remove non-digits except E+ and E- in string

Question

I have the vector vec that I want to convert to numeric type. Therefore I need to get rid of non-digits (including '+') first. The problem: When I exclude them also the '+' and '-' from 'E+' and 'E-' symbols are also removed.

How can I remove all non-digits except for 'E-', 'E+' and '.' from vec?

vec = c('1234', '+ 42', '1E+4', 'NR 12', '4.5E+04', '8.6E-02')

My approaches:

gsub('[^0-9E.]', '', vec) # removes '-' and '+' in 'E-' and 'E+'

gsub('[^0-9(E\\+).]', '', vec) # includes the '+' from '+ 42' 

My desired output is:

c('1234', '42', '1E+4', '12', '4.5E+04', '8.6E-02')

Answer 1

You may extract the numbers using the following regex:

[-+]?[0-9]*\.?[0-9]+([eE][-+]?[0-9]+)?

Details

• [-+]? - either + or -
• [0-9]* - 0+ digits
• \.? - an optional .
• [0-9]+ - 1+ digits
• ([eE][-+]?[0-9]+)? - an optional capturing group (add ?: after ( to use a non-capturing group) matching 1 or 0 occurrences of
  • [eE] - e or E
  • [-+]? - an optional - or +
  • [0-9]+ - 1 or more digits

R demo:

vec <- c('1234', '+ 42', '1E+4', 'NR 12', '4.5E+04', '8.6E-02')
res <- regmatches(vec, regexpr("[-+]?[0-9]*\\.?[0-9]+([eE][-+]?[0-9]+)?", vec))
unlist(res)
## => [1] "1234"    "42"      "1E+4"    "12"      "4.5E+04" "8.6E-02"

If multiple matches per item in a character vector are expected replace regexpr with gregexpr.

Answer 2

You can change your regex to check if + or - is not preceded by E or e and in that case don't remove them (using look behind and enable perl=TRUE) and otherwise, include + and - in your main character set, so they are removed in any other case with empty string. Try changing your line from this,

gsub('[^0-9E.]', '', vec)

to,

gsub('(?<![Ee])[+-]|[^0-9E.+-]', '', vec, perl=TRUE)