Viewrendered3 Ready For Beta Testing

[Thomas Passin]

I have submitted a pull request for the Vewrendered3 plugin.  If anyone wants to try it out (and find all those beta bugs for me), and you don't want for Edward to get into devel, you can get it from the "VR3" branch my GitHub account at

https://github.com/tbpassin/leo-editor

Please note that there are three things to get for this plugin:

1. plugins/viewrendered3.py

2. plugins/viewrendered3/  # get the entire directory (it contains two stylesheets).

3. plugins/qt_gui.py

I recommend setting up a hot key to toggle the rendering pane on and off.  I use the following line in my @shortcuts node in MyLeoSettings.py:

vr3-toggle = Alt+0

This version of the plugin only works with docks.  So if you are using newer versions of Leo 6.2 from devel, you will need to run Leo with --use-docks.

So far as I know, VR3 and VR can be run at the same time.  I bind vr-toggle to Alt+F10.  When both plugins are running, plugin docstrings and help-for-command (F11) will display in VR3.

Note that the @setting to render MathJax is a little different than it is for VR.  This is spelled out in the plugin help.

[Edward K. Ream]

On Tue, Mar 10, 2020 at 12:30 PM Thomas Passin <tbp1...@gmail.com> wrote:

I have submitted a pull request for the Vewrendered3 plugin. 

Please resubmit this PR with it based on devel, not master.

Edward

[Thomas Passin]

I updated devel on my repo a few minutes ago to be level with Leo's devel, and merged VR3 into it.  Then I made a new pull request.  I hope that's what you need.  Otherwise I'll need a bit of handholding...

[Edward K. Ream]

On Wed, Mar 11, 2020 at 8:33 AM Thomas Passin <tbp1...@gmail.com> wrote:

I updated devel on my repo a few minutes ago to be level with Leo's devel, and merged VR3 into it.  Then I made a new pull request.  I hope that's what you need.  Otherwise I'll need a bit of handholding...

Something's not right. There is a large diff against qt_gui.py.

I suggest making a new private branch off of devel, then making the minimal changes needed to get VR3 to work.

Edward

[Thomas Passin]

My clone of the Leo repository already has a devel branch that I had just updated.  Github Desktop seemed confused when I went to merge my private VR3 into my private devel, so I just copied the modified files into devel instead.  Could that have caused what you see?  The actual changes to qt_gui were in only two places, to add viewrendered3 to the list of viewrenderedxx that would be used for displaying plugin docstrings, help-for-command, etc.

Can I just clone a branch of the Leo repo?  I thought - at least, using Github Desktop - that I could only get the whole repo with all its branches.

[Edward K. Ream]

On Wed, Mar 11, 2020 at 12:04 PM Thomas Passin <tbp1...@gmail.com> wrote:

Can I just clone a branch of the Leo repo?  I thought - at least, using Github Desktop - that I could only get the whole repo with all its branches.

I never use Github desktop, so I couldn't say.

Yes, I recommend cloning the Leo repo, and switching to the devel branch. Then add vr3.py and any other needed files. Carefully check the diffs, then file a PR.

Edward

[Thomas Passin]

All right, I'll see if I can do that.  The thing is, I was already working with a clone of the Leo repo.  Github Desktop clones the repo, sets the clone up on your own Github account, and sets it to point to the Leo repo as the upstream - all those actions you would otherwise do one by one, if you remembered all the git commands, get done in one fell swoop.  What I don't know is whether I can re-clone the Leo repo since I've already got one.

Maybe if I start with a new location on my hard drive...

[Matt Wilkie]

Git's branching and merging always confuses me, let alone across forks. Meaning: don't feel bad, you're not alone!

The network graph at https://github.com/tbpassin/leo-editor/network shows your VR3 branch (top black line for me) as starting in the middle of nowhere, not rooted to any Leo branch. Maybe this has something to do with it?

What I don't know is whether I can re-clone the Leo repo since I've already got one. ... Maybe if I start with a new location on my hard drive...

You can create as many clones of the same repo as you can make folders on a hard drive. ;-)  If GH Desktop makes doing that hard perhaps use the command line to create the clone and then try and open it in Desktop?

-matt

[Thomas Passin]

On Thursday, March 12, 2020 at 6:31:16 PM UTC-4, Matt Wilkie wrote:

Git's branching and merging always confuses me, let alone across forks. Meaning: don't feel bad, you're not alone!

