Combination Regmatches and Replacement for specific Character-CodePudding

I've tried replace character which match with specific character or followed by "BT", but my codes failed. This is my codes:

df <- data.frame(
  exposure = c("123BT", "113BB", "116BB", "117BT")
)

df %>%
  mutate(
    exposure2 = case_when(exposure == regmatches("d \\BT") ~ paste0("-", exposure),
                     TRUE ~ exposure)
  )

the error is:

Error: Problem with `mutate()` column `exposure2`.
i `exposure2 = case_when(...)`.
x argument "m" is missing, with no default
Run `rlang::last_error()` to see where the error occurred.

Whereas my target is:

df <- data.frame(
  exposure = c("123BT", "113BB", "116BB", "117BT"),
exposure2 = c(-123, 113, 116, -117)
)

CodePudding user response：

I recommend you use library stringr, you can extract your numbers with regex (\\d) :

library(stringr)
library(dplyr)

df %>%
  mutate(
    exposure2 = case_when(str_detect(exposure,"BT") ~ paste0("-", str_extract(exposure, "(\\d) ")),
                          TRUE ~ str_extract(exposure, "(\\d) "))
  )

Output:

  exposure exposure2
1    123BT      -123
2    113BB       113
3    116BB       116
4    117BT      -117

If you still prefer use regmatches you can get same result with:

df %>%
  mutate(
    exposure2 = case_when(exposure %in% regmatches(exposure, regexpr("\\d BT", exposure)) ~ paste0("-", regmatches(exposure, regexpr("\\d ", exposure))),
                          TRUE ~ regmatches(exposure, regexpr("\\d ", exposure)))
  )

CodePudding user response：

First, a concise solution that you can easily implement in your dplyr::mutate. Using gsub we remove characters and coerce the result as.integer. The result, we multiply by 1 or -1 according to if the string contains "BT" or not; for this we use grepl (gives boolean) and add 1L (coerces to integer) to get indices 1 or 2.

c(1, -1)[grepl('BT', df$exposure)   1L]*as.integer(gsub('\\D', '', df$exposure))
# [1] -123  113  116 -117

Above is the recommended solution. The solution you envision is much more complex since it processes the information not very efficient. I implement the logic in a small function¹ to demonstrate.

f <- \(x) {
  rm <- regmatches(x, regexpr("\\d BT", x))
  o <- gsub('\\D', '', x)
  o <- ifelse(x %in% rm, paste0('-', o), o)
  as.integer(o)
}

f(df$exposure)
# [1] -123  113  116 -117

_{¹Notes: For regmatches you need matching info, e.g. from regexpr. The regex should actually look sth like "\\d BT".}

Data:

df <- structure(list(exposure = c("123BT", "113BB", "116BB", "117BT"
)), class = "data.frame", row.names = c(NA, -4L))

CodePudding user response：

library(readr) 
(-1)^grepl('BT', df$exposure)  * parse_number(df$exposure)
[1] -123  113  116 -117