Evaluate Single Python Code Blocks in a Quarto File like in R Studio or Jupyter Lab

Question

Quarto is a literate programming framework. It allows you to combine in a single file code (in various programming languages) and text (in markdown).^*

When a Quarto document is rendered, the output from the code is combined with the text into a single document (e.g. pdf or html).

I have installed Quarto-emacs, which does a great job of rendering and live-previewing quarto files. Among other things, I have also installed Polymode, which associates different code and text blocks with their respective major-modes in a single buffer. Now I can write, code, and preview the results on the fly. Thanks, Emacs!

However, whenever I change some python code block, I have to re-render the whole document (by running markdown-preview or markdown-compile). This can take a while. Moreover, I can't run the code block and then continue to experiment in the interpreter (REPL).

So what I want is to run only a selected python code block or the selected and all previous code blocks (like in a Jupyter notebook) in the interpreter (REPL). This works quite well in RStudio, but I would rather stick to Emacs, which in general has great support for Python and R.

It sort of works with R code blocks in Emacs. When I hit C-c b (ess-eval-buffer) on an R code block, a buffer with an R console is opened and the block is executed there. I don't know how to evaluate all previous code blocks, however.

But when I hit C-c c (elpy-shell-send-buffer), the whole buffer is evaluated instead of just the present code block. Of course, this results in a syntax error. When I try C-c C-y w elpy-shell-send-codecell, I get "Not in a codecell".

I also tried several polymode functions:

• polymode-eval-region-or-chunk produces Undefined ‘polymode-eval-region-function’ in buffer test.qmd[python] on python blocks but works on R blocks.
• polymode-eval-buffer, polymode-eval-buffer-from-beg-to-point, and polymode-eval-buffer-from-point-to-end only run the R blocks, not the Python blocks.

Is there any way to make this work as seamlessly as in RStudio?

^{*This is a generalization of R Markdown and similar to both org-mode with org-babel and Juypter notebooks, which Emacs can display and run using ein.}

Answer 1

EDIT

Probably a cleaner way than the original answer is to just define a suitable polymode-eval-region-function. Inspecting again poly-R.el, we find that we can define an appropriate function as follows:

(defun poly-python-eval-region (beg end msg)
    (python-shell-send-region beg end nil msg t))

(defun poly-python-mode-setup ()
  (setq-local polymode-eval-region-function #'poly-python-eval-region))

(add-hook 'python-mode-hook #'poly-python-mode-setup)

Make sure to evaluate the above code before opening the quarto file, as it configures the polymode-eval-region-function via the python-mode-hook.

Also, make sure that you start a python shell (by placing your cursor within the python cell and doing M-x run-python) before calling polymode-eval-chunk.

Subsequently, you should do M-x polymode-eval-chunk from within a python code block, and then Emacs will correctly make use of the poly-python-eval-region function.

END EDIT

Although I haven't tested this with elpy, I guess it should work when using elpy also. From poly-R.el, we find that the region is narrowed by advising ess-eval-buffer (and similar functions) with pm-execute-narrowed-to-span.

However, we can not simply advise python-shell-send-buffer because the save-restriction within that function breaks the 'effect' of the advice. We can solve it by defining a custom my-python-shell-send-buffer:

(defun my-python-shell-send-buffer (&optional send-main msg)
  "See docstring python-shell-send-buffer."
  (interactive (list current-prefix-arg t))
  (python-shell-send-region (point-min) (point-max) send-main msg t))

(advice-add 'my-python-shell-send-buffer :around 'pm-execute-narrowed-to-span)

Now, with the cursor placed inside a python code block, we can simply send the block via M-x my-python-shell-send-buffer (or bind it to some shortcut).

Before we can use this, we first have to do M-x run-python from the .qmd file buffer (but I guess you know that already).

Beware that python-shell-send-buffer only prints output (i.e. from print statements), not return values. But, when sending your example code block, the plot should show 'pop-up '.

Answer 2

Thanks for the tips guys! I just installed quarto-mode today and trying it out now. Your solution is useful and has allowed me to run the Python code in a shell. However, the display of the input is slightly weird for me:

As you can see, the output is there, but the input is weirdly formatted. Do you have the same?

In addition, were you able to get LSP (via eglot or lsp-mode) properly working? I am using eglot and ran into some issues: https://github.com/joaotavora/eglot/issues/1229

Quarto looks great, and quarto-mode looks useful even though it's still relatively new (and unfortunately inactive in terms of development). I'm keen to understand how you deal with these issues and how you use quarto as I'd like to stay with Emacs for this as well.