The network graph at https://github.com/tbpassin/leo-editor/network shows your VR3 branch (top black line for me) as starting in the middle of nowhere, not rooted to any Leo branch. Maybe this has something to do with it?

Thanks for the encouragement.   I did start a new VR3 branch in my clone. That's what you are supposed to do to work on a new feature. I tried to merge it with the clone's devel branch, but something weird went on and I couldn't see how to do it within Github Desktop. So I just copied the changed and new files into (my clone's) devel branch.  I (think that I remember) committing that before  making the PR.

I'm more used to working with Mercurial Hg.  It would have opened kdiff3 and shown the two versions side by side, highlighting the differences, and I could just walk through the changes, accepting the ones I wanted for the merge.  I'm not sure how to do the equivalent thing in GitHub. Still, I don't see why just copying the final files over into devel (on my clone) wouldn't have worked.

[Matt Wilkie]

Thanks for the encouragement.

 You're welcome :)

   I did start a new VR3 branch in my clone. That's what you are supposed to do to work on a new feature.

Yes.

 

Still, I don't see why just copying the final files over into devel (on my clone) wouldn't have worked.

This is the route I've used several times when I can't figure out how to do X in git properly. That and deleting the whole local repo (after putting my changes somehwere safe) and cloning a totally fresh copy.

FWIW, I've found Git Extensions to be a closer match than GH Desktop to my experience with TortoiseHG Workbench:

https://gitextensions.github.io/ (I wish the GitHub founders had discovered mecurial before git! Or fossil.)

-matt

[Matt Wilkie]

Thomas, I think maybe you have repo directories intermingling and/or not a clean repo. I just tested this locally and it worked (as in, didn't have a large diff):

md code-test
cd code-test
git clone --depth=1 https://github.com/tbpassin/leo-editor.git/ -b VR3 leo-vr3
xcopy leo-vr3\leo\plugins\viewrendered3.py ..\leo-editor\leo\plugins\
xcopy leo-vr3\leo\plugins\viewrendered\* ..\leo-editor\leo\plugins\viewrendered\
xcopy leo-vr3\leo\plugins\qt_gui.py ..\leo-editor\leo\plugins\

The git command does a shallow clone using --depth (for speed, only most recent commit) and grabs only the VR3 branch.

and then explored the repo with Git Extensions and made the commits. Here's a screenshot comparing the commit of qt_gui.py with origin/devel branch:

-matt

[Thomas Passin]

I think that between the time that I modified qt_gui.py and the time that I pushed it to my clone of the devel branch, that it got reformatted in origin/ devel.  Between my VR3 branch and my devel branch, I see many changes in the file diffs - I'm sure they are what Edward noticed - that are minor reformatting differences, like an extra space before or after a "#" character, or an argument broken over to the next line.  This makes me think that someone had run the new automatic reform script over the entire branch.

I completely understand why no one would want to slog through so many of these changes to verify that they are harmless.  I'll work out how to recover, and go on from there.

[Edward K. Ream]

On Fri, Mar 13, 2020 at 9:48 PM Thomas Passin <tbp1...@gmail.com> wrote:

I think that between the time that I modified qt_gui.py and the time that I pushed it to my clone of the devel branch, that it got reformatted in origin/ devel.  Between my VR3 branch and my devel branch, I see many changes in the file diffs - I'm sure they are what Edward noticed - that are minor reformatting differences, like an extra space before or after a "#" character, or an argument broken over to the next line.  This makes me think that someone had run the new automatic reform script over the entire branch.

I completely understand why no one would want to slog through so many of these changes to verify that they are harmless.  I'll work out how to recover, and go on from there.

Thanks. There is still time to make VR3 part of 6.2.

Edward

[Thomas Passin]

Thanks, Ed,

I have not been able to get my repo on GitHub straightened out so that you won't see large diffs in qt_gui.py from all those formatting changes.  So I've attached a zip file with my files and the new plugins/viewrendered3 directory that contains two stylesheets.  Would you please just add them so we can get this done?

[Edward K. Ream]

On Sat, Mar 14, 2020 at 11:05 AM Thomas Passin <tbp1...@gmail.com> wrote:

Thanks, Ed,

I have not been able to get my repo on GitHub straightened out so that you won't see large diffs in qt_gui.py from all those formatting changes.  So I've attached a zip file with my files and the new plugins/viewrendered3 directory that contains two stylesheets.  Would you please just add them so we can get this done?

Will do.

Edward

[Thomas Passin]

Thank you, Edward!

