根据特定条件从R中的字符串中提取数字

Question

I'm trying to extract some numbers from a string (comments) based on a specific criteria. The numbers I would like to extract directly follow a date in 24 hour format and always contain a decimal place and are less than 20 (there are other numbers in the string but I'm not interested in these). I've managed to extract the numbers I would like with the R code below but have no way of relating these back to the ID's they came from. Some ID's have multiple numbers of interest, while some only have one. For example, I need some way to associate the ID number in the dummy data given below with every number of interest. As you can see ID 1 contains three results of interest (4.1, 6.9 and 4.3) while ID 2 has only 1 result of interest (6.5).

Any help would be fantastic!

(An example of the format of comment.txt)

    ID  comments
    1   abc1200 4.1  abc1100 6.9 etd1130 4.3 69.0
    2   abc0900 6.5 abcde 15
    3   3.2 0850 9.5 abc 8.2 0930 12.2 agft 75.0
    4   ashdfalsk 0950 10.5 dvvxcvszv asdasd assdas d 75.0


#rm(list=ls(all=TRUE))

#import text and pull out a list of all numbers contained withtin the free text
raw_text <- read.delim("comment.txt")
numbers_from_text <- gregexpr("[0-9]+.[0-9]", raw_text$comments)

numbers_list <- unlist(regmatches(raw_text$comments, numbers_from_text))
numbers_list <- as.data.frame(numbers_list)

#pull out those numbers that contain an decimal place and create a running count
format<-cbind(numbers_list,dem=(grepl("\\.",as.character(numbers_list$numbers_list)))*1,row.number=1:nrow(numbers_list))

#if the number does not contain a decimal (a date) then create a new row number which is the addition of the first row
#else return NA
test <- cbind(format,new_row = ifelse(format$dem==0, format$row.number+1, "NA"))

#match the cases where the new_row is equal to the row.number and then output the corresponding numbers_list
match <-test$numbers_list[match(test$new_row,test$row.number)]

#get rid of the NA's for where there wasnt a match and values less than 20 to ensure results are correct
match_NA <- subset(match, match!= "<NA>" & as.numeric(as.character(match))<20)

match_NA <- as.data.frame(match_NA) 

Answer 1

Something like this seems to work, matching numerics starting with a blank which contain a period, then converting to numeric and extracting which ones are less than 20.

library(stringr)
temp <- apply(comments, 1, function(x) {
  str_extract_all(x,"[[:blank:]][0-9]+[.][0-9]")
})

library(purrr)
temp <- lapply(flatten(temp), function(x) as.numeric(str_trim(x)))
lapply(temp, function(x) x[x <20])

[[1]]
[1] 4.1 6.9 4.3

[[2]]
[1] 6.5

[[3]]
[1]  3.2  9.5  8.2 12.2

[[4]]
[1] 10.5