subNonStandardCharacters {Ecfun}

sub nonstandard characters with replacement

Description

First convert to ASCII, stripping standard accents and special characters. Then find the first and last character not in standardCharacters and replace all between them with replacement. For example, a string like "Ruben" where "e" carries an accent and is mangled by some software would become something like "Rub_n" using the default values for standardCharacters and replacement.

Usage

subNonStandardCharacters(x,
   standardCharacters=c(letters, LETTERS, 
      ' ','.', '?', '!', ',', 0:9, '/', '*', 
      '$', '%', '\"', "\'", '-', '+', '&', 
      '_', ';', '(', ')', '[', ']', '\n'),
   replacement='_',
   gsubList=list(list(pattern = 
      '\\\\\\\\|\\\\',
      replacement='\"')), ... )

Arguments

x

character vector in which it is desired to find the first and last character not in standardCharacters and replace that substring by replacement.

standardCharacters

a character vector of acceptable characters to keep.

replacement

a character to replace the substring starting and ending with characters not in standardCharacters.

gsubList

list of lists of pattern and replacement arguments to be called in succession before looking for nonStandardCharacters

...

optional arguments passed to strsplit

Details

1. for(il in 1:length(gsubList)) x <- gsub(gsubList[[il]][["pattern"]], gsubList[[il]][['replacement']], x)

2. x <- stringi::stri_trans_general(x, "Latin-ASCII")

3. nx <- length(x)

4. x. <- strsplit(x, "", ...)

5. for(ix in 1:nx) find the first and last standardCharacters in x.[ix] and substitute replacement for everything in between.

NOTES:

** To find the elements of x that have changed, use either subNonStandardCharacters(x) != x or grep(replacement, subNonStandardCharacters(x)), where replacement is the replacement argument = "_" by default.

** On 13 May 2013 Jeff Newmiller at the University of California, Davis, wrote, 'I think it is a fools errand to think that you can automatically "normalize" arbitrary Unicode characters to an ASCII form that everyone will agree on.' (This was a reply on [e-mail], subject: "Re: [R] Matching names with non-English characters".)

** On 2014-12-15 Ista Zahn suggested stri_trans_general. (This was a reply on [e-mail], subject: "[R] Comparing Latin characters with and without accents?".)

Value

a character vector with everything between the first and last character not in standardCharacters replaced by replacement.

Author(s)

Spencer Graves with thanks to Jeff Newmiller, who described this as a "fool's errand", Milan Bouchet-Valat, who directed me to iconv, and Ista Zahn, who suggested stri_trans_general.

See Also

sub, strsplit, grepNonStandardCharacters, subNonStandardNames subNonStandardNames iconv in the base package does some conversion, but is not consistent across platforms, at least using R 3.1.2 on 2015-01.25. stri_trans_general seems better.

Examples

##
## 1. Consider Names = Ruben, Avila and Jose, where 
##    "e" and "A" in these examples carry an accent.  
##    With the default values for standardCharacters and
##    replacement, these might be converted to something
##    like Rub_n, _vila, and Jos_, with different software
##    possibly mangling the names differently.  (The
##    standard checks for R packages in an English locale 
##    complains about non-ASCII characters, because they
##    are not portable.)
##
nonstdNames <- c('Ra`l', 'Ra`', '`l', 'Torres, Raul',
           "Robert C. \\Bobby\\\\", NA, '', '  ', 
           '$12', '12%')
#  confusion in character sets can create
#  names like Names[2]
Name2 <- subNonStandardCharacters(nonstdNames)
str(Name2)

# check 
Name2. <- c('Ra_l', 'Ra_', '_l', nonstdNames[4],
            'Robert C. "Bobby"', NA, '', '  ', 
            '$12', '12%')
str(Name2.)

all.equal(Name2, Name2.)

##
## 2.  Example from iconv
##
icx <- c("Ekstr\u{f8}m", "J\u{f6}reskog", 
         "bi\u{df}chen Z\u{fc}rcher")
icx2 <- subNonStandardCharacters(icx)

# check 
icx. <- c('Ekstrom', 'Joreskog', 'bisschen Zurcher')

all.equal(icx2, icx.)