the very first passing implementation, mainly

• added altair as a dependency
• dropped axis.py
• dropped some altair-specific tests (perhaps should drop more)
• every plot function now returns a "non-interactive" altair chart, therefore I dropped interactive and ax as arguements

Some remarks

• definitely should try "reenact" this in the notebook https://pandas.pydata.org/pandas-docs/stable/visualization.html
• perhaps it's worth considering to add the support for repeat in api - for example, plotting the groupby object, or something alike

the current version of pdvega will not work in JupyterLab: the main reason is that the new MIME-based rendering used by JupyterLab is not yet supported in the vega3 library that pdvega depends on

Just wanted to clarify that this is correct, even with the vega3 jupyterlab extension?

If that is the case I guess this can be kept open to track any progress...

Hey @jakevdp!

This is a really cool project!

This is jumping the gun a bit -- but a new accessor extensions API landed in pandas (dev) (and AccessorProperty will be deprecated). It's really slick! It was designed especially for projects like pdvega.

Here's a PR using the new extension api with backwards compatibility.

Adding figsize support to core plotting functions - same approach as for pdvega.plotting functions.

also, moved warn_if_keywords_unused to the end of each function

I think what's promising about this library is people can use this library to create a plot that people can easily export to share or further customize (or even blend with other plots).

trying to use pdvega like in documentation anytime I get the error message 'module 'pandas.core' has no attribute 'index'' e.g. import numpy as np import pandas as pd import pdvega from vega_datasets import data iris = data.iris() pdvega.andrews_curves(iris, 'species')

I am using python 3.8 I think it is because pandas deprecated index

This works fine.

df = pd.DataFrame(np.random.rand(5, 3))
df.vgplot.scatter(x=0, y=1, c=2)

But, if you want to use color and y based on same column, i.e column 1. It throws

df.vgplot.scatter(x=0, y=1, c=1)

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-40-d247c37c9696> in <module>()
      1 df = pd.DataFrame(np.random.rand(5, 3))
----> 2 df.vgplot.scatter(x=0, y=1, c=1)

d:\apps\anaconda2\lib\site-packages\pdvega\_core.pyc in scatter(self, x, y, c, s, alpha, interactive, width, height, **kwds)
    504         spec = finalize_vegalite_spec(spec, interactive=interactive,
    505                                       width=width, height=height)
--> 506         return Axes(spec, data=data[cols])
    507 
    508     def area(self, x=None, y=None, stacked=True, alpha=None,

d:\apps\anaconda2\lib\site-packages\pdvega\_axes.pyc in __init__(self, spec, data)
      5     """Class representing a pdvega plot axes"""
      6     def __init__(self, spec=None, data=None):
----> 7         self.vlspec = VegaLite(spec, data)
      8 
      9     @property

d:\apps\anaconda2\lib\site-packages\vega3\base.pyc in __init__(self, spec, data)
     21         """Initialize the visualization object."""
     22         spec = utils.nested_update(copy.deepcopy(self.DEFAULTS), spec)
---> 23         self.spec = self._prepare_spec(spec, data)
     24 
     25     def _prepare_spec(self, spec, data):

d:\apps\anaconda2\lib\site-packages\vega3\vegalite.pyc in _prepare_spec(self, spec, data)
     15 
     16     def _prepare_spec(self, spec, data):
---> 17         return prepare_spec(spec, data)
     18 
     19 

d:\apps\anaconda2\lib\site-packages\vega3\utils.pyc in prepare_spec(spec, data)
     86         # We have to do the isinstance test first because we can't
     87         # compare a DataFrame to None.
---> 88         data = sanitize_dataframe(data)
     89         spec['data'] = {'values': data.to_dict(orient='records')}
     90     elif data is None:

d:\apps\anaconda2\lib\site-packages\vega3\utils.pyc in sanitize_dataframe(df)
     64             # For floats, convert nan->None: np.float is not JSON serializable
     65             col = df[col_name].astype(object)
---> 66             df[col_name] = col.where(col.notnull(), None)
     67         elif str(dtype).startswith('datetime'):
     68             # Convert datetimes to strings

e:\github\pandas\pandas\core\frame.pyc in __setitem__(self, key, value)
   2547         else:
   2548             # set column
-> 2549             self._set_item(key, value)
   2550 
   2551     def _setitem_slice(self, key, value):

e:\github\pandas\pandas\core\frame.pyc in _set_item(self, key, value)
   2623         self._ensure_valid_index(value)
   2624         value = self._sanitize_column(key, value)
-> 2625         NDFrame._set_item(self, key, value)
   2626 
   2627         # check if we are modifying a copy

