Using regular expressions in R to extract information from string - regex

Question

I searched the stack overflow a little and all I found was, that regex in R are a bit tricky and not convenient compared to Perl or Python.
My problem is the following. I have long file names with informations inside. The look like the following:
20150416_QEP1_EXT_GR_1234_hs_IP_NON_060.raw
20150416_QEP1_EXT_GR_1234-1235_hs_IP_NON_060.raw
20150416_QEP1_EXT_GR_1236_hs_IP_NON_060_some_other_info.raw
20150416_QEP1_EXT_GR_1237_hs_IP_NON_060
I want to extract the parts from the filename and convert them conveniently into values, for example the first part is a date, the second is machine abbreviation, the next an institute abbreviation, group abbreviation, sample number(s) etc...
What I do at the moment is constructing a regex, to make (almost) sure, I grep the correct part of the string:
regex <- '([:digit:]{8})_([:alnum:]{1,4})_([:upper:]+)_ etc'
Then I use sub to save each snipped into a variable:
date <- sub(regex, '\\1', filename)
machine <- sub(regex, '\\2', filename)
etc
This works, if the filename has the correct convention. It is overall very hard to read and I am search for a more convenient way of doing the work. I thought about splitting the filename by _ and accessing the string by index might be a good solution. But sometimes, since the filenames often get created by hand, there are terms missing or additional information in the names and I am looking for a better solution to this.
Can anyone suggest a better way of doing so?
EDIT
What I want to create is an object, which has all the information of the filenames extracted and accessible... such as my_object$machine or so....

Answer 1

The help page for ?regex actually gives an example that is exactly equivalent to Python's re.match(r"(?P<first_name>\w+) (?P<last_name>\w+)", "Malcolm Reynolds") (as per your comment):
## named capture
notables <- c(" Ben Franklin and Jefferson Davis",
"\tMillard Fillmore")
#name groups 'first' and 'last'
name.rex <- "(?<first>[[:upper:]][[:lower:]]+) (?<last>[[:upper:]][[:lower:]]+)"
(parsed <- regexpr(name.rex, notables, perl = TRUE))
gregexpr(name.rex, notables, perl = TRUE)[[2]]
parse.one <- function(res, result) {
m <- do.call(rbind, lapply(seq_along(res), function(i) {
if(result[i] == -1) return("")
st <- attr(result, "capture.start")[i, ]
substring(res[i], st, st + attr(result, "capture.length")[i, ] - 1)
}))
colnames(m) <- attr(result, "capture.names")
m
}
parse.one(notables, parsed)
The normal way (i.e. the R way) to extract from a string is the following:
text <- "Malcolm Reynolds"
x <- gregexpr("\\w+", text) #Don't forget to escape the backslash
regmatches(text, x)
[[1]]
[1] "Malcolm" "Reynolds"
You can use however Perl-style group naming by using argument perl=TRUE:
regexpr("(?P<first_name>\\w+) (?P<last_name>\\w+)", text, perl=TRUE)
However regmatches does not support it, hence the need to create your own function to handle that, which is given in the help page:
parse.one <- function(res, result) {
m <- do.call(rbind, lapply(seq_along(res), function(i) {
if(result[i] == -1) return("")
st <- attr(result, "capture.start")[i, ]
substring(res[i], st, st + attr(result, "capture.length")[i, ] - 1)
}))
colnames(m) <- attr(result, "capture.names")
m
}
Applied to your example:
text <- "Malcolm Reynolds"
x <- regexpr("(?P<first_name>\\w+) (?P<last_name>\\w+)", text, perl=TRUE)
parse.one(text, x)
first_name last_name
[1,] "Malcolm" "Reynolds"
To go back to your initial problem:
filenames <- c("20150416_QEP1_EXT_GR_1234_hs_IP_NON_060.raw", "20150416_QEP1_EXT_GR_1234-1235_hs_IP_NON_060.raw", "20150416_QEP1_EXT_GR_1236_hs_IP_NON_060_some_other_info.raw", "20150416_QEP1_EXT_GR_1237_hs_IP_NON_060")
regex <- '(?P<date>[[:digit:]]{8})_(?P<machine>[[:alnum:]]{1,4})_(?P<whatev>[[:upper:]]+)'
x <- regexpr(regex,filenames,perl=TRUE)
parse.one(filenames,x)
date machine whatev
[1,] "20150416" "QEP1" "EXT"
[2,] "20150416" "QEP1" "EXT"
[3,] "20150416" "QEP1" "EXT"
[4,] "20150416" "QEP1" "EXT"