbbox(pylab.gcf().canvas.get_renderer())

trim_height_below_mm = ((bbox0.ymin)/72.0*25.4)
trim_height_above_mm = old_height_mm - ((bbox3.ymax+4)/72.0*25.4)

new_size        = fig.get_size_inches()
new_width_mm    = new_size[0]*25.4 
new_height_mm   = new_size[1]*25.4 - trim_height_below_mm - trim_height_above_mm

fig.set_size_inches(new_width_mm/25.4, new_height_mm/25.4, num = 1, forward=True)

print new_size, old_size, new_width_mm, new_height_mm
print trim_height_below_mm, trim_height_above_mm

column_width_mm     = new_width_mm
column_height_mm    = new_height_mm 

ax_0.set_position([fig_wspace_mm/column_width_mm, (1.0*fig_hspace_mm + 0.0*axis_height_mm - trim_height_below_mm)/column_height_mm, axis_width_mm/column_width_mm, axis_height_mm/column_height_mm])
ax_1.set_position([fig_wspace_mm/column_width_mm, (2.0*fig_hspace_mm + 1.0*axis_height_mm - trim_height_below_mm)/column_height_mm, axis_width_mm/column_width_mm, axis_height_mm/column_height_mm])
ax_2.set_position([fig_wspace_mm/column_width_mm, (3.0*fig_hspace_mm + 2.0*axis_height_mm - trim_height_below_mm)/column_height_mm, axis_width_mm/column_width_mm, axis_height_mm/column_height_mm])
ax_3.set_position([fig_wspace_mm/column_width_mm, (4.0*fig_hspace_mm + 3.0*axis_height_mm - trim_height_below_mm)/column_height_mm, axis_width_mm/column_width_mm, axis_height_mm/column_height_mm])

pylab.show()