Saving Python matplotlib figures with source-code blocks

Question

So I am trying to create a Data Science workflow using org-mode source-blocks. Basically, I need to emulate the jupyter notebook functionality where you create a plot and it shows the figure. Now, following this link, you can do this:

#+begin_src python :results file
import matplotlib, numpy
matplotlib.use('Agg')
import matplotlib.pyplot as plt
fig=plt.figure(figsize=(4,2))
x=numpy.linspace(-15,15)
plt.plot(numpy.sin(x)/x)
fig.tight_layout()
plt.savefig('images/python-matplot-fig.png')
return 'images/python-matplot-fig.png' # return filename to org-mode
#+end_src

The problem with this is that, you have to save the figure manually and then you have to return the path to the figure (Because org is wrapping that code in a function).

That is super awkward to me, because I don't want to be saving the figure manually nor have that return statement at the end, because I want to export these org files to my blog using orig-publish, and that return statement is not valid code outside of the org source blocks code context. I then tried to do this:

#+begin_src python  :file example.png :dir images/ :results file
  import numpy as np
  import matplotlib.pyplot as plt
  import seaborn as sns

  sns.set()  # Enable Seaborn theme

  random_points = np.random.randint(0, 10, (100, 2))

  plt.scatter(random_points[:, 0], random_points[:, 1])
  plt.show()
#+end_src

#+RESULTS:
[[file:images/example.png]]

But the problem is that, it does not save any figure at the images/example.png path. What I would want to do would be that org-mode would save that figure on the path I specified in the header, and then allow me to visualize it in the results.

How could I achieve this? Thank you!

Answer 1

I'm a little late but I just figured out a pretty neat way to do this using :file and without using :var.

As the other anwser already pointed out plt.show() doesn't produce output.

However you can use :results output file instead of :results file (which is the same as :results value file). This means that the data to create the image is taken from stdout instead of the value returned directly by python. Because of this we can use plt.savefig(sys.stdout.buffer) to output the image:

#+BEGIN_SRC python :results output file :file example.png :output-dir images/
  import sys
  import matplotlib.pyplot as plt
  import numpy as np

  x = np.linspace(0, 2 * np.pi, 100)
  y = np.sin(x)
  plt.plot(x, y)
  plt.savefig(sys.stdout.buffer)
#+END_SRC

#+RESULTS:
[[file:images/example.png]]

Answer 2

The basic problem is that plt.show() does not produce anything (other than showing the image as a side-effect and blocking), so there is nothing to save (it actually produces the None object, so that's what you get in the result):

#+begin_src python :results file
  import numpy as np
  import matplotlib.pyplot as plt

  random_points = np.random.randint(0, 10, (100, 2))

  plt.scatter(random_points[:, 0], random_points[:, 1])
  plt.show()
#+end_src

#+RESULTS:
[[file:None]]

The None that you see in the link is the return value of plt.show().

So you have to save the image:

#+begin_src python :results file :var f="images/example.png"
  import numpy as np
  import matplotlib.pyplot as plt

  random_points = np.random.randint(0, 10, (100, 2))

  plt.scatter(random_points[:, 0], random_points[:, 1])
  plt.savefig(f)
#+end_src

#+RESULTS:
[[file:None]]

That still does not work, because savefig() saves the image but still returns None. So we need to change things to return the file name:

#+begin_src python :results file :var f="images/example.png"
  import numpy as np
  import matplotlib.pyplot as plt

  random_points = np.random.randint(0, 10, (100, 2))

  plt.scatter(random_points[:, 0], random_points[:, 1])
  plt.savefig(f)
  return f
#+end_src

#+RESULTS:
[[file:images/example.png]]

Usign a :var f="..." header allows us to define it in one place and use it in the block. But I don't know how to avoid that and use :file/:dir headers instead.