Join mehrzeiligen Strings in einem String

Question

Ich versuche, eine Protokolldatei zu analysieren, die wie folgt aussieht:

rawData <- structure(list(V1 = c("24 01 2016, 11:50:17 pm: ‎Line to skip", 
     "24 01 2016, 11:50:17 pm: ‎Line to skip", "24 01 2016, 11:51:47 pm: User1: Message one is here", 
     "24 01 2016, 11:53:04 pm: User2: A long message that spans multiple lines, so I have to write a really long and tedious message here to illustrate my point. The point is that this message is really long and ", 
     "can", "[span]", "Several lines.", "24 01 2016, 11:51:47 pm: User3: My first message", 
     "27 10 2017, 12:54:03 am: ‎‪+44 ‬012 3456789 left")), .Names = "V1", row.names = c(NA, 
     -9L), class = "data.frame") 

Jede Nachricht beginnt mit einem Datum

24/01/2016, 11:50:17 pm: ‎Line to skip 
24/01/2016, 11:50:17 pm: ‎Line to skip 
25/01/2016, 11:51:47 pm: User1: Message one is here 
25/01/2016, 11:53:04 pm: User2: A long message that spans multiple lines, so I have to write a really long and tedious message here to illustrate my point. The point is that this message is really long and 

can 
[span] 

Several lines. 
24/01/2016, 11:51:47 pm: User3: My first message 
27/10/2017, 12:54:03 am: ‎‪+44 ‬012 3456789 left 
28/10/2017, 02:54:03 pm: User3: My second message! 

---

, und ich habe Problem beim Parsen von Nachrichten, die sich über mehrere Zeilen erstrecken (wie in Zeile 4).

Hier ist, was ich bisher:

> head(parseR()) 
     date  time sender                                          message 
1 2016-01-25 23H 51M 47S User1                                       Message one is here 
2 2016-01-25 23H 53M 4S User2 A long message that spans multiple lines, so I have to write a really long and tedious message here to illustrate my point. The point is that this message is really long and 
3 2016-01-24 23H 51M 47S User3                                        My first message 
4 2017-10-28 14H 54M 3S User3 

vorschlagen Kann jemand:

suppressMessages(library(lubridate)) 
suppressMessages(library(dplyr)) 
suppressMessages(library(plyr)) 
suppressMessages(library(tidyr)) 

parseR <- function(file='data/chat_log.txt',drop="44"){ 
    rawData <- read.delim(file, quote = "", 
        row.names = NULL, 
        stringsAsFactors = FALSE, 
        header = F) 


    # remove blank lines 
    # rawData<-rawData[!apply(rawData == "", 1, all),] 

    empty_lines = grepl('^\\s*$', rawData) 
    rawData = rawData[! empty_lines] 

    # join multi line messages into single line 
    # rawData$V1<-gsub("[\r\n]", " ", rawData$V2) 

    sepData<-suppressWarnings(separate(rawData, V1, c("datetime", "sender", "message"), sep = ": ", extra = "merge")) 

    sepData$message <- trimws(sepData$message) 
    sepData$sender<-factor(sepData$sender) 

    data <- sepData %>% 
    filter(!is.na(message)) %>% 
    filter(!grepl(drop, sender)) %>% 
    droplevels() 

    cleanData<-separate(data, datetime, c("date", "time"), sep = " ", remove =TRUE) 
    cleanData$date<-ymd(cleanData$date) 
    cleanData$time<-hms(cleanData$time) 

    return(cleanData) 
} 

---

Allerdings, wenn ich den zurückgegebenen Datenrahmen überprüfen, werden muliti Linie Nachrichten nicht korrekt analysiert werden eine Möglichkeit, leere Zeilen zu entfernen und Text, der nicht mit einem Datum beginnt, in eine Zeichenfolge zu integrieren?

Wunschformat für Zeile 4:

25/01/2016, 11:53:04 pm: User2: A long message that spans multiple lines, so I have to write a really long and tedious message here to illustrate my point. The point is that this message is really long and can [span] Several lines. 

Answer 1

Hier ist ein Caveman Ansatz für dieses Problem. Ich nehme den Zeitstempel, um ein eindeutiger Anfang der Leitungsidentifizierung zu sein. Wenn keine vorhanden ist, wird die Linie (oder das Element) in die vorherige eingefügt. Das folgende Beispiel ist an einen Vektor angepasst, kann aber leicht geändert werden, um mit anderen Klassen wie Matrizen oder Datenrahmen zu arbeiten.

rd <- c("24 01 2016, 11:50:17 pm: Line to skip", 
     "24 01 2016, 11:50:17 pm: Line to skip", "24 01 2016, 11:51:47 pm: User1: Message one is here", 
     "24 01 2016, 11:53:04 pm: User2: A long message that spans multiple lines, so I have to write a really long and tedious message here to illustrate my point. The point is that this message is really long and ", 
     "can", "[span]", "Several lines.", "24 01 2016, 11:51:47 pm: User3: My first message", 
     "27 10 2017, 12:54:03 am: ‪+44 ‬012 3456789 left") 
rd 

out <- rep(NA, length(rd)) 

gr <- 1 
for (i in 1:length(rd)) { 
    # if starting with timestamp, save into out and move on (gr) 
    find.startline <- grepl("^\\d{2} \\d{2} \\d{4}, \\d{2}:\\d{2}:\\d{2} (am|pm):", rd[i]) 
    if (find.startline) { 
    out[gr] <- rd[i] 
    gr <- gr + 1 
    } 

    if (!find.startline) { 
    # if doesn't start with timestamp, append to previous (ss) 
    ss <- gr - 1 
    out[ss] <- paste(out[ss], rd[i]) 
    } 
} 

# if there are any multiline comments, some residual NAs should be present, removed 
out <- out[!is.na(out)] 
out 

[1] "24 01 2016, 11:50:17 pm: Line to skip"                                                 
[2] "24 01 2016, 11:50:17 pm: Line to skip"                                                 
[3] "24 01 2016, 11:51:47 pm: User1: Message one is here"                                              
[4] "24 01 2016, 11:53:04 pm: User2: A long message that spans multiple lines, so I have to write a really long and tedious message here to illustrate my point. The point is that this message is really long and can [span] Several lines." 
[5] "24 01 2016, 11:51:47 pm: User3: My first message"                                               
[6] "27 10 2017, 12:54:03 am: *+44 ,012 3456789 left" 

Answer 2

Ich schlage vor, etwas Ähnliches wie Roman Lösung aber in Tidyverse Welt:

rawData %>% 
    mutate( 
    MgsNo = (!substr(V1, 1, 1) %>% # take first character 
       as.numeric %>% # convert to numeric - produces NAs for non-numeric values 
       is.na) %>% # produces True(1) and False(0) (by ! I reverse those) 
     cumsum) %>% # then cumulative sum as Mgs NO e.g. 1,1,1,0,0,1 -> 1,2,3,3,3,4 
    group_by(MgsNo) %>% 
    do(MgsBody = paste(.$V1 , collapse = "")) %>% # concatenate all in each MgsNo group 
    select(MgsBody) %>% 
    pull