1

I have a problem where I would appreciate any ideas on how to do it with R. The problem is this: I have a latex-table stored. The numbers in the table are all equipped with three digits after the decimal point. I want to cut these digits off, leaving the others in the table. (Think of the numbers representing estimation results, but with dimension "dollar". Then, a value of 145.553 does not make much sense, and 145 is enough). The person who created these tables did not think too much about this, so here I go trying to avoid going through the table by hand. :)

So far, I only found different solutions for how to extract numbers from strings, not how to change them so that the string itself is unaltered otherwise.

Example: 

 strings <- c(
 "a.name      & $-436.735  $   & $-710.832$   \\\\", 
 "std(a.name) & $(1403.604)$   & $(1274.283)$ \\\\", 
 )

The solution should return 

 strings <- c(
 "a.name      & $-436  $   & $-710$   \\\\", 
 "std(a.name) & $(1403)$   & $(1274)$ \\\\", 
 )

and, of course, if it was possible to do the rounding correctly, then it would be even better. But this is not of upmost importance.

I tried using gsub with \\.... to identify the strings that contain a period followed by three other numbers, but this also gives me the variable names, a.name in my example.

Does anyone have an idea how I could accomplish what I would like to do?

Thanks in advance!

coffeinjunky
  • 11,254
  • 39
  • 57

2 Answers2

3

This uses base R's gregexpr, regmatches, and regmatches<- to round any number with a decimal part.

It will work correctly even for numbers like .789 (i.e. with no digits before the decimal point) and -0.4 (which should round to a number without a minus sign). The one situation where it might perform less than ideally is that it will not remove the trailing decimal from a number like 10. .

string <- c("a.name      & $-436.735  $   & $-710.832$   \\\\", 
            "std(a.name) & $(1403.604)$   & $(1274.283)$ \\\\")

f <- function(x) {
    pat <- "(-)?[[:digit:]]+\\.[[:digit:]]*"
    m <- gregexpr(pat, x)
    regmatches(x,m) <- lapply(regmatches(x,m), function(X) round(as.numeric(X)))
    x
}

f(string)
# [1] "a.name      & $-437  $   & $-711$   \\\\"
# [2] "std(a.name) & $(1404)$   & $(1274)$ \\\\"
Josh O'Brien
  • 159,210
  • 26
  • 366
  • 455
  • Thanks a lot Josh. This is a really smart solution, and I learn a lot here! – coffeinjunky Nov 05 '13 at 22:10
  • 1
    Glad that helps. The nice thing about this template is that, once you've extracted the matches, you can use the anonymous function in `lapply()` (which here just rounds the numbers) to perform pretty much any manipulations you desire. – Josh O'Brien Nov 05 '13 at 22:19
2
gsub(strings, pattern ="\\.[[:digit:]]{3}", replacement = "")
#[1] "a.name      & $-436  $   & $-710$   \\\\" "std(a.name) & $(1403)$   & $(1274)$ \\\\"

To get the rounding, I'd do something along these lines but the brackets make it a little ugly - 

sapply(
  strsplit(
    strings,
    "\\$|\\$\\(|\\)\\$"
  ),
  function(x) 
    paste(
      x[1],'$',
      ifelse(as.numeric(x[2]) <0, round(as.numeric(x[2]),0),paste0("(",round(as.numeric(x[2]),0),")")),'$',
      x[3],'$',
      ifelse(as.numeric(x[4]) <0, round(as.numeric(x[4]),0),paste0("(",round(as.numeric(x[4]),0),")")),'$',
      x[5]
      )
)
#[1] "a.name      &  $ -437 $    &  $ -711 $    \\\\"   "std(a.name) &  $ (1404) $    &  $ (1274) $  \\\\"
TheComeOnMan
  • 12,535
  • 8
  • 39
  • 54
  • Wow, this is awesome. I have never seen the `[[:digit:]]{3}` part before. I am not quite sure how it works, but it does. Thanks a lot Codoremifa! – coffeinjunky Nov 05 '13 at 19:17
  • Btw, would be happy to get any hints on good resources where I can learn more about regular expressions. If you know any, feel free to share. – coffeinjunky Nov 05 '13 at 19:19
  • 2
    To be honest, I'm learning regular expressions myself by moving around on SO. I was also trying the rounding, etc. but that seems to be a little complicated. See edit. – TheComeOnMan Nov 05 '13 at 19:20