related to issue #2133

I just came across this issue. I've been using a workaround in the mean time.

>>> ax.set_rgrids(range(-40, 40, 10))
...
ValueError: radial grids must be strictly positive
>>> ax.set_yticks(range(-40, 40, 10))  # Works great!

It's natural to want to plot this:

It's an antenna gain pattern, plotting gain in dB, going down to -25dB in
value at the two nulls. Gridlines are at -10, -20, -30.

This was plotted just by removing the test for rmin <= 0 in
PolarAxes.set_rgrids(), in polar.py

On 7 July 2013 16:53, Thomas A Caswell notifications@github.com wrote:

related to issue #2133#2133

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/2203#issuecomment-20572708
.

Dan Timson
System Architect
CRFS Ltd
Building 7200
Cambridge Research Park
Beach Drive
Cambridge
CB25 9TL

Tel: +44 (0)1223 815615
Direct: +44 (0)1223 815522
e: dtimson@crfs.com
w: www.crfs.com

I would argue that if there is a 'natural' way to deal with negative radius values, it should be via reflection as discussed in the other issue. I don't know what the best solution it is not obvious. @dmcdougall you have thoughts?

Why not something along these lines. Reflect any data that falls below
rmin instead of automatically reflecting only those values less than zero?
Or would it make more sense that it should clip those values?

Personally I have little use for reflected values at all, but it may be counter intuitive to some. For me, I'd advocate clipping data that falls below rmin (just like we would clip data that falls below ymin).

I see polar plots as essentially a Cartesian plot which has been mapped into a circle, or even more preferably an annulus (however I don't see how to create a polar plot in the shape of an annulus here).

I agree - clipping (i.e. if y<rmin, then set y=rmin) is the most
appropriate. For engineering plots of dB vs angle, large negative values
mean negligible signal levels, so it's OK if they all map to the origin of
the polar plot. This gives nice looking plots, as you would want.

On 8 July 2013 19:29, Bradley M. Froehle notifications@github.com wrote:

Personally I have little use for reflected values at all, but it may be
counter intuitive to some. For me, I'd advocate clipping data that falls
below rmin (just like we would clip data that falls below ymin).

I see polar plots as essentially a Cartesian plot which has been mapped
into a circle, or even more preferably an annulus (however I don't see how
to create a polar plot in the shape of an annulus here).

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/2203#issuecomment-20625987
.

Dan Timson
System Architect
CRFS Ltd
Building 7200
Cambridge Research Park
Beach Drive
Cambridge
CB25 9TL

Tel: +44 (0)1223 815615
Direct: +44 (0)1223 815522
e: dtimson@crfs.com
w: www.crfs.com

So, the current behavior (in 1.2.1 and 1.3.0) is to clip values < rmin. If so, what is the value of plotting gridlines at negative values?

@mdboom We are suggesting that in certain fields it is often useful for rmin to be negative.

One way to think about this is that for plotting things are drawn in the physical location (theta, r - rmin) but labeled as r. For example, this plot would have rmin = -15 and rmax = 0.

MEP: https://github.com/matplotlib/matplotlib/wiki/MEP24

@bfroehle @dtimson I have created a MEP to deal with this and would like both of your inputs.

It is highly unlikely this will get done for 1.4.0, punting to 1.5.x.

I've taken a look at what's in MEP24. I don't think you need a specified
minimum value of r: just allow any value, and leave it up to the user to
ensure their dataset takes the values they want to plot. After all, this is
what happens for Cartesian plots.

As I said in an email just over a year ago, all that's needed (I assume) is
to remove the test for rmin <= 0 in PolarAxes.set_rgrids(), in polar.py.

Dan

On 29 June 2014 03:41, Thomas A Caswell notifications@github.com wrote:

MEP: https://github.com/matplotlib/matplotlib/wiki/MEP24

@bfroehle https://github.com/bfroehle @dtimson
https://github.com/dtimson I have created a MEP to deal with this and
would like both of your inputs.

It is highly unlikely this will get done for 1.4.0, punting to 1.5.x.

—
Reply to this email directly or view it on GitHub
#2203 (comment)
.

Dan Timson
System Architect
CRFS Ltd
Building 7200
Cambridge Research Park
Beach Drive
Cambridge
CB25 9TL

Tel: +44 (0)1223 815615
Direct: +44 (0)1223 815522
e: dtimson@crfs.com
w: www.crfs.com

@dtimson I don't understand what you mean. The reason we clip r<=0 is that it negative r values were turning into effective rotations by pi.

This is not a bit of the code base I am familiar with nor use on a regular basis, could you put together a PR showing what you are suggesting?

OK, if you've got people putting in radial coords with negative radii (!),
it would be shame to break their code. So use clipping if you like.

Dan

On 18 July 2014 13:13, Thomas A Caswell notifications@github.com wrote:

@dtimson https://github.com/dtimson I don't understand what you mean.
The reason we clip r<=0 is that it negative r values were turning into
effective rotations by pi.

This is not a bit of the code base I am familiar with nor use on a regular
basis, could you put together a PR showing what you are suggesting?

—
Reply to this email directly or view it on GitHub
#2203 (comment)
.

