I guess its more a question than a feature request:

The data:

data1 = np.random.normal(0, 0.1, 1000)
data2 = np.random.normal(1, 0.4, 1000) + np.linspace(0, 1, 1000)
data3 = 2 + np.random.random(1000) * np.linspace(1, 5, 1000)
data4 = np.random.normal(3, 0.2, 1000) + 0.3 * np.sin(np.linspace(0, 20, 1000))

data = np.vstack([data1, data2, data3, data4]).transpose()

df = pd.DataFrame(data, columns=['data1', 'data2', 'data3', 'data4'])
df.head()

now, plotting in Pandas is done like that:

df.plot(title='Line plot')

The unnamed dataframe-index is automatically used for the x-axis.

Question: Would it be possible to have a similar behaviour in chartify?

ch = chartify.Chart(blank_labels=True, x_axis_type='linear')
ch.plot.line(
   # Data must be sorted by x column
   data_frame=df,
   x_column= "index", # internally do a df.index.get_values() or just the fixed term "index"? 
   y_column="value",
   color_column="variable")

According to the tidy data paradigm, this solution works. But it seeems like a lot of extra miles to get the same result like in pandas:

melted_data = pd.melt(df.reset_index(), 
                      id_vars='index',  
                      value_vars=df.columns)

ch = chartify.Chart(blank_labels=True, x_axis_type='linear')
ch.plot.line(
    data_frame=melted_data,
    x_column= "index",   
    y_column="value",
    color_column="variable")

I have the feeling i won't be the last person to ask exactly this queston about grabbing the index-data. ;-)

Thanks for finding @Danela!

label_test = pd.DataFrame({'value': [.20, .40, .05, .6, .2, .8],
'bucket': [1, 2, 3, 1, 2, 3],
'platform': ['android', 'android', 'android', 'ios', 'ios', 'ios'],
'value2': [1.0, 2.0, 3, 6., 8., 10.]
})
ch = chartify.Chart(x_axis_type='categorical')
ch.plot.bar(label_test, ['bucket', 'platform'], 'value')
ch.plot.text(label_test, ['bucket', 'platform'], 'value', 'value2')
ch.show()

https://readthedocs.org/projects/chartify/builds/

I believe that heatmap lacks an order parameter for categories. The values are sorted the way they occur in the data frame (?), which is not very convenient.

When plotting bar charts the color_column must be contained in the categorical columns. For use cases that include highlighting specific values it's not always preferable to have the color dimension on the axis.

Would be good to look into whether this adjustment should be made to the other categorical plots as well (stacked bar, interval, parallel, scatter, etc)

Is this a BUG REPORT or FEATURE REQUEST?:

Uncomment only one, leave it on its own line:

type: bug

type: feature

Environment:

• Chartify version(s): installed latest via pip install chartify
• Operating System(s): Wind 10
• Python version(s): Anaconda 4.5.4

What happened:
jupyter notebook, tried to import charify

ImportError: DLL load failed: The specified module could not be found.
What you expected to happen:
charitify to import
How to reproduce it (as minimally and precisely as possible):
get the error anytime I try to import charity, I uninstalled it and reinstalled via pip still same error
Anything else we need to know?:
ImportError Traceback (most recent call last)
in ()
----> 1 import chartify

C:\Users\msavage\AppData\Local\Continuum\anaconda3\lib\site-packages\chartify_init_.py in ()
15 # limitations under the License.
16 """Top-level package for chartify."""
---> 17 from chartify._core.chart import Chart
18 from chartify._core.colors import color_palettes
19 from chartify._core.options import options

C:\Users\msavage\AppData\Local\Continuum\anaconda3\lib\site-packages\chartify_core\chart.py in ()
24
25 import bokeh
---> 26 import bokeh.plotting
27 from bokeh.embed import file_html
28

C:\Users\msavage\AppData\Local\Continuum\anaconda3\lib\site-packages\bokeh\plotting_init_.py in ()
1 # this is just for testing, otherwise the figure module is shadowed
2 # by the figure function and inacessible
----> 3 from . import figure as _figure
4
5 # extra imports -- just things to add to 'from bokeh.plotting import'

