R devel - Nov 2004 - Possible bug in read.fwf (PR#7350)

Full_Name: Shigeru Mase
Version: R 1.9.1 and 2.0.0
OS: Debian Linux
Submission from: (NULL) (219.35.150.115)


I tried to read the following text file "test.txt" with 5 lines using 
read.fwf() function:

# comment 1
1234567   # comment 2
1 234567  # comment 3
12345  67 # comment 4
# comment 5

In R 1.9.1, I got the following result (since I have R 2.0.0 installed now 
I cannot reproduce it now but, at least, my former record says so):
> (x <- read.fwf("test.txt", width=c(2,2,3), skip=1, n=4))
   V1 V2  V3
 1 12 34 567
 2  1 23 456
 3 12 34   5

But, in R 2.0.0, the above command seems to run into endless loop, giving no
response. On the other hand, 
> (x <- read.fwf("test.txt", width=c(2,2,3), skip=1))
  V1 V2  V3
1  1 23 456   # skipped first 2 lines
2 12 34   5   # skipped the last comment line

By the way, the following gave the intended result, but skip the last
comment line:
> (x <- read.fwf("test.txt", width=c(2,2,3), skip=0))
  V1 V2  V3
1 12 34 567
2  1 23 456
3 12 34   5

I have been puzzled and, seemingly, I could not find particular indications
of relevent changes in the help document. I searched relevent bug reports on
this site but seems there are none. I got the same behaviour report from a user
of R 2.0.0 under MS Windows XP.

There are two bugs, due to problems in this change (a version of which is
in NEWS).

r30192 2004-07-08 04:05:38 +0100 (Thu, 08 Jul 2004)

multiline records, skipped columns in read.fwf.  Added read.fortran for
fixed format specifications like 3F4.2.

1) If left with 0 lines to read, read.fwf loops away.  
2) 'skip' is passed on to read.table, but the skipped lines have been 
removed.

Fixed for 2.0.1.

On Mon, 8 Nov 2004 mase@is.titech.ac.jp wrote:
> Full_Name: Shigeru Mase
> Version: R 1.9.1 and 2.0.0
> OS: Debian Linux
> Submission from: (NULL) (219.35.150.115)
> 
> 
> I tried to read the following text file "test.txt" with 5 lines
using
> read.fwf() function:
> 
> # comment 1
> 1234567   # comment 2
> 1 234567  # comment 3
> 12345  67 # comment 4
> # comment 5
> 
> In R 1.9.1, I got the following result (since I have R 2.0.0 installed now 
> I cannot reproduce it now but, at least, my former record says so):
> 
> > (x <- read.fwf("test.txt", width=c(2,2,3), skip=1, n=4))
>    V1 V2  V3
>  1 12 34 567
>  2  1 23 456
>  3 12 34   5
> 
> But, in R 2.0.0, the above command seems to run into endless loop, giving
no
> response. On the other hand, 
> 
> > (x <- read.fwf("test.txt", width=c(2,2,3), skip=1))
>   V1 V2  V3
> 1  1 23 456   # skipped first 2 lines
> 2 12 34   5   # skipped the last comment line
> 
> By the way, the following gave the intended result, but skip the last
> comment line:
> 
> > (x <- read.fwf("test.txt", width=c(2,2,3), skip=0))
>   V1 V2  V3
> 1 12 34 567
> 2  1 23 456
> 3 12 34   5
That's correct: comment lines are skipped, and that is what 1.9.1 did.
> I have been puzzled and, seemingly, I could not find particular indications
> of relevent changes in the help document. I searched relevent bug reports
on
> this site but seems there are none. I got the same behaviour report from a
user
> of R 2.0.0 under MS Windows XP.
-- 
Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595