👍 I guess we can ignore the appveyor error?

This interacts with #5772 in a safe way, but we should probably centralize this logic in one place or the other.

just thinking out loud, would it make sense to use a logarithm for the
number of bins? I am imagining edge cases where users want really long time
series. we have had users hitting limits in the agg renderer before with
very big plots.

On Thu, Dec 31, 2015 at 2:30 PM, Thomas A Caswell notifications@github.com
wrote:

This interacts with #5772
#5772 in a safe way, but
we should probably centralize this logic in one place or the other.

—
Reply to this email directly or view it on GitHub
#5724 (comment)
.

On 2015/12/31 9:30 AM, Thomas A Caswell wrote:

This interacts with #5772
#5772 in a safe way, but
we should probably centralize this logic in one place or the other.

If so, the logic definitely belongs in ticker, not in axis. It probably
makes sense for ticker to handle the case of _tick_space == 0.

One could go farther, and argue that axis should only supply the
estimated label size and axis length, and leave everything else to the
ticker.

On 2015/12/31 9:44 AM, Benjamin Root wrote:

just thinking out loud, would it make sense to use a logarithm for the
number of bins? I am imagining edge cases where users want really long time
series. we have had users hitting limits in the agg renderer before with
very big plots.

I don't understand; would you elaborate, please? How would a logarithm
be used?

I will rebase my PR on top of this one.

@WeatherGod, I think using a log of the axis size would add more complexity than it would be worth. This is all for trying to automatically make a plot look decent on the screen. The axis size really can't get all that large. The very large case would be something like a poster; but at that point the user is going to need to specify all sorts of things manually anyway.

Any objections to merging this?

This is a clear and simple improvement that needs to go in, so I merged it. I suspect that somewhere along the line the tick strategy might need to be revisited, but that needn't block this incremental improvement.