I am 👍 modulo a bit of discussion about the names and the details of the shape.

It is a bit annoying that x -> X make sense but p -> P does not, but I do not really see any better options here.

Maybe make the limbs a little narrower? This look very blocky to me.

Yeah, the options for plus were not great, the silver medallist was #.

I will mock-up some options with various limb sizes.

For filled markers, I think a blocky design makes sense--you have to have room inside to fill it. This PR looks good to me. Using the hash instead of the P might be an improvement.

Here are some marker options for a rough vote.

The labels at the top of the columns indicate the ratio for the breadth of the "limbs".
The markers in the top half have a larger linewidth than those in the bottom half.

plus

x

For a uniform look it should be the same breadth ratio for both markers. The current implementation uses a ratio of 1/3.

My vote: 1/4

I lean towards 1/5, but am ok with 1/4

👍 to using the same ratio for both.

I am sure someone will ask for the ability to control this ratio via rcparams, but lets put off doing that until someone asks for it 😄 .

I think 1/3 is most functional, 1/4 is still acceptable, 1/5 is too skinny. The point of a filled marker is to have a filling and an outline. Make it too skinny, and the marker will have to be very large for there to be enough room for an outline and a fill; and the filling should occupy at least as much area as the outline. My assumption is that normally markers will be small compared to the Axes dimensions. In the examples given here, the smallest ones are larger than I would expect most real-life markers to be.

How about we put this up as a vote on matplotlib-users?

On Tue, Aug 9, 2016 at 2:11 AM, Eric Firing notifications@github.com
wrote:

I think 1/3 is most functional, 1/4 is still acceptable, 1/5 is too
skinny. The point of a filled marker is to have a filling and an outline.
Make it too skinny, and the marker will have to be very large for there to
be enough room for an outline and a fill; and the filling should occupy at
least as much area as the outline. My assumption is that normally markers
will be small compared to the Axes dimensions. In the examples given here,
the smallest ones are larger than I would expect most real-life markers to
be.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
#6907 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AARy-NOitf-ddoZRcdyJKO9n32s8TcPNks5qeBoWgaJpZM4JeFmo
.

@efiring That is a good point that I had not considered.

I'm inclined to merge this as-is. The AppVeyor failure is unrelated. I do not support setting up a vote on the mailing list--way too much fuss for a simple decision. It might be too late to get this into 2.0--I don't think that backporting it is essential.

I used the markers as they are in the PR and I have not wished to alter the limb sizes. I am fine on that front. Otherwise, I do indeed think a hash is a better symbol for the filled plus marker. I will make that
change in the next within a day or so.

one problem with the hash symbol... it can't be used in our current
matplotlibrc reader. We should probably update the reader so that comments
are properly parsed.

On Sat, Sep 10, 2016 at 5:05 AM, Hassan Kibirige notifications@github.com
wrote:

I used the markers as they are in the PR and I have not wished to alter
the limb sizes. I am fine on that front. Otherwise, I do indeed think a
hash is a better symbol for the filled plus marker. I will make that
change in the next within a day or so.

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
#6907 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AARy-K2hAKn4AINKRkAVBV-hxmFajjrHks5qonLBgaJpZM4JeFmo
.

backported to v2.x as 00495d9.
Thank you, @has2k1!