Correctly save column position when moving by lines

[ianhi]

Fixes: https://github.com/axelfahy/jupyterlab-vim/issues/23

Binder link: https://mybinder.org/v2/gh/ianhi/jupyterlab-vim/markdown-cols?urlpath=lab

Here is a test snippet:

#include "syscalls.h"
/* getchar:  simple buffered version */
int getchar(void)
{
  static char buf[BUFSIZ];
  static char *bufp = buf;
  static int n = 0;
  if (n == 0) {  /* buffer is empty */

    n = read(0, buf, sizeof buf);
    bufp = buf;
  }
  return (--n >= 0) ? (unsigned char) *bufp++ : EOF;
}

new behavior

old (broken):

This Pulls in the updated version from the codemirror source https://github.com/codemirror/CodeMirror/blob/9d0f9d19de70abe817e8b8e161034fbd3f907030/keymap/vim.js#L1922

The new version checks for the functions in the switch so I had to pull out the moveByLines function and give it a name. Other than that I just tweaked how the switch works and added some new lines from the codemirror version.

cc @BrendanMartin and @lukashergt

• [x] Add a line to History.md briefly documenting the change.

[ianhi]

closing and reopening to see if I can get that binder bot to comment.

[lukashergt]

closing and reopening to see if I can get that binder bot to comment.

Not sure whether this helps here, but I've found empty commits useful for similar situations:

git commit --allow-empty -m "wakey wakey GitHub Actions"

[ianhi]

Here's the binder link: https://mybinder.org/v2/gh/ianhi/jupyterlab-vim/markdown-cols?urlpath=lab

[ianhi]

I think that the builds are failing because the package.json scripts use jlpm which is jupyterlabs shipped version of yarn, but the build env doesn't install jlab. So there are a few solutions:

1. Install jupyterlab in the build env
2. switch the package.json scripts to use npm

Do you have a preference for either?

Also, are there meant to be automated version bumps for the python - if so I don't think that those are working.

[ianhi]

Also, are there meant to be automated version bumps for the python - if so I don't think that those are working.

Nevermind I was confused, vresion.py uses the package.json i see.

[axelfahy]

I think that the builds are failing because the package.json scripts use jlpm which is jupyterlabs shipped version of yarn, but the build env doesn't install jlab. So there are a few solutions:

1. Install jupyterlab in the build env

2. switch the package.json scripts to use npm

Do you have a preference for either?

I think you are right about jlpm. I don't have any preference, what do you think is cleaner?

[ianhi]

I think you are right about jlpm. I don't have any preference, what do you think is cleaner?

Let's just also install jlab, I think that might be necessary anyway as build:prod calls jupyter labextension build .

[ianhi]

oh hmm, but that makes the npm publish action need jlab as well...

But keeping jlpm will keep us consistent with the cookiecutter which I think is a positive. Do you have anything against combining the two publish actions?

[ianhi]

Actually I'm even more confused, because the build requirments in pyproject.toml really ought to have picked up jlab as a dep :/

[ianhi]

aaah, that's because python setup.py apparently doesn't look inside of pyproject.toml, if we use the fairly new python -m build instead I think it will work.

[ianhi]

@axelfahy I think that should fix it, but theres no way to test without merging first, so lets go for it?

[axelfahy]

@ianhi Yes, let's go for it then, I'll merge.

[axelfahy]

@ianhi By Is it okay to remove the action to publish to npm? Without it, the version installed using the extension manager (directly from jupyterlab) will not be updated, right?

[ianhi]

I added that action into the the single publish action - it should publish to both pypi and npm.