Proper Factor Function

[yhat]

Proper Factor Function

[glamp]

Yeah I was thinking of writing a library with R-Like functions for python. Purely functional style with thinks like factor, table, etc.

We could throw it in ggplot or make it a separate library. I'd probably go with the later.

Thoughts?

On Oct 17, 2013, at 9:09 PM, Stefan Eng [email protected] wrote:

R has the 'factor' function: http://stat.ethz.ch/R-manual/R-devel/library/base/html/factor.html

Is there interest in implementing this so we can write:

p = ggplot(mtcars, aes(factor(cyl), mpg))
p + geom_boxplot()
instead of:

p = ggplot(mtcars, aes('factor(cyl)', mpg))
p + geom_boxplot()
—
Reply to this email directly or view it on GitHub.

[stefaneng]

Yeah that would be a pretty cool idea. Separate library would probably be better. I wouldn't mind working on that at some point.

[glamp]

cool. i'll put something together.

On Thu, Oct 17, 2013 at 9:59 PM, Stefan Eng [email protected]:

Yeah that would be a pretty cool idea. Separate library would probably be
better. I wouldn't mind working on that at some point.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/40#issuecomment-26567422
.

[pzwang]

Nifty. Would be cool to see if we can compile those expression graphs directly with Numba to Blaze kernels.

[jankatins]

I don't think python allows this, as for p = ggplot(mtcars, aes(factor(cyl), mpg)) "cyl" needs to be available as a variable [edit] when aes(...) is called[/edit](as well as mpg! Both are columns in the dataframe: mtcars["mpg"]) . R has some lazy evaluation rules, which makes this possible but I think python won't allow for it.

[edit] As far as I understand R lazy evaluation rules, the function call gets the code passed in and aes can then evaluate the code against the dataset, but in the above example in python aes gets the result of factor(cyl) that would mean that both factor and cyl must be available in that context. [/edit]

I think there are already functions in either pandas or patsy (the library which implements R like formulas for statsmodels) to convert a pandas.Series to factors (see pandas.Categorical). But I'm not sure how that could be used in this case unless you want to NOT pass in a dataframe but the series directly.

What could make sense is using pandas.Categorial internaly in https://github.com/yhat/ggplot/blob/master/ggplot/ggplot.py#L70 or using patsy formulas where ggplot accepts R formulas.

[jankatins]

Also see #188

[jankatins]

Proper factor() would need updates in pandas. see pandas-dev/pandas#5313

[glamp]

fixed in 0.9.3