Errors in the transformation parser: aes(x="np.log(column)", y="Not_Existing")

[jankatins]

The above example gives an error:

gg = ggplot(_df, aes(x="np.log(c_ratio_in)", y="np.log(ppy_ratio_in)")) + geom_point()
gg

 File "<string>", line 1
    data.get('np')data.get('log')(data.get('ppy_ratio_in'))
                     ^
SyntaxError: invalid syntax

-> There should be a better way to detect if something is a column or not -> Check if the potential column name is really in data.columns and only then surround it with the data.get(...)

This should also help against non existing columns:

_df.get("NotExisting")

-> No Error! -> The final plt simple shows nothing :-(

If there is an error in the eval, the error message is not helpfull:

gg = ggplot(_df, aes(x="log(c_ratio_in)", y="log(ppy_ratio_in)")) + geom_point()
gg
[...]
c:\data\external\ggplot\ggplot\ggplot.pyc in _build_df_from_transforms(data, aes)
    481                         item = "data.get('%s')" % item
    482                     lambda_column += item
--> 483                 data[name] = eval(lambda_column)
    484     return data

c:\data\external\ggplot\ggplot\ggplot.pyc in <module>()

TypeError: 'NoneType' object is not callable

-> At least the resulting eval line should be printed and then the exception reraised.

[jankatins]

patsy basically does the same job and it would probably save a lot of time to use patsy for the job: http://patsy.readthedocs.org/en/latest/formulas.html#the-formula-language

Unfortunately I haven't found a way to support strings/factors without getting dummy coding, but numeric columsn work like this:

from patsy import dmatrix
_f = "np.log(c_ratio_in) + np.log(ppy_ratio_in) -1"
dm = dmatrix(_f, _df, return_type="dataframe")

[glamp]

see 0.9.3