R: Turning a data frame row into a character vector

Question

Short version:

I do not understand the behaviour of as.character when trying to convert a single row of a data frame to a character vector.

> mydf <- data.frame("myvar1"=c("mystring","2"),"myvar2"=c("mystring","3"))
> mydf # nice!
myvar1   myvar2
1 mystring mystring
2        2        3
> as.character(mydf[1,])
[1] "2" "2"
> as.character(as.vector(mydf[1,]) ) 
[1] "2" "2"

Maybe somebody could give me an explanation for the last 2 output lines and the correct approach? Thanks a lot.

Background/Purpose:

I want to use lre() in order to detect consecutive occurrences of values in a row of a data frame (with columns of different data types).

Problem: lre() requires a vector, vectors require a definite data type (integer, character, factor, ...). My idea here is to turn the data frame row into a character vector to avoid data loss through conversion.

Solution 1

Your data frame columns aren't characters they are factors.

When you create a data frame the default is that characters are factors. You can see this clearly if you select a column

R> mydf[,1]
[1] mystring 2       
Levels: 2 mystring

To avoid this behaviour set the stringsAsFactors argument to FALSE

mydf = data.frame("myvar1"=c("mystring", "2"),
                    "myvar2"=c("mystring", "3"), 
                     stringsAsFactors=FALSE)

You should also look at this question: How to convert a data frame column to numeric type?

Solution 2

Try this:

 mydf <- data.frame("myvar1"=c("mystring","2"),"myvar2"=c("mystring","3"), stringsAsFactors=F)
 as.character(mydf[1,])
 [1] "mystring" "mystring"

Your strings have been coerced into factors, and you have been shown the factor levels.

Comments

[nilsole]

Thank you! I guess have to take some lessons in data types. (y)

[nilsole]

btw cs + @coffeinjunky : Do you think it is a good idea to turn data (integers, floats, ...) into characters in order to avoid data loss through conversion? or does R offer better ways? thx :)

[csgillespie]

R is used by professional statisticians. I don't think you have to worry about data loss - if you do you are probably doing it wrong ;)

[coffeinjunky]

I don't really see any advantages but many disadvantages of converting your data to characters, i.e. I would not advise doing so.

[nilsole]

ok, the conversion is for the purpose of using it for rle() only, in order to assimilate the rows and to detect identical values. Ideas on better ways? :)