e:\github\pandas\pandas\core\generic.pyc in _set_item(self, key, value)
   2290 
   2291     def _set_item(self, key, value):
-> 2292         self._data.set(key, value)
   2293         self._clear_item_cache()
   2294 

e:\github\pandas\pandas\core\internals.pyc in set(self, item, value, check)
   3992         removed_blknos = []
   3993         for blkno, val_locs in _get_blkno_placements(blknos, len(self.blocks),
-> 3994                                                      group=True):
   3995             blk = self.blocks[blkno]
   3996             blk_locs = blklocs[val_locs.indexer]

e:\github\pandas\pandas\core\internals.pyc in _get_blkno_placements(blknos, blk_count, group)
   5020 
   5021     # FIXME: blk_count is unused, but it may avoid the use of dicts in cython
-> 5022     for blkno, indexer in lib.get_blkno_indexers(blknos, group):
   5023         yield blkno, BlockPlacement(indexer)
   5024 

e:\github\pandas\pandas\_libs\lib.pyx in pandas._libs.lib.get_blkno_indexers()
   1164 @cython.boundscheck(False)
   1165 @cython.wraparound(False)
-> 1166 def get_blkno_indexers(int64_t[:] blknos, bint group=True):
   1167     """
   1168     Enumerate contiguous runs of integers in ndarray.

ValueError: Buffer has wrong number of dimensions (expected 1, got 0)

Is this expected? If not, happy to work on the patch.

flake8 testing of https://github.com/jakevdp/pdvega on Python 2.7.14

$ flake8 . --count --select=E901,E999,F821,F822,F823 --show-source --statistics

./pdvega/_axes.py:64:33: F821 undefined name 'data'
            self._vlspec.data = data
                                ^
./pdvega/_axes.py:66:26: F821 undefined name 'data'
            self._data = data
                         ^
2     F821 undefined name 'data'

I wanted to try this out but didn't want to install it locally, so I made a binder instead.

The README badge will start working once this PR is merged, till then: https://mybinder.org/v2/gh/betatim/pdvega/binder?filepath=examples%2Fpdvega_example.ipynb

I recently brought pdvega to the Anaconda users community through the conda-forge channel. you can now install it with:

  conda install -c conda-forge pdvega

Thanks for this awesome tool.

-Eddie

pdvega could add an entrypoint to register itself with pandas: https://dev.pandas.io/development/extending.html#plotting-backends

Something like

# in setup.py
setup(  # noqa: F821
    ...,
    entry_points={
        "pandas_plotting_backends": [
            "altair = pdvega.<module>",
        ],
    },
)

where <module> is whatever module has the plot top-level method with the right signature.

As a rule, I think we should use Altair code internally rather than dicts... it will make things easier to debug if and when Vega-Lite/Altair changes.

e.g. {'maxbins': 10} should be alt.Bin(maxbins=10) etc.

The plotting library doesn't seem to work when I try and plot a datetime object. It can handle just dates but when there is an associated time the plot builds without error but no line is plotted.

Code here that doesn't work:

import pandas as pd import matplotlib.pyplot as plt import pdvega

rng = pd.date_range('1/1/2011', periods=72, freq='H') rng = [pd.Timestamp(r) for r in rng] ts = pd.Series(np.random.randn(len(rng)), index=rng)

ts.vgplot.line() #this doesn't throw any errors but no data is shown

ts.plot() #this works on the other hand plt.show()

Maybe a bit niche, but ran into this issue with lineplot: if there is a column of all np.nan, then it is ignored, but if there is a column of all None, then it makes the plot really wacky.

Generate some data:

import pandas as pd
import numpy as np
import pdvega
%matplotlib inline

# generate some data
np.random.seed(111)
df = pd.DataFrame(np.random.randn(50, 4), 
        index=pd.date_range('1/1/2000', periods=50),
                  columns=list('ABCD'))
df = df.cumsum()

# this plot is fine
df.vgplot()

# this column is ignored in the plot
df['nan'] = np.nan
df.vgplot()

(looks the same as above)

# this column makes everything weird
df['none'] = None
df.vgplot()

Oddly enough this doesn't happen if the A and B columns are int:

np.random.seed(111)
df = pd.DataFrame(np.random.randint(low=0, high=5, size=[50, 2]), 
        index=pd.date_range('1/1/2000', periods=50),
                  columns=list('AB'))
df = df.cumsum()

# add a column of all none
df['nan'] = np.nan

# add a column of all none
df['none'] = None
df.vgplot()