C:\Users\msavage\AppData\Local\Continuum\anaconda3\lib\site-packages\bokeh\plotting\figure.py in ()
6 from six import string_types
7
----> 8 from ..core.properties import Any, Auto, Either, Enum, Int, List, Seq, Instance, String, Tuple
9 from ..core.enums import HorizontalLocation, MarkerType, VerticalLocation
10 from ..models import ColumnDataSource, Plot, Title, Tool, GraphRenderer

C:\Users\msavage\AppData\Local\Continuum\anaconda3\lib\site-packages\bokeh\core\properties.py in ()
254 from .property.container import RelativeDelta; RelativeDelta
255
--> 256 from .property.dataspec import AngleSpec; AngleSpec
257 from .property.dataspec import ColorSpec; ColorSpec
258 from .property.dataspec import DataSpec; DataSpec

C:\Users\msavage\AppData\Local\Continuum\anaconda3\lib\site-packages\bokeh\core\property\dataspec.py in ()
39 from .instance import Instance
40 from .primitive import Float, String
---> 41 from .visual import FontSize, MarkerType
42
43 #-----------------------------------------------------------------------------

C:\Users\msavage\AppData\Local\Continuum\anaconda3\lib\site-packages\bokeh\core\property\visual.py in ()
28
29 # External imports
---> 30 import PIL.Image
31 from six import string_types
32

C:\Users\msavage\AppData\Local\Continuum\anaconda3\lib\site-packages\PIL\Image.py in ()
56 # Also note that Image.core is not a publicly documented interface,
57 # and should be considered private and subject to change.
---> 58 from . import _imaging as core
59 if PILLOW_VERSION != getattr(core, 'PILLOW_VERSION', None):
60 raise ImportError("The _imaging extension was built for another "

ImportError: DLL load failed: The specified module could not be found.

Currently the _get_svgs and _figure_to_png methods have code copied from bokeh.io.export. I stripped out the wait_until_render_complete functions because I couldn't figure out a quick way to get it to work with chromedriver. Probably worth restoring at some point.

Question seems to come up a lot. Example solution:

outside_groups = ['heavy', 'light', 'etc']
inner_groups = ['US', 'JP', 'CA']
sort_order = list(product(outside_groups, inner_groups))```
Then `categorical_order_by=sort_order`

Hi,

Just an FYI that projects downstream from Bokeh (such as Dask and Holoviews) have begun contributing minimal Bokeh-specific test suites to the Bokeh project to run during our Continuous Integration testing. Certainly we intend not to make any breaking changes at this point, but the "downstream" tests allow us to be extra careful not to disturb any downstream projects. (These tests are currently advisory-only in order to avoid too-tight coupling between project state, but we observe the results regularly).

I'd like to invite you to contribute a test suite for Chartify. The relevant files are here:

https://github.com/bokeh/bokeh/blob/master/scripts/ci/install.downstream
https://github.com/bokeh/bokeh/blob/master/scripts/ci/test.downstream

I'm happy to answer any questions or assist however I can.

Example:

import pandas as pd
import numpy as np
import chartify
df = pd.DataFrame({'values': [1000, 10, 100, 20]})
df['labels'] = df['values']


# breaks
ch = chartify.Chart(x_axis_type='categorical', y_axis_type='log')
ch.plot.bar(df, 'labels', 'values')
ch.show()

# works
factors = [str(x) for x in [1000, 100, 10, 1]]
ch = chartify.Chart(x_axis_type='categorical', y_axis_type='log')
ch.plot._set_categorical_axis_default_factors(True, factors)
ch.figure.vbar(x=factors, top=[1000, 100, 10, 1], bottom=0.1, width=1, fill_color='blue')
ch.show()

Related to:
bokeh/bokeh#6671
bokeh/bokeh#6536

Create a text plot method for dual categorical axes to add move flexibility to heatmap labeling.

Example use case here:
https://stackoverflow.com/questions/54502504/how-to-display-data-from-two-columns-in-chartify-heatmap/54577536#54577536

https://stackoverflow.com/questions/53381384/how-to-add-multiple-y-axis-with-chartify-to-draw-elbow-curves

For vertical legends, the legend sort order is the reverse of the area plotting order. The default order should be changed to be made consistent.

