CODEWITHSUNDEEP
rregexstringlogfile
tho_mi
tho_mi

Reputation: 718

R - Split character vector using regex

I've got some kind of logfile I'd like to read and analyse. Unfortunately the files are saved in a pretty "ugly" way (with lots of special characters in between), so I'm not able to read in just the lines with each one being an entry. The only way to separate the different entries is using regular expressions, since the beginning of each entry follows a specified pattern.

My first approach was to identify the pattern in the character vector (I use read_file from the readr-package) and use the corresponding positions to split the vector with strsplit. Unfortunately the positions seem not always to match, since the result doesn't always correspond to the entries (I'd guess that there's a problem with the special characters).

A typical line of the file looks as follows:

16/10/2017, 21:51 - George: This is a typical entry here

The corresponding regular expressions looks as follows:

([[:digit:]]{2})/([[:digit:]]{2})/([[:digit:]]{4}), ([[:digit:]]{2}):([[:digit:]]{2}) - ([[:alpha:]]+):

The first thing I want is a data.frame with each line corresponding to a specific entry (in a next step I'd split the pattern into its different parts).

What I tried so far was the following:

regex.log = "([[:digit:]]{2})/([[:digit:]]{2})/([[:digit:]]{4}), ([[:digit:]]{2}):([[:digit:]]{2}) - ([[:alpha:]]+):"
log.regex = gregexpr(regex.log, file.log)[[1]]
log.splitted = substring(file.log, log.regex, log.regex[2:355]-1)

As can be seen this logfile has 355 entries. The first ones are separated correctly. How can I separate the character vector using a regular expression without loosing the information of the regular expression/pattern?

Upvotes: 0

Views: 830

Answers (1)

Jthorpe
Jthorpe

Reputation: 10204

Use capturing and non-capturing groups to identify the parts you want to keep, and be sure to use anchors:

file.log  =  "16/10/2017, 21:51 - George: This is a typical entry here"
regex.log = "^((?:[[:digit:]]{2})\\/(?:[[:digit:]]{2})\\/(?:[[:digit:]]{4}), (?:[[:digit:]]{2}):(?:[[:digit:]]{2}) - (?:[[:alpha:]]+)): (.*)$"

gsub(regex.log,"\\1",file.log)
>> "16/10/2017, 21:51 - George"
gsub(regex.log,"\\2",file.log)
>> "This is a typical entry here"

Upvotes: 2

Related Questions