在python中使用ggplot绘图库
pyggplot是ggplot的python调用（非移植），ggplot是R中使用广泛的绘图库。
源码：https://github.com/TyberiusPrime/pyggplot#pyggplot
Python中的R调用封装：https://github.com/davidthaler/Python-wrapper-for-R-Forecast
由于rpy2原项目托管在sourceforge已经无法访问，新的github镜像：https://github.com/randy3k/rpy2
一个rpy2的再次封装库：https://github.com/quadrismegistus/RpyD2
pyggplot是 R ggplot2 库的封装。使用了rpy2进行调用，将Pandas数据帧对象传送到R中，然后使用rpy2进行调用。
示例
http://nbviewer.ipython.org/url/tyberiusprime.github.io/pyggplot/pyggplot%20samples.ipynb
安装
通过 PyPI
$ pip install pyggplot
You may be required to update pandas, rpy2, so you may be required to run
$ pip install --upgrade pyggplot
用法
import pandas as pd
import numpy as np
import ggplot
df = pd.DataFrame({'x': np.random.rand(100),
'y': np.random.randn(100),
'group': ['A','B'] * 50})
p = pyggplot.Plot(df)
p.add_scatter('x','y', color='group')
p.render('output.png')
## or if you want to use it in IPython Notebook
# p.render_notebook()
更多使用方法
获取pandas.DataFrame对象，然后add layers使用多个不同的 add_xyz functions (e.g. add_scatter).
参考 ggplot关于layers (geoms)的文档, 然后替换 geom_* with add_*.
访问： http://docs.ggplot2.org/0.9.3.1/index.html
You do not need to separate aesthetics from values - the wrapper will treat a parameter as value if and only if it is not a column name.
(so y = 0 is a value, color = 'blue' is a value - except if you have a column 'blue', then it is a column!. And y = 'value' does not work, but that seems to be a ggplot issue).
当 DataFrame 传送到 R：
行索引转为列，通过 'reset_index',
multi level column indices are flattened by concatenating them with ' ', that is (X, 'mean') becomes 'x mean'.
Error messages are not great - most of them translate to 'one or more columns were not found', but they can appear as a lot of different actual messages such as
argument "env" is missing, with no default
object 'y' not found
object 'dat_0' not found
requires the following missing aesthetics: x
non numeric argument to binary operator
without actually quite pointing at what is strictly the offending value. Also, the error appears when rendering (or printing in the IPython Notebook), not when adding the layer.
Open questions
the stat support is not great - it doesn't easily map into pythonic objects. For now, do your stats in pandas - more powerful anyhow!
how could error messages be improved?
Other ggplots' for python
http://ggplot.yhathq.com/ is a port of ggplot2 for python based on matplotlib - unfortunatly not yet feature complete as of early 2015.
https://github.com/sirrice/pyplot is another wrapper for ggplot closer to R's syntax, and does not rely on rpy2 - calls command line R.