Probably worth checking how this works with other plot types as well:

• Stacked bars
• Area

Getting started area, and the more examples links in the readme are broken when read on https://pypi.org/project/chartify/.

Label orientation method requires bokeh >= 0.12.16

import chartify

Generate example data

data = chartify.examples.example_data()

quantity_by_fruit = (data.groupby('fruit')['quantity'].sum().reset_index())
print(quantity_by_fruit.head())
quantity_by_fruit.iloc[0,0] = 'AAAAAAAAAAAAAAAAAAAA'
ch = chartify.Chart(False, y_axis_type='categorical')
ch.set_title("Vertical bar plot")
ch.plot.bar(
data_frame=quantity_by_fruit,
categorical_columns='fruit',
numeric_column='quantity')
ch.show()

Could you please also publish chartify to conda ? Not just to pip.

Thank you.

Because they don't utilize _named_data_source. This probably applies to some of the other plot types as well.

Currently only accepts lists.

Thanks for finding Dominic Fitzgerald!

Is this a BUG REPORT or FEATURE REQUEST?:

type: bug

Environment:

• Chartify version(s): 2.3.4
• Operating System(s): Windows 10
• Python version(s): 3.7.0

What happened:
If ch.set_legend_location("outside_top") is evaluated after a subtitle is created, it will cover the subtitle. Not an issue if the subtitle is created after ch.set_legend_location("outside_top") .

What you expected to happen:
Expected subtitle to appear

How to reproduce it (as minimally and precisely as possible):
Create a chart with a subtitle then add the legend to "outside_top"

Anything else we need to know?:
It's still possible to have both subtitle and outside_top legend if the subtitle is added afterwards.

ch = chartify.Chart()
ch.figure.circle([1, 2, 3], [1, 2, 3])
ch.set_title('hi')
ch.save('test.html', 'html')
ch.show()

RuntimeError: Models must be owned by only a single document, Circle(id='1668', ...) is already in a doc

https://github.com/authomatic/chromedriver_installer

type: bug

Environment:

• Chartify version(s): 2.3.5
• Operating System(s): MacOS 10.14.2
• Python version(s): 3.6.5

What happened:
Calling line method with all 0 y_column raises OverflowError

What you expected to happen:
Return an instance of chart

How to reproduce it (as minimally and precisely as possible):

import chartify
import pandas as pd

df = pd.DataFrame(data={'col1': [1, 2, 3], 'col2': [0, 0, 0]})

ch = chartify.Chart()
ch.plot.line(data_frame=df,
             x_column='col1',
             y_column='col2')

The plot methods apply default formatting. As a result the user-applied formatting methods can seem to not work if they're applied before plotting and thus overwritten. Make the order of operations clearer, or provide a warning if the user-supplied default is overwritten, or ideally only apply the defaults if there is no user-supplied formatting.

Is this a BUG REPORT or FEATURE REQUEST?:

type: feature

Maybe this is already possible with Chartify and I missed it, but it would be great if Chartify provided an easy way to overlay/add horizontal and vertical lines (of different colors, different dash patterns) to a graph! e.g. to represent levels, or dates of specific events, on the graph.

Hi,

I've an issue regarding how to print the legend with a simple set of data like below:

accel_speed  accel_g1  accel_g2  accel_g3  accel_g4  accel_g5  accel_g6

0 0 0.000000 0.000000 0.000000 0.000000 0.0 0.0
1 1 0.000000 0.000000 0.000000 0.000000 0.0 0.0
2 2 0.000000 0.000000 0.000000 0.000000 0.0 0.0
3 3 0.000000 0.000000 0.000000 0.000000 0.0 0.0
4 4 0.000000 0.000000 0.000000 0.000000 0.0 0.0
5 5 0.000000 0.000000 0.000000 0.000000 0.0 0.0
6 6 2.813216 0.000000 0.000000 0.000000 0.0 0.0
7 7 3.653817 0.000000 0.000000 0.000000 0.0 0.0
8 8 2.807802 0.000000 0.000000 0.000000 0.0 0.0
9 9 2.312193 0.555871 0.000000 0.000000 0.0 0.0
10 10 3.660337 1.154770 0.000000 0.000000 0.0 0.0

that I would like to line plot as accel_speed vs accel_gX, so currently I'm the code below to fill the chart

        for k in data.keys():
            if k != 'accel_speed':
                ch.plot.line(data_frame=df, y_column=k, x_column='accel_speed')

But I don't know how to specify the legend. Finally is there a way to plot such multi column data in one shot such as ch.plot.multi_line()?

