A little function for generating parameters for Panda's read_csv from an R data.frame

dtype.R

dtype <- function(s,numpy="np",fields=names(s),pretty=TRUE){
	classes <- sapply(s,class)
	dtypes = character()
	parse_dates = character()

	for(i in 1:length(fields)){
		cls <- classes[i]
		fld <- fields[i]
		if(cls == "numeric"){
			dtypes[i] <- sprintf("%s.float64",numpy)
		}else if(cls == "integer"){
			dtypes[i] <- sprintf("%s.int64",numpy)
		}else if(cls == "factor"){
			dtypes[i] <- "str"
		}else if(cls == "character"){
			dtypes[i] <- "str"
		}else if(cls == "logical"){
			dtypes[i] <- sprintf("%s.bool",numpy)
		}else if(cls == "Date"){
			dtypes[i] <- "str"
			parse_dates[length(parse_dates) + 1] <- fld
		}else{
			stop(sprintf("Don't know how to handle field %s of class %s",fld,cls))
		}
	}

	if(pretty){
		start = "\n    "
		end = "\n    "
	}else{
		start = ""
		end = " "
	}

	out = sprintf("\ndtype={%s%s},\nparse_dates=[%s%s],\n",
			start,paste0(sprintf('"%s":%s,%s',fields,dtypes,end),collapse = ""),
			start,paste0(sprintf('"%s",%s',parse_dates,end),collapse = ""))

	cat(out)
	invisible(out)
}

R DataFrame 2 Pandas.md

When you've got a data.frame in R and want to write it to csv and then import it into a Python Pandas instance, lazy code like this will fail to parse your string and date fields:

<some_file.R>
write.csv(my_data_frame, "my_data.csv" row.names=FALSE)

<another_file.py>
import pandas as pd
my_data = pd.read_csv("my_data.csv")

The soludtion is of course to pass pd.read_csv() a dtype and parse_dates parameters. This little function generates these parameters for you. For example:

> DF = data.frame(widgets=5L,size="large",today=Sys.Date())
> dtype(DF)

dtype={
    "widgets":np.int64,
    "size":str,
    "today":str,
    },
parse_dates=[
    "today",
    ],