R help - Dec 2001 - Detecting numerical value in character variable

I have a variable that can have either numeric or character values.
When numeric, I take one action; when not-numeric, I take another action.
Unfortunately, my approaches are awkward, so I look for others' approaches.

To detect a numeric value, I have semi-successfully used two appoaches.
I somewhat simplify here using direct character values like "123"
rather than a variable.
1. !is.na(as.numeric("123"))
   which responds "TRUE", but
      !is.na(as.numeric("abc"))
   responds
      FALSE    #so I know it is not numeric
      Warning message:
      NAs introduced by coercion
   This all works well enough except the error message looks bad 
   when printed, and hints that I use the wrong appoach.

2.
!as.logical(gsub("1","T",gsub("-1","F",as.character(regexpr("[^0-9]","123")))))
   This responds "TRUE" for the string "123" having only
numeric characters.
   However, notice how harsh this is on the reader.
   
   Unfortunately, "regexpr" here responds in -1 and 1 rather than
FALSE and TRUE,
   so this becomes an extra verbose appoach.

My question: CAN ONE BETTER DETECT NUMERIC DATA IN A CHARACTER VARIABLE?
One first imagines trying,
   is.numeric("123")
but this responds FALSE, telling us merely that this is a character string.


This problem arises in an R program I have used for years to balance my
checkbook,
producing 5 lines identical to my bank's statement.
I input my checkbook data from a file with one natural column having entries
like
(excluding # comments),
   3117             #check number
   SALARY:10-1-01   #salary deposited on 10/1/2001
   TRANSF:10-23-01  #transfer between accounts on 10/23/2001
These non-numerical descriptive entries speed balancing my checkbook,
especially when I error.



-- 
Jameson C. Burt, NJ9L   Fairfax, Virginia, USA
jameson at coost.com       http://www.coost.com
(202) 690-0380 (work)
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

On Wed, 5 Dec 2001, Jameson C . Burt wrote:
> I have a variable that can have either numeric or character values.
> When numeric, I take one action; when not-numeric, I take another action.
> Unfortunately, my approaches are awkward, so I look for others'
approaches.
>
> To detect a numeric value, I have semi-successfully used two appoaches.
> I somewhat simplify here using direct character values like "123"
rather than a variable.
> 1. !is.na(as.numeric("123"))
>    which responds "TRUE", but
>       !is.na(as.numeric("abc"))
>    responds
>       FALSE    #so I know it is not numeric
>       Warning message:
>       NAs introduced by coercion
>    This all works well enough except the error message looks bad
>    when printed, and hints that I use the wrong appoach.
That is the best current approach.  Set options(warn=-1) around the piece
of code using it.

Another approach in 1.4.0 (real soon now) is to use type.convert, and
check if the answer is mode "numeric").
>
> 2.
!as.logical(gsub("1","T",gsub("-1","F",as.character(regexpr("[^0-9]","123")))))
>    This responds "TRUE" for the string "123" having
only numeric characters.
>    However, notice how harsh this is on the reader.
Well, numbers can have decimal points in, and you are only testing if
any character is non-numeric.

regexpr("[^\.0-9]","123") == -1

would be pretty good.  This would not allow exponential notation nor Inf
or -Inf, though.
>    Unfortunately, "regexpr" here responds in -1 and 1 rather than
FALSE and TRUE,
>    so this becomes an extra verbose appoach.
See above.
> My question: CAN ONE BETTER DETECT NUMERIC DATA IN A CHARACTER VARIABLE?
> One first imagines trying,
>    is.numeric("123")
> but this responds FALSE, telling us merely that this is a character string.
Correct, as documented.

-- 
Brian D. Ripley,                  ripley at 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 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

Jameson C . Burt <jameson at monumental.com>
writes:
>I have a variable that can have either numeric or character values.
>When numeric, I take one action; when not-numeric, I take another action.
>Unfortunately, my approaches are awkward, so I look for others'
approaches.
>
>To detect a numeric value, I have semi-successfully used two appoaches.
>I somewhat simplify here using direct character values like "123"
rather than a
>variable.
>1. !is.na(as.numeric("123"))
>   which responds "TRUE", but
>      !is.na(as.numeric("abc"))
>   responds
>      FALSE    #so I know it is not numeric
>      Warning message:
>      NAs introduced by coercion
>   This all works well enough except the error message looks bad 
>   when printed, and hints that I use the wrong appoach.
>
>2.
!as.logical(gsub("1","T",gsub("-1","F",as.character(regexpr("[^0-9]","123")))
>))
>   This responds "TRUE" for the string "123" having only
numeric characters.
>   However, notice how harsh this is on the reader.
>   
>   Unfortunately, "regexpr" here responds in -1 and 1 rather than
FALSE and
>TRUE, 
>   so this becomes an extra verbose appoach.
>
>My question: CAN ONE BETTER DETECT NUMERIC DATA IN A CHARACTER VARIABLE?
>One first imagines trying,
>   is.numeric("123")
>but this responds FALSE, telling us merely that this is a character string.
>
>
>This problem arises in an R program I have used for years to balance my 
>checkbook,
>producing 5 lines identical to my bank's statement.
>I input my checkbook data from a file with one natural column having entries
>like
>(excluding # comments),
>   3117             #check number
>   SALARY:10-1-01   #salary deposited on 10/1/2001
>   TRANSF:10-23-01  #transfer between accounts on 10/23/2001
>These non-numerical descriptive entries speed balancing my checkbook,
>especially when I error.
Your first solution is fine:

        a <- c("a", "b", 3, 4, "f")
        b <- as.numeric(a)
        a[!is.na(b)]


but gives warnings. Suppress them with options():

        a <- c("a", "b", 3, 4, "f")
        options(warn = -1)
        b <- as.numeric(a)
        a[!is.na(b)]
        
Remember to reinstate warnings:

        options(warn = 1)

When you are finished. See help(options).

Mark

--
Mark Myatt


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._