Thank you,
Roberto Fichera.

Is there is a way to avoid chromedriver install to just render png output?

Is this possible to switch to another way of generating pngs?

What's advantage of crhomedriver?

ps. I also noticed selenium dependency - how is chartify related to automated testing?

Thanks!

Is this a BUG REPORT or FEATURE REQUEST?:

Not quite sure, but I guess feature

type: feature

• Chartify version(s): 2.3.5

What happened:
I intend to plot an unstacked area chart, to compare different evolutions are respective results. With different areas not having all the same amount of rows.
Meaning, for instance, area A1 can go from 1 to 10 but area A2 may only have values from 1 to 8.
What is currently built now always assumes "nan" values to be considered 0, while this is not true and displays incorrect information.
Attachment 1 will shows the problem above.

What you expected to happen:
I expect the areas to stop their plot when there are no more X values belonging to them, instead of assuming 0.

How to reproduce it (as minimally and precisely as possible):
Create a dataset where one of the group_by columns does not contain the same amount of values as others

If any other information is required please let me know.
Thank you

FEATURE REQUEST

E.g.:

ch.plot.scatter(
    data_frame=price_and_quantity_by_country,
    x_column='total_price',
    y_column='quantity',
    color_column='country')

ch.plot.text(
    data_frame=price_and_quantity_by_country,
    x_column='total_price',
    y_column='quantity',
    text_column='country',
    color_column='country',
    x_offset=1,
    y_offset=-1,
    font_size='10pt')

I understand there is a very good reason to separate the data- and the text-plot, but wouldn't it be more efficient to inject the text-marker-"data" directly into scatter.plot, or any kind of plot?

Like a switch. If text data is present, render the data & text?

Something like this:

ch.plot.scatter(
    data_frame=price_and_quantity_by_country,
    x_column='total_price',
    y_column='quantity',
    color_column='country',
    text_column='country',
    x_offset=1,
    y_offset=-1,
    font_size='10pt')

THIS IS FEATURE REQUEST

Environment:

• Chartify version(s): '2.3.5'
• Operating System(s): 'win32'
• Python version(s): 3.7 (Conda)

I suggest to support the with statement in the following way:

import chartify
import pandas

class Chart(chartify.Chart):
    def __enter__(self):
        return self
    def __exit__(self, type, value, traceback):
        self.show()

df = pandas.DataFrame({'x': [1, 2, 3], 'y': [1, 2, 3]})

with Chart() as ch:
    ch.plot.line(df, 'x', 'y')
    ch.set_title("Example")

This would allow a very simple and clean way of structuring one's plots.

Is this a BUG REPORT or FEATURE REQUEST?:

type: feature

Anything else we need to know?:

A tutorial using mybinder.org (like bokeh's tutorial) makes it much more convenient to test the library before installing it.

Is this a BUG REPORT or FEATURE REQUEST?:

Uncomment only one, leave it on its own line:

type: bug

type: feature

Environment:

• Chartify version(s): 2.3.4
• Operating System(s): Windows 10
• Python version(s): 3.7.0

What happened:
Tried to change the xaxis tick orientation of a plot with x_axis_type = 'datetime'
I am able to change the orientation for other plot types I've used

What you expected to happen:
x-axis tick orientation to change as it does for other types of plot

How to reproduce it (as minimally and precisely as possible):
Create a datetime chart and try to change its axis tick orientation

Anything else we need to know?:

traceback:

