问题
Is it possible to manipulate the record/observation/row delimiter when reading in data (i.e. read.table) from a text file? It's straightforward to adjust the field delimiter using sep="", but I haven't found a way to change the record delimiter from an end-of-line character.
I am trying to read in pipe delimited text files in which many of the entries are long strings that include carriage returns. R treats these CRs as end-of-line, which begins a new row incorrectly and screws up the number of records and field order.
I would like to use a different delimiter instead of a CR. As it turns out, each row begins with the same string, so if I could use use something like \nString to identify true end-of-line, the table would import correctly. Here's a simplified example of what one of the text files might look like.
V1,V2,V3,V4
String,A,5,some text
String,B,2,more text and
more text
String,B,7,some different text
String,A,,
Should read into R as
V1 V2 V3 V4
String A 5 some text
String B 2 more text and more text
String B 7 some different text
String A N/A N/A
I can open the files in a text editor and clean them with a find/replace before reading in, but a systematic solution within R would be great. Thanks for your help.
回答1:
We can read them in and collapse them afterwards. g will have the value 0 for the header, 1 for the next line (and for follow on lines, if any, that are to go with it) and so on. tapply collapses the lines according to g giving L2 and finally we re-read the lines:
Lines <- "V1,V2,V3,V4
String,A,5,some text
String,B,2,more text and
more text
String,B,7,some different text
String,A,,"
L <- readLines(textConnection(Lines))
g <- cumsum(grepl("^String", L))
L2 <- tapply(L, g, paste, collapse = " ")
DF <- read.csv(text = L2, as.is = TRUE)
DF$V4[ DF$V4 == "" ] <- NA
This gives:
> DF
V1 V2 V3 V4
1 String A 5 some text
2 String B 2 more text and more text
3 String B 7 some different text
4 String A NA <NA>
回答2:
If you're on Linux/Mac, you should really be using a command line tool, like e.g. sed, instead. Here are two slightly different approaches:
# keep the \n
read.csv(pipe('sed \'N; s/\\([^,]*\\)\\n\\([^,]*$\\)/"\\1\\n\\2"/\' test.txt'))
# V1 V2 V3 V4
#1 String A 5 some text
#2 String B 2 more text and\nmore text
#3 String B 7 some different text
#4 String A NA
# get rid of the \n and replace with a space
read.csv(pipe('sed \'N; s/\\([^,]*\\)\\n\\([^,]*$\\)/\\1 \\2/\' test.txt'))
# V1 V2 V3 V4
#1 String A 5 some text
#2 String B 2 more text and more text
#3 String B 7 some different text
#4 String A NA
来源：https://stackoverflow.com/questions/16115887/possible-to-change-the-record-delimiter-in-r