Thanks a lot for the PR. Can you give us an example? If this is to go in, it will need an example in the gallery and tests. OTOH maybe a good idea to get some input as to whether this will be of general enough interest to include in the core library.

Unfortunately, I haven't been able to get self-hosting matplotlib working, so I didn't have the confidence to include my tests or dig into the gallery code.

I did post a test script and example of output to the devel mailing list.

I've implemented essentially the same thing twice before - in Java and Matlab. The Java Graphics2D implementation is most similar to this one, but I can't post it online.

There is another open-source implementation of this idea inspired by my Matlab implementation above, but it is not integrated as a PathEffect and doesn't understand the plot's aspect ratio -- so the results are not as pretty.

Overall, I believe this would be reasonably frequently used and it presents little risk of any maintenance burden in the future.

Even if it was not used frequently, I believe it warrants inclusion to serve as an example of what can be accomplished by building on PathEffect.

Please include your script above in the description and the plot. The mailing list seems to have eaten your *.py and some folks on here are not subscribed to the mailing list.

In the end you will have to at least make tests, or find someone to do it for you.

Where is the preferred place to add appropriate demos / tests?

Is it the examples gallery? Such as the brief PathEffects demo. Or perhaps up with the line styles, or the contour plotting?

Or is it in the tests directory? There is a PathEffect test there too. However, since I was unable to bootstrap matplotlib, I have not yet gotten these tests to run -- my attempts all crash with a warning about deprecation of ImageComparisonTest and some error about nose. No idea what to do from there -- it looks like matplotlib deprecated its own test harness...

You could hit us up for more guidance on discourse.matplotlib.com. Not sure what the issue is, but the test definitely run, so there is something missing or incompatible in your install.

The idea looks nice (dunno if it should go to the library -- in which case some tests are needed -- or just be an example, in which case you can skip that but some more docs/examples would still be nice).

@anntzer Thanks for all the style suggestions -- all have been incorporated.

I ended up having to use np.int(np.ceil(x)) instead of math.ceil(x) to ensure an integer type for indexing. If there is a better way to do this, I'm all ears.

@jklymak I think I beat on my local setup enough to run tests and demos. It'll take a couple days to put something pretty together. Thanks for the feedback.

Ah yes, I forgot about the float/int problem. I tend to use math.ceil in that case (yes, contradicting my earlier suggestion to just use np), but with the explicitly prefixed name. int(np.ceil(...)) is fine too (though no need for np.int).

I've pushed up a first attempt at some examples. I still don't have a local matplotlib dev environment set up, so there is a chance I missed something when I pulled this together.
Feedback appreciated.

For the interested, the example now builds with the result here.

Demo looks great! https://25998-1385122-gh.circle-artifacts.com/0/home/circleci/project/doc/build/html/gallery/misc/tickedstroke_demo.html#sphx-glr-gallery-misc-tickedstroke-demo-py

There are some formal errors in the docstring causing sphinx to complain:

/home/circleci/project/lib/matplotlib/patheffects.py:docstring of matplotlib.patheffects.TickedStroke:50: WARNING: Unexpected indentation.
/home/circleci/project/lib/matplotlib/patheffects.py:docstring of matplotlib.patheffects.withTickedStroke:43: WARNING: Unexpected indentation.

Please fix these.

I've fixed all the things you specifically called out, but I'm not sure that covers the above messages -- where did you pull these from? I'd like to see more context as they don't seem to refer to line number in the source, and I don't know what indentation it expects.

They are from the circleCI build output. The numbers are relative to the start of the docstring (flake8 does not know about file line numbers).

They are from the circleCI build output. The numbers are relative to the start of the docstring (flake8 does not know about file line numbers).

I'm still missing something. The docstrings it references don't have nearly that many lines. Perhaps it is characters from the start.

StackOverflow says this is often caused by a missing blank line -- but a root cause is something to do with inheritance of documentation. I certainly don't see what to fix to make this go away.

Finally managed to get doc formatting issue fixed.

Gallery Examples

Docs

@jklymak I've added tests to this PR, I'd appreciate if you review and remove the 'Needs tests' label when you can.

@jklymak Please re-review when you have an opportunity.

Just a couple of small things with the test. If I forget, please ping me for a re-review. Thanks for your hard work on this, it seems like a nice feature...

Thanks for all the constructive feedback.

Looks like effort is pushing towards 3.4.0 now -- do you see anything remaining that would keep this from making that milestone?

Please rebase on top of master. The CI failure is fixed there.

Happy to do so. Should I squash it all down to a single commit, or keep this all spread out?

Should I squash it all down to a single commit, or keep this all spread out?

Please squash.

Apart from thetar and squashing, this is good to go.

Thanks much!

Thanks @ramcdona! Congratulations on your first PR to Matplotlib 🎉 We hope to hear from you again.

Thanks @ramcdona! Congratulations on your first PR to Matplotlib 🎉 We hope to hear from you again.

Thanks for working with me to get this in. I know lots of folks who will be glad to see this included.

We'll see if I have another itch to scratch. There are lots of other PathEffects that could be done -- architecture, drafting, cartography, etc. all have interesting line styles that would be likely be implemented this way rather than just a pattern of dash/dots.