[Edward K. Ream]

​On Sat, Mar 14, 2020 at 3:58 PM Thomas Passin wrote:

Thank you, Edward!

​You're welcome.

Recent revs merge the VR3 code into both the 6.2 branch and the devel branch.

This should be safe because the only real change to leo\plugins\qt_gui.py was to the ​<< load viewrendered plugin >> section.

Your version of qt_gui was way out of date, so I made the change by hand. The new code is:

pc = g.app.pluginsController
# Load viewrendered (and call vr.onCreate) *only* if not already loaded.
if (
    not pc.isLoaded('viewrendered.py')
    and not pc.isLoaded('viewrendered2.py')
    and not pc.isLoaded('viewrendered3.py')
):
    vr = pc.loadOnePlugin('viewrendered.py')
    if vr:
        g.blue('viewrendered plugin loaded.')
        vr.onCreate('tag', {'c': c})

Please test this in either branch and report any problems.

Edward

[Thomas Passin]

On Sunday, March 15, 2020 at 7:08:32 AM UTC-4, Edward K. Ream wrote:

Your version of qt_gui was way out of date, so I made the change by hand. The new code is:

That's so strange, because I got it right from the devel branch and just added my little bit of code.  Anyway, thanks again.

[Edward K. Ream]

Something is very wrong at your end.

I have just pushed some changes to both 6.2 and devel. They fix pylint complaints and related problems:

- g.toUnicode typically does not need a guard.

- The test at line looked backwards to me.

- I changed test against type to tests on isinstance.

  More convenient now that we only use Python 3.

- The diff at line 1351 looked like a mistake, which I fixed.

- The diff at line 2079 fixes a common mistake: defaults for kwargs should not be dicts or lists.

- The diff at line 2894 changed a useless line, arising from '==' instead of '='.

etc.

In future, please run pylint on your code. It will catch these kinds of things.

For now, please carefully review the diffs.

Edward

[Thomas Passin]

On Sunday, March 15, 2020 at 11:35:26 AM UTC-4, Edward K. Ream wrote:

In future, please run pylint on your code. It will catch these kinds of things.

I thought that pyflakes was doing that job now.  It didn't say anything. Well, I didn't run it on qt_gui.py, because I just copied it from the devel branch and added my one line.  I didn't even import it into Leo.  So it's true that I don't know what pyflakes would have reported.

[Thomas Passin]

Now that's funny.  I just imported qt_gui.py in my Leo outline - the same qt_gui.py file I zipped up for you - and neither pyflakes nor pylint found anything to complain about.  Maybe your pylint settings are different from mine.  I *think* I'm using whatever you set up for Leo 6.1.  I don't recall that I changed any of those settings.

[Edward K. Ream]

On Sun, Mar 15, 2020 at 3:46 PM Thomas Passin <tbp1...@gmail.com> wrote:

On Sunday, March 15, 2020 at 11:35:26 AM UTC-4, Edward K. Ream wrote:

In future, please run pylint on your code. It will catch these kinds of things.

I thought that pyflakes was doing that job now. 

Pyflakes and pylint completely separate. Pyflakes does an amazingly good job of catching local binding errors. Pyflakes is fast enough so it can (should!) be run whenever saving a .py file.

Pylint is much slower, and checks different things. I typically run pylint on changed files about once a day.

Edward

[Edward K. Ream]

On Sun, Mar 15, 2020 at 4:23 PM Thomas Passin <tbp1...@gmail.com> wrote:

Now that's funny.  I just imported qt_gui.py in my Leo outline - the same qt_gui.py file I zipped up for you - and neither pyflakes nor pylint found anything to complain about.  Maybe your pylint settings are different from mine.  I *think* I'm using whatever you set up for Leo 6.1.  I don't recall that I changed any of those settings.

The problems were in VR3.py, not qt_gui.py.

Why you are confused about this? Rev a2a482898 clearly changes viewrendered3.py only.

Edward

[lewis]

I do have nbformat installed.
Since build eb6f4ec070 I get this in console log:
----------
Empty setting name in False
set parent @settings
Leo 6.3-devel, devel branch, build eb6f4ec070
2020-03-28 18:58:09 -0500
VR3: *** No Markdown ***
VR3: *** No nbformat
[snip]

However I do not get that message in Leo's log pane. Why are there extra messages in the console log about Markdown and nbformat?

Regards

Lewis

[Edward K. Ream]

Thomas and I are actively working on vr3. These messages are for debugging. They will likely change in the near future.

Edward