behavior when plotting no data

[eyurtsev]

Minor issue, but some of the plotting functions work when there's no data and some don't.

# These work
plot([])
scatter([], [])
bar([], [])

# These raise a ValueError
hexbin([], [])
hist([])
hist2d([])

[tacaswell]

I am not convinced that this is an actual issue as the functions that raise ValueError are actually doing computations and then using the results to create artists where as other functions are just creating artists.

A Line2D object with no data points still makes sense (the loops in draw are just no-ops) even if it seems a bit silly. On the other hand, I am not sure what it means to compute the histogram of an empty list (although numpy does let you do this....).

[WeatherGod]

My own personal rule is that no data is still a valid use case (I do lots
of automation). Notice in the recently added quiver3d that we explicitly
handle the no data case prior to any calculation.

IIRC, I personally pushed for numpy's histogram to allow for empty input
arrays. I could have sworn I did the same for mpl, though...

On Thu, Dec 4, 2014 at 1:27 PM, Thomas A Caswell notifications@github.com
wrote:

I am not convinced that this is an actual issue as the functions that
raise ValueError are actually doing computations and then using the
results to create artists where as other functions are just creating
artists.

A Line2D object with no data points still makes sense (the loops in draw
are just no-ops) even if it seems a bit silly. On the other hand, I am not
sure what it means to compute the histogram of an empty list (although
numpy does let you do this....).

—
Reply to this email directly or view it on GitHub
#3886 (comment)
.

[mdboom]

Yeah, I guess I agree with @WeatherGod here: it would be nice to support empty data wherever we can (though we certainly don't currently in all cases).

[eyurtsev]

@tacaswell for a histogram it would be okay to produce an empty plot with the x and y axes, but without doing any binning. This is akin to how I would describe to someone my experimental data: first I would draw the axes, then I populate it with data. :)

[tacaswell]

Yeah, I have come around on this.

[umairidris]

I have made some simple modifications to fix this issue. Please see #4119

Note: I was not able to reproduce this error with hist2d

[tacaswell]

It should support empty data on master now...

On Tue, Feb 17, 2015, 18:34 umairidris notifications@github.com wrote:

I have made some simple modifications to fix this issue. Please see #4119
#4119

Note: I was not able to reproduce this error with hist2d

—
Reply to this email directly or view it on GitHub
#3886 (comment)
.

[tacaswell]

OK, this is odd, I have a clear memory of changing this, but can't find it...

[tacaswell]

#3883 is what I was thinking of and the constraints did to get as relaxed as I thought they did.