How to write and read printable ASCII characters to/from UTF-8 encoding file?

Question

I want to write to a file with UTF-8 encoding containing the character 10001100 which is Œ the Latin capital ligature OE in extended ASCII table,

zz <- file("c:/testbin", "wb")
writeBin("10001100",zz)
close(zz)

When I open the file with office(encoding=utf-8), I can see Œ what I can not read is with readBin?

zz <- file("c:/testbin", "rb")
readBin(zz,raw())->x
x
[1] c5
readBin(zz,character())->x
Warning message:
In readBin(zz, character()) :
incomplete string at end of file has been discarded
x
character(0)

Potherb

I get 8c from charToRaw, can you put the output of sessionInfo() into your question? – Proustite

Answer 1

There are multiple difficulties here.

• Firstly, there are actually several "Extended ASCII" tables. Since you are on Windows you are probably using CP1252 which is one of them, also called Windows-1252 or ANSI, and the Win default "latin" encoding. However the code for Œ varies within this family of tables. In CP1252, "Œ" is represented by 10001100 or "\x8c", as you wrote. However it does not exist in ISO-8859-1. And in UTF-8 it corresponds to "\xc5\x92" or "\u0152", as rlegendi indicated.

So, to write UTF-8 from CP1252-as-binary-as-string, you have to convert your string into it a "raw" number (the R class for bytes) and then a character, change its "encoding" from CP1252 to UTF-8 (in fact convert its byte value to the corresponding one for the same character in UTF-8), after that you can re-convert it to raw, and finally write to the file:

char_bin_str <- '10001100'
char_u <- iconv(rawToChar(as.raw(strtoi(char_bin_str, base=2))),
              # "\x8c"    8c     140    '10001100'
                from="CP1252",
                to="UTF-8")

test.file <- "~/test-unicode-bytes.txt"

zz <- file(test.file, 'wb')
writeBin(charToRaw(char_u), zz)
close(zz)

• Secondly, when you readBin(), do not forget to give a number of bytes to read which is big enough (n=file.info(test.file)$size here), otherwise it reads only the first byte (see below):

  zz <- file(test.file, 'rb') x <- readBin(zz, 'raw', n=file.info(test.file)$size) close(zz)

  x [1] c5 92

• Thirdly, if in the end you want to turn it back into a character, correctly understood and displayed by R, you have first to convert it into a string with rawToChar(). Now, the way it will be displayed depends on your default encoding, see Sys.getlocale() to see what it is (probably something ending with 1252 on Windows). The best is probably to specify that your character should be read as UTF-8 – otherwise it will be understood with your default encoding.

  xx <- rawToChar(x) Encoding(xx) <- "UTF-8"

  xx [1] "Œ"

This should keep things under control, write the correct bytes in UTF-8, and be the same on every OS. Hope it helps.

---

PS: I am not exactly sure why in your code x returned c5, and I guess it would have returned c5 92 if you had set n=2 (or more) as a parameter to readBin(). On my machine (Mac OS X 10.7, R 3.0.2 and Win XP, R 2.15) it returns 31, the hex ASCII representation of '1' (the first char in '10001100', which makes sense), with your code. Maybe you opened your file in Office as CP1252 and saved it as UTF-8 there, before coming back to R?

Chopfallen

Answer 2

Try this instead (I replaced the binary value with the UTF encoding because I think it is better when you want such an output):

writeBin(charToRaw("\u0152"), zz)

Googins

why charToRaw("Œ") is c5 92 not 8c ? – Potherb

How did you got 8c there? :-) – Googins

in the ascii table ,you can see it. – Potherb

Hm, I don't know which table you use, can you give me a link? According to this document, this should be 92 not 8c as an UTF-8 encoded character: fileformat.info/info/unicode/char/0152/charset_support.htm The sequence c58c is letter O with macron fileformat.info/info/unicode/char/014c/index.htm – Googins