Converting text into ASCII/ISO-8859-1

If you have these escape sequences in a shell variable, in dash, mksh or bash:

printf %b "$string_with_backslash_escapes"

This isn't POSIX: the %b specifier is POSIX but it requires a 0 after each backslash. This also interprets other backslash escapes: \n as a newline, \t as a tab, etc.

Here's a perl one-liner that converts octal escape sequences only.

perl -pe 's[\\(?:([0-7]{1,3})|(.))] [defined($1) ? chr(oct($1)) : $2]eg'

http://www.rapidmonkey.com/unicodeconverter/reverse.jsp interprets octal values as Latin-1 characters, I don't know why Unicode and UTF-8 are mentioned in the page. I have no idea what http://www.rapidmonkey.com/unicodeconverter/advanced.jsp does.

Apparently LibreOffice tries to use ISO-8859-1 by default, which is causing the problem. In response to this bug report, a new parameter --infilter has been added. The following command produces U+2014 em dash:

libreoffice  --convert-to csv --infilter=CSV:44,34,76,1 --headless --outdir dir file.xlsx

I tested this with LO 5.0.3.2. From the bug report, it looks like the earliest version containing this option is LO 4.4.

See also: https://ask.libreoffice.org/en/question/13008/how-do-i-specify-an-input-character-coding-for-a-convert-to-command-line-usage/