Execute line/selection and advance

[cbrnr]

I was wondering if it is possible to execute the current line (or a selection) in the REPL and advance to the next line. Currently, ShiftEnter executes but does not advance.

There was a discussion in microsoft/vscode-jupyter#3897, but although the original issue was identical to my question here, it was closed.

[brettcannon]

We don't have any special integration with the REPL, so there's no specific support from us for that use case (i.e. only the keyboard shortcuts that the REPL itself has are available).

[anthonykim1]

@cbrnr Just got the matter merged, should be available to you like tomorrow!

[cbrnr]

Yes, the list example I've mentioned now works correctly! However, it only works if nothing is selected (i.e., by putting the cursor in one of the lines of this list definition). If I select all lines of this assignment and then hit Shift+Enter, I still get the extra empty line.

[cbrnr]

@anthonykim1 I was wondering if supporting IPython would now be possible with the new smart Shift+Enter feature (see #17172 for the related issue). As I noted previously, this is currently only possible with the following workaround:

1. In the Python terminal session (which opens automatically after the first Shift+Enter), I can immediately exit() and then start ipython --no-autoindent.
2. Now all commands are sent to IPython.

One problem is that this only works correctly with --no-autoindent. If I don't pass this argument, the following code snippet does not get correctly copied to IPython when pressing Shift+Enter:

def f(x):
    if x:
        return "Yes"
    return "No"

This results in the following error:

In [1]: def f(x):
   ...:         if x:
   ...:                     return "Yes"
   ...:                         return "No"
  Cell In[1], line 4
    return "No"
    ^
IndentationError: unexpected indent

Apparently, IPython automatically inserts indentation, which is very useful when manually typing some lines of code. Therefore, --no-autoindent is not a good fix, because users will generally want this behavior.

The other problem is that it's still difficult to use IPython as the (default) interactive interpreter, but that's a different issue.

Do you think this is worth pursuing? In my opinion, IPython offers a dramatically improved interactive experience, so this would be well worth it. And yes, I know that VS Code has its dedicated Jupyter-based interactive mode, but I'd prefer a bare-bones non-Jupyter approach (and probably many others would too). BTW, PyCharm has had the option to use IPython as its default interactive interpreter for years.

[anthonykim1]

Hi @cbrnr. Thanks for your feedback and sharing your experience!
Definitely interesting to see the workaround to use the IPython. I do realize there are multiple people wanting for native IPython support also previously mentioned from: #22139 and #17172 which you can go upvote more, but I'm not really able to guarantee the native support for IPython right away at the moment (one of the reason being the size of IPython being pretty significant).

Speaking of the Jupyter based interactive window, do you use them at all? I'm curious to think what your feedback would be on interactive window in relation to the the Python REPL we have for VS Code.

Also, what do you think makes you prefer IPython usage? Is it the pretty highlighting, magic commands, or something else?

I would love to hear your opinions on them as planning out the roadmap for our REPL rewrite and future plans regarding REPL.

[cbrnr]

Sorry for my late reply @anthonykim1. Regarding IPython, I don't think the extension should ship IPython, but if it can find it in the (active) Python environment, there should be an option to use it. That's how PyCharm implements that feature.

I am aware of the Jupyter-based interactive window, but I don't really like it. Not because the UI is bad (it's great actually), but because running Python via Jupyter can introduce subtle differences in behavior (such as how repr gets a special treatment if a __repr_html__ is available). I really prefer to work with plain interactive Python, because it minimizes the chances of unintended behavior IMO.

And yeah, I prefer IPython over the regular interactive Python interpreter for the reasons you mentioned. It's just so much more convenient, e.g. changing the working directory is just cd and doesn't require an import and os.chdir(). I also like that I can run shell commands by prefixing with !. I also like tiny things, for example getting help is as easy as typing ?sum or sum?, much quicker than help(sum). And you can actually exit with exit (Python tells you to type exit() LOL).

[starball5]

Smart Send is a thing now (see #21838).

Also related on Stack Overflow: VS Code Python move to next line on run ctrl + enter where I've written up an answer post.