AttributeError                            Traceback (most recent call last)
<ipython-input-47-1e689c0d8eb9> in <module>()
     28 monthyearch.axes.set_yaxis_label('Number of Reported Crimes')
     29 # monthyearch.axes.set_xaxis_tick_values(pd.date_range('2014-01-31', '2017-12-31', freq='6M'))
---> 30 monthyearch.axes.set_xaxis_tick_orientation('diagonal')
     31 
     32 

c:\users\simon\appdata\local\programs\python\python37-32\lib\site-packages\chartify\_core\axes.py in set_xaxis_tick_orientation(self, orientation)
    192 
    193         self._chart.figure.xaxis.major_label_orientation = level_1
--> 194         self._chart.figure.xaxis.subgroup_label_orientation = level_2
    195         self._chart.figure.xaxis.group_label_orientation = level_3
    196         return self._chart

c:\users\simon\appdata\local\programs\python\python37-32\lib\site-packages\bokeh\models\plots.py in __setattr__(self, attr, value)
     36     def __setattr__(self, attr, value):
     37         for x in self:
---> 38             setattr(x, attr, value)
     39 
     40     def __dir__(self):

c:\users\simon\appdata\local\programs\python\python37-32\lib\site-packages\bokeh\core\has_props.py in __setattr__(self, name, value)
    287 
    288             raise AttributeError("unexpected attribute '%s' to %s, %s attributes are %s" %
--> 289                 (name, self.__class__.__name__, text, nice_join(matches)))
    290 
    291     def __str__(self):

AttributeError: unexpected attribute 'subgroup_label_orientation' to DatetimeAxis, similar attributes are major_label_orientation

Uncomment only one, leave it on its own line:

type: bug

type: feature

Bokeh added the "BokehJS x.y.z successfully loaded." banner because loading resources from the CDN is the default loading strategy and we wanted to inform users if BokehJS failed to load (which can happen due to networks issues or other reasons). Then users can self-diagnose when their plots don't appear after calling bokeh.io.show.

Since Chartify inlines the Bokeh resources, there isn't the risk to BokehJS not being loaded and I think the banner could be confusing to users via:

• "Why does it say Bokeh? I thought I was using Chartify"
• "It says Bokeh is loaded, but bokeh isn't in my namespace"

All-in-all, it feels cleaner to hide the BokehJS banner.

type: feature

I have a use case where I would like to create a scatter plot that is color coded by a categorical variable, and include some other text in the tooltip. Currently, I am unable to do this with Chartify because the data frame I pass in has extra columns stripped from it, so they are not available to the tooltip. If you could either not strip those columns out, or add a parameter to the scatter plot function that would allow me to keep a text column, such as label_column it would be very helpful.

Here's an example to illustrate the issue:

I have a dataframe, df, that looks like:

I would like t column to appear in a tooltip in a scatter plot with the following code:

import pandas as pd
import chartify
from bokeh.models import HoverTool

df = pd.DataFrame({'x': [1, 2, 3], 
    'y': [4, 5, 6], 
    'l': ['a', 'b', 'c'], 
    't': ['Sunflower', 'Shallow', 'thank you, next']})

ch = chartify.Chart(blank_labels=True)
ch.plot.scatter(
        data_frame=df,
        x_column='x',
        y_column='y',
        color_column='l')
hover = HoverTool(tooltips=[
    ("Title", '@t'),
    ("Cluster", "@l"),
])
ch.figure.add_tools(hover)
ch.show()

But what happens is that the tooltip shows "Title: ???, Cluster: a". As far as I can tell, there is no way to pass in a list of text to HoverTool -- it has to be a part of the ColumnDataSource.

It looks like the internals of PlotNumericXY.scatter is filtering out columns here. Is this necessary? Can it be made optional?

Is there any way to have the charts render on web so I can style and animate them on the fly and potentially make them interactive with data from my database?

Is this a BUG REPORT or FEATURE REQUEST?:
type: feature

Currently, when attempting to use gridplot with Chartify (as a list of charts), it returns "AttributeError: 'Chart' object has no attribute 'select'." When attempting the same with bokeh.layouts.layout, it returns "ValueError: Only LayoutDOM items can be inserted into a layout."

It would be great to have this feature integrated to show/export many chart objects at once.

Is this a BUG REPORT or FEATURE REQUEST?:
type: feature

Box-and-whisker plots are useful tools for statistical analysis and data presentation. It would be very useful if the feature could be added.

PNGs don't look as good as SVG or HTML output.

PNG:

HTML:

Is this a BUG REPORT or FEATURE REQUEST?:

Uncomment only one, leave it on its own line:

type: bug

type: feature

Environment:

• Chartify version(s): beta (pip install)
• Operating System(s): Linux 4.19.3-300.fc29.x86_64 x86_64
• Python version(s): 3.7

What happened:
I'm getting an extra two orders of magnitude (or more if I set a floating point on the tick value) for values with floating points in stacked bar charts.