Dan Timson
System Architect
CRFS Ltd
Building 7200
Cambridge Research Park
Beach Drive
Cambridge
CB25 9TL

Tel: +44 (0)1223 815615
Direct: +44 (0)1223 815522
e: dtimson@crfs.com
w: www.crfs.com

The issue was that we have two competing feature requests here.

1. Treat the radial as a regular cartesian number line that is just simply
   anchored at whatever the minimum r value is (thus allowing negative values
   in situations such as log scale).
2. Treat the negative radials as a rotation by pi.

The original behavior way back when did neither of these, IIRC.
Unfortunately, I don't see how we can implement both as they seem to be
mutually exclusive. I am not sure how to proceed. Perhaps we can get some
inspirations from other projects such as Matlab, R's ggplot, and mathmatica?

Cheers!
Ben Root

On Fri, Jul 18, 2014 at 9:13 AM, dtimson notifications@github.com wrote:

OK, if you've got people putting in radial coords with negative radii (!),
it would be shame to break their code. So use clipping if you like.

Dan

On 18 July 2014 13:13, Thomas A Caswell notifications@github.com wrote:

@dtimson https://github.com/dtimson I don't understand what you mean.
The reason we clip r<=0 is that it negative r values were turning into
effective rotations by pi.

This is not a bit of the code base I am familiar with nor use on a
regular
basis, could you put together a PR showing what you are suggesting?

—
Reply to this email directly or view it on GitHub
<
https://github.com/matplotlib/matplotlib/issues/2203#issuecomment-49423918>

.

Dan Timson
System Architect
CRFS Ltd
Building 7200
Cambridge Research Park
Beach Drive
Cambridge
CB25 9TL

Tel: +44 (0)1223 815615
Direct: +44 (0)1223 815522
e: dtimson@crfs.com
w: www.crfs.com

—

Reply to this email directly or view it on GitHub
#2203 (comment)
.

Is there anyone who actually wants option 2, treating negative radials as a
rotation by pi? If not, let's just go with option 1. allowing negative
values. The option 2 behaviour would in any case be easily achievable by
minor processing of the input data.

Dan

On 18 July 2014 15:01, Benjamin Root notifications@github.com wrote:

The issue was that we have two competing feature requests here.

1. Treat the radial as a regular cartesian number line that is just simply
   anchored at whatever the minimum r value is (thus allowing negative values
   in situations such as log scale).
2. Treat the negative radials as a rotation by pi.

The original behavior way back when did neither of these, IIRC.
Unfortunately, I don't see how we can implement both as they seem to be
mutually exclusive. I am not sure how to proceed. Perhaps we can get some
inspirations from other projects such as Matlab, R's ggplot, and
mathmatica?

Cheers!
Ben Root

On Fri, Jul 18, 2014 at 9:13 AM, dtimson notifications@github.com
wrote:

OK, if you've got people putting in radial coords with negative radii
(!),
it would be shame to break their code. So use clipping if you like.

Dan

On 18 July 2014 13:13, Thomas A Caswell notifications@github.com
wrote:

@dtimson https://github.com/dtimson I don't understand what you
mean.
The reason we clip r<=0 is that it negative r values were turning into
effective rotations by pi.

This is not a bit of the code base I am familiar with nor use on a
regular
basis, could you put together a PR showing what you are suggesting?

—
Reply to this email directly or view it on GitHub
<

https://github.com/matplotlib/matplotlib/issues/2203#issuecomment-49423918>

.

Dan Timson
System Architect
CRFS Ltd
Building 7200
Cambridge Research Park
Beach Drive
Cambridge
CB25 9TL

Tel: +44 (0)1223 815615
Direct: +44 (0)1223 815522
e: dtimson@crfs.com
w: www.crfs.com

—

Reply to this email directly or view it on GitHub
<
https://github.com/matplotlib/matplotlib/issues/2203#issuecomment-49428874>

.

—
Reply to this email directly or view it on GitHub
#2203 (comment)
.

Dan Timson
System Architect
CRFS Ltd
Building 7200
Cambridge Research Park
Beach Drive
Cambridge
CB25 9TL

Tel: +44 (0)1223 815615
Direct: +44 (0)1223 815522
e: dtimson@crfs.com
w: www.crfs.com

Is it possible to have a rminclip={True, False, 'auto'} parameter and support both? If you set the rmin=0 and have clip=False, it goes through the origin and produces an effective rotation by pi. If you have the rmin=-100, and clip=True, then it shows negative dB until they hit -100 and then clips them so they don't go through the origin. 'auto' would be False if rmin == 0 and True otherwise?

In dB polar plots, negative and positive radii produce the same magnitude decibel value, so you sometimes see them show polarity as different colors: http://www.sweetwater.com/insync/media/2013/05/supercardioid.gif

Is there anyone who actually wants option 2, treating negative radials as a rotation by pi?

Isn't that necessary to make things like http://mathworld.wolfram.com/Rose.html ?

@dtimson to answer your question: I, for one, was expecting that negative radial values be plotted as rotations by pi. When using radial plots to show vectors, it is standard that a negative magnitude be shown as a positive vector in the opposite direction.

I think this is now fixed by #4699 @QuLogic ?

Yes, dunno why it wasn't tagged before.