[R] Avoiding factors and levels in data frames

[(Ted Harding)]

On 01-Sep-08 08:20:25, ONKELINX, Thierry wrote:
> 
> Try to add options(stringsAsFactors = FALSE) in your Rprofile.site
> (in the etc directory). Using as.is = TRUE seems safer than
> stringsAsFactors = FALSE in the read.fwf function. Because as.is
> is set to FALSE by default and stringsAsFactors is not set.
> 
> HTH,
> 
> Thierry

Can I ask for some elucidation about how the code operates here?
Apparently read.fwf() calls read.table(), and ?read.fwf refers
you to ?read.table for things like 'as.is' and 'stringsAsFactors'.

When I look at the code for read.table, I see in the paramater
list:

function (file, .... , as.is = !stringsAsFactors, ... ,
          stringsAsFactors = default.stringsAsFactors(), ... )

with *no further reference whatever* to 'stringsAsFactors' in the
body of the function. In particular, there is no test that I can
see of whether or not 'stringsAsFactors' has been set by the user
in the call.

The standard result of default.stringsAsFactors() is TRUE.

I've written a tiny test function:

  temp<-function(as.is = !stringsAsFactors,
        stringsAsFactors = default.stringsAsFactors()){
  print(c(as.is=as.is, sAF=stringsAsFactors))
  }

  temp()
# as.is   sAF 
# FALSE  TRUE 

  temp(stringsAsFactors = FALSE)
# as.is   sAF 
#  TRUE FALSE 

  temp(as.is=FALSE,stringsAsFactors = FALSE)
# as.is   sAF 
# FALSE FALSE 

So, if read.table is called with 'as.is=FALSE' (which is the default
set by read.fwf(), with any reference to 'stringsAsFactors' in the
call being part of the "..." which is passed to read.table()), then
read.table will be called with 'as.is=FALSE' regardless of whether
'stringsAsFactors=FALSE' has been set explicitly in calling read.fwf().

The only way to get 'as.is' to be TRUE would be to set it explicitly
in the call to read.fwf() (and in that case one need not bother with
'stringsAsFactors', since its only purpose seems to be to determine
the value of 'as.is'). Or, of course, to set default.stringsAsFactors
to be FALSE; but in many case people will want to have per-case
control over what happens in cases like this.

Well, that's how it seems to me, on reading the code. Is this what
Thierry really means when he says "stringsAsFactors is not set"?

If that is the case, then it seems to indicate some conflict or
inconsistency between read.fwf() and read.table() in this respect.
In any case, it strikes me as something of an undesirable tangle!

With thanks for any comments,
Ted.

> -----Oorspronkelijk bericht-----
> Van: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org]
> Namens Asher Meir
> Verzonden: zondag 31 augustus 2008 11:02
> Aan: r-help at r-project.org
> Onderwerp: [R] Avoiding factors and levels in data frames
> 
> Hello all.
> 
> I am an experienced R user, I have used R for many years for a wide
> variety of applications. However, I keep on running into one obstacle:
> I never want factors or levels in my data frames, but I keep on
> getting them. Is there any way to globally turn this whole feature of
> data frames off? Using options(stringAsFactors=FALSE) does not seem to
> work.
> Alternatively, if I have a data frame with levels, can I just get rid
> of them in that data frame?
> 
> Here is an example: I have a large text file, of which part is in the
> fixed-width tabular form I need. I created a widths vector and a
> column names vector. I then read the file as follows:
> 
> raw1<-read.fwf(fn1,widths=widmax,col.names=headermax,stringsAsFactors=FA
> LSE)
> 
> But raw1 still has factors! It is an old class data frame:
> 
>> is(raw1)
> [1] "data.frame" "oldClass"
> 
> And it still has levels:
>> raw1[1,1]
> [1] Gustav wind
> 229 Levels: - - - - - - -     - - - - WIN       - - - M ... Z INDICATES
> C
> 
> My question is:
> 1. Can I get rid of the levels in raw1?
> 2. Even better -- can I stop it getting read in as a data frame with
> factors?
> 3. Even better -- can I just tell R to never use factors in my data
> frames?
> 
> Or any other solution that occurs to people -- maybe this is the wrong
> way to go about reading in fixed width data in this kind of file.
> 
> I would appreciate any help.
> 
> Asher

--------------------------------------------------------------------
E-Mail: (Ted Harding) <Ted.Harding at manchester.ac.uk>
Fax-to-email: +44 (0)870 094 0861
Date: 01-Sep-08                                       Time: 10:22:55
------------------------------ XFMail ------------------------------