What you expected to happen:
When setting the y axis tick values to '0%", I should get percentage values ranging from 0-100%.

How to reproduce it (as minimally and precisely as possible):
Plot all chromosome data using stacked bar plot

RdGy = chartify.color_palettes['RdGy']
shifted_RdGy = RdGy.shift_palette('black', percent=20)
shifted_RdGy.show()

(chartify.Chart(blank_labels=True,
x_axis_type='categorical')
.style.set_color_palette('diverging', palette=shifted_RdGy)
.plot.bar_stacked(
data_frame=results,
categorical_columns='#ID',
categorical_order_by='labels',
categorical_order_ascending=True,
numeric_column='Covered_percent',
stack_column='sample',
normalize=False)
.set_legend_location('outside_right', orientation='vertical')
.axes.set_yaxis_tick_format('0.0%')
.axes.set_xaxis_tick_orientation('vertical')
.show('png'))

Anything else we need to know?:
Included the df and ipynb if you want to replicate.
Allen_concatData.txt
pileup.zip

Null values automatically get cast to zero. It would be good to give users the ability to drop them.

Thanks galmeida!

unable to chartify in notebook environment due to below errow.

WebDriverException: Message: 'chromedriver' executable needs to be in PATH. Please see https://sites.google.com/a/chromium.org/chromedriver/home

Is it possible to change the font family for the entire plot?

Uncomment only one, leave it on its own line:

type: bug

type: feature

It would be nice to allow users to specify their own Bokeh resource loading. I believe it would be possible to implement this by having the import chartify invocation defer to a previous call to bokeh.io.output_notebook.

This would enable users to:

• keep notebook sizes small by loading BokehJS from the CDN (or other locations) and not inlining it
• load only the required resources (ie not loading bokeh-widgets.min.js if it's not needed)

This would almost make chartify work in different notebook contexts, by letting users specify
output_notebook(notebook_type='zeppelin')

Example code:

# user specifies their own bokeh/notebook configuration
from bokeh.io import output_notebook
output_notebook()

# load chartify package into namespace but defer to above bokeh.io.output_notebook call
import chartify

Request for Example of stacked horizontal bar chart from pandas groupby object, preferably with red/green for good/bad values for each bar.
Why? I can't get this to work in matplotlib 2.0, which seems to choke on this code

Is this a BUG REPORT or FEATURE REQUEST?:

Uncomment only one, leave it on its own line:

type: feature

Environment:

• Chartify version(s): n/a
• Operating System(s): Win 10
• Python version(s): 3.6.1

What happened:
See https://stackoverflow.com/questions/53583715/trying-to-print-two-color-stacked-barh-using-matplotlib for details.

Note the issue seems to be creating the graph properly from a pandas groupby object.

Here is the unredacted code

if  'Incomplete Evidence' in df.Evaluation:    
    gb = df.groupby(group_by).Evaluation.value_counts(
                    ).unstack(level=-1).nlargest(limit, columns='Incomplete Evidence').stack()
elif 'Complete Evidence' in df.Evaluation:
     gb = df.groupby(group_by).Evaluation.value_counts(
                    ).unstack(level=-1).nlargest(limit, columns='Complete Evidence').stack()
else:
     gb = df.groupby(group_by).Evaluation.value_counts()

.# rbar = [x for x in gb.unstack(level=-1)['Incomplete Evidence']]
.# gbar = [x for x in gb.unstack(level=-1)['Complete Evidence']]

. ### Temporarily using a bar chart
chart = gb.unstack(level=-1).plot.bar(color=chart_colors, stacked=True)
.# chart = plt.barh(0, rbar, color='r')
.# chart = plt.barh(1, gbar, left=rbar, color='g')

What you expected to happen:

A normal horizontal stacked bar with red/green from a groupby object.

How to reproduce it (as minimally and precisely as possible):

See code above - no way to create object in matplotlib to work as expected.

Anything else we need to know?:

Otherwise users will get a blank chart with no explanation.

Allow users to create heatmap plots directly from cross-tabbed data without needing to melt the input data.