Cool. Much improved. 👍

@mdboom thanks for the comments. I think I got them all as they came in 😄

...aaaaaand here's the part where I pitch the idea that we add in an option to pass your own bootstrapper.

Something like:

def boxplots_stats(..., bootstrap_fxn=None):
    if bootstrap_fxn is None:
        bootstrap_fxn = _bootstrap_median

    # ...
    def _compute_conf_interval(data, med, iqr, bootstrap):
        if bootstrap is not None:
            # Do a bootstrap estimate of notch locations.
            # get conf. intervals around median
            CI = bootstrap_fxn(data, N=bootstrap)
            notch_min = CI[0]
            notch_max = CI[1]
        else:
            N = len(data)
            notch_min = med - 1.57 * iqr / np.sqrt(N)
            notch_max = med + 1.57 * iqr / np.sqrt(N)

    # yada yada 

I am starting to think that box plots are a complicated enough topic that they should be spun off into a sub-project (which is allowed to do things like require pandas and scipy). Probably take violin plots with them too.

Isn't that essentially seaborn?

We would also still want basic boxplot/violinplot functionality for those
who do not need to delve into pandas or seaborn.

On Wed, Oct 28, 2015 at 10:28 PM, Elliott Sales de Andrade <
notifications@github.com> wrote:

Isn't that essentially seaborn?

—
Reply to this email directly or view it on GitHub
#5343 (comment)
.

rebased with current master (branch had gotten stale)

tagged this as a bug fix, but I am not sure if that is the correct tag for this.

@tacaswell should I rebase on 1.5.X or 2?

on to master then we will back-port the merge to where ever we decide this will go.

It looks like you committed the conflicts in the SVG files:

.../result_images/test_axes/boxplot_autorange_false_whiskers-expected.svg:470: parser error : StartTag: invalid element name
<<<<<<< HEAD
    ^

Bump -- just gave this another rebase to keep it current with master

Build failures are unrelated. Something's wacky with colorbars, e.g.,
FAIL: matplotlib.tests.test_colorbar.test_colorbar_closed_patch.test

Yes, this came about due to a merge of a different PR last night, I think.

On Thu, Feb 18, 2016 at 12:16 PM, Paul Hobson notifications@github.com
wrote:

Build failures are unrelated. Something's wacky with colorbars, e.g.,
FAIL: matplotlib.tests.test_colorbar.test_colorbar_closed_patch.test

—
Reply to this email directly or view it on GitHub
#5343 (comment)
.

Sorry, I broke all the branches. Merged a PR that passed before we put the zero-tolerance in. There were some regions of those images where the 8bit blue value changed by 1

The failure on appveyor is

FAIL: Github issue #1256 identified a bug in Line.draw method 
---------------------------------------------------------------------- 
Traceback (most recent call last):
  File "C:\conda\envs\test-environment\lib\site-packages\nose\case.py", line 197, in runTest
    self.test(*self.arg)
  File "c:\projects\matplotlib\lib\matplotlib\testing\decorators.py", line 152, in wrapped_callable
    func(*args, **kwargs)
  File "c:\projects\matplotlib\lib\matplotlib\tests\test_lines.py", line 61, in test_invisible_Line_rendering
    assert_true(slowdown_factor < slowdown_threshold)
AssertionError: False is not true 

which is known to be flaky

Bump -- give me a shout if y'all want any changes made to this.

Read through this, other than my one comment 👍.

The docstrings are much better.

Can this get a note in api_changes? To check my understanding:

• this is a minor API change (ex the plot will be different for same inputs) but
  1. only in semi-pathological edge case
  2. the change will make edge case more consistent with normal case
  3. the way it was is considered confusing by those skilled in the art of boxplots
  4. there is a way to get the old plots back

@tacaswell your understanding matches mine.

and to make sure I'm clear -- I don't modify api_changes.rstdirectly, I just add a short, single file: api_changes/2016-03-12-PMH.rst?

Yes. The individual files helps prevent rebase-due-to-doc-conflicts
thrashing.

On Sun, Mar 13, 2016 at 6:41 PM Paul Hobson notifications@github.com
wrote:

@tacaswell https://github.com/tacaswell your understanding matches mine.

and to make sure I'm clear -- I don't modify api_changes.rstdirectly, I
just add a short, single file: api_changes/2016-03-12-PMH.rst?

—
Reply to this email directly or view it on GitHub
#5343 (comment)
.

Backport wasn't clean (conflict in removed svg file) so doing it via #6153

thanks for the merge and help, everyone!