[R] read.table(..., header == FALSE, colClasses = <vector with names attribute>)

[Benjamin Tyner]

Yes, it makes sense now; lesson learned. Thank you both! Sometimes it 
seems that no matter how good the documentation, some useR will 
inevitably (ab)use the code in ways that were never intended by the 
authors. Then when the code and/or documentation changes, it is not 
always obvious to the useR whether the intent of the authors has 
changed, or whether the useR had just been "getting the right answer for 
the wrong reason" all along. In this particular case, the change was 
documented as stemming from a "new feature" (as opposed to a bugfix or 
more stringent argument checking) that might appear to be a non fully 
backwards compatible change. For example one might have the (apparently) 
bad habit of using col.names as a shortcut to rename headers on-the-fly ...

    > getRversion()
    [1] ‘3.2.2’

    > read.table(textConnection("x y\na 3.14"), header = TRUE, 
colClasses = c(x = "character", y = "numeric"), col.names = c("foo", "bar"))
      foo  bar
    1   a 3.14

but indeed, the names attribute has zero effect on the result:

    > read.table(textConnection("x y\na 3.14"), header = TRUE, 
colClasses = c(y = "character", x = "numeric"), col.names = c("foo", "bar"))
      foo  bar
    1   a 3.14

so I agree it is good that we are checking for that now.

Regards
Ben

On 10/24/2017 08:55 AM, Martin Maechler wrote:
>>>>>> Benjamin Tyner <btyner at gmail.com>
>>>>>>      on Tue, 24 Oct 2017 07:21:33 -0400 writes:
>      > Jeff,
>      > Thank you for your reply. The intent was to construct a minimum
>      > reproducible example. The same warning occurs when the 'file' argument
>      > points to a file on disk with a million lines. But you are correct, my
>      > example was slightly malformed and in fact gives an error under R
>      > version 3.2.2. Please allow me to try again; in older versions of R,
>
>      >    > read.table(file = textConnection("a\t3.14"), header = FALSE,
>      > colClasses = c(x = "character", y = "numeric"), sep="\t")
>      >      V1   V2
>      >    1  a 3.14
>
>      > (with no warning). As of version 3.3.0,
>
>      >    > read.table(file = textConnection("a\t3.14"), header = FALSE,
>      > colClasses = c(x = "character", y = "numeric"), sep="\t")
>      >      V1   V2
>      >    1  a 3.14
>      >    Warning message:
>      >    In read.table(file = textConnection("a\t3.14"), header = FALSE,  :
>      >      not all columns named in 'colClasses' exist
>
>      > My intent was not to complain but rather to learn more about best
>      > practices regarding the names attribute.
>
> which is a nice attitude, thank you.
>
> An even shorter MRE (as header=FALSE is default, and the default
> sep="" works, too):
>
>> tt <- read.table(textConnection("a 3.14"), colClasses = c(x="character", y="numeric"))
> Warning message:
> In read.table(file = textConnection("a 3.14"), colClasses = c(x = "character",  :
>    not all columns named in 'colClasses' exist
> If you read in the help page -- you did read that before posting, did you?---
> how 'colClasses' should be specified ,
>
>      colClasses: character.  A vector of classes to be assumed for the
> 	      columns.  If unnamed, recycled as necessary.  If named, names
> 	      are matched with unspecified values being taken to be ‘NA’.
>
> 	      Possible values are ..................
> 	      .........
>
> and the 'x' and 'y' names you used, are matched with the
> colnames ... which on the other hand are "V1" and "V2"  for
> you, and so you provoke a warning.
>
> Once you have read (and understood) the above part of the help
> page, it becomes, easy, no?
>
>> tt <- read.table(textConnection("a 3.14"), colClasses = c("character","numeric"))
>> t2 <- read.table(textConnection("a 3.14"), colClasses=c(x="character",y="numeric"), col.names=c("x","y"))
>> t2
>    x    y
> 1 a 3.14
> i.e., no warning in both of these two cases.
>
> So please, please, PLEASE: at least non-beginners like you *should*
> take the effort to read the help page (and report if these seem
> incomplete or otherwise improvable)...
>
> Best,
> Martin Maechler
> ETH Zurich