格式化字符串以在R / Shiny中搜索引擎样式

Question

I am working on a seemingly simple problem that nevertheless seems to be an annoying regex calculation.

I am designing a shiny app that allows users to search a database for strings and count the number of string matches.

From the stringr package, my ultimate call is:

str_count(text, pattern=REGEX(user_input))

My goal is to transform the user input into an appropriate regex - while allowing the user to input the data in standard search term format.

So the following user input:

artist picasso "picasso painting" france

should be formed into the following regex:

artist|picasso|picasso painting|france

Where the solution knows to treat "picasso painting" as one word due to the quotes.

Any help is appreciated!

Answer 1

Here is a base R solution:

regex.escape <- function(string) {
  gsub("([][{}()+*^${|\\\\?])", "\\\\\\1", string)
}

sort.by.length.desc <- function (v) v[order( -nchar(v)) ] 

s <- "artist picasso \"picasso (painting)\" france zoo"
keys <- c(t(read.table(text=s, header=FALSE)))          # Read in the values
keys <- sort.by.length.desc(keys)                       # Sort the values
pattern = paste(regex.escape(keys), collapse="|")       # Create the pattern
## Test
## cat(pattern, sep="\n")                               # This shows the regex pattern
txt <- "The artist was born in france and named picasso picasso (painting)"
length(unlist(gregexpr(pattern, txt)))                  # Count the number of occurrences
[1] 4

See the R demo . There are 4 matches, thus, the output is 4 .

Details :

• The regex.escape function escapes the most important chars that a regex engine may interpret as special characters
• The sort.by.length.desc orders the items of the character vector by character vector length in a descending order
• The c(t(read.table(text=s, header=FALSE))) reads the user input and stores as a character vector in keys
• The pattern = paste(regex.escape(keys), collapse="|") creates a pattern with alternation operators (looks like picasso \\(painting\\)|picasso|artist|france|zoo , cat(pattern, sep="\\n") displays the resulting pattern as a literal string)
• The length(unlist(gregexpr(pattern, txt))) line counts occurrences of a match using base R gregexpr function.

Answer 2

Split it up by doing global match using "[^"]*"|\\S+ .
Blindly remove leading/trailing double quotes ^"|"$ .
Push the matches into an array.
Sort the array longest on top (descending ?).
Replace each element's metachars ([\\[$^()*+|{}-\\\\]) with \\\\$1 .
Finally, join the elements together with an alternation | .