Question

I have a file, file1.tex, containing TeX commands, like \em and \par. All of the commands are in the format \ + some string of uppercase and lowercase letters from A-Z and are all followed by a space.

I need to use a command like this, which replaces all spaces with \, a slash and a space.

sed -i "s/\ /\\\\\ /g" ./file1.tex

I do not want these to replace the empty spaces which appear immediately after TeX commands though. For example, I want this:

\noindent This is a sentence {\em which has some words}.
This is another \hfill sentence \ldots with some more words.

To become:

\noindent This\ is\ a\ sentence\ {\em which\ has\ some\ words}.
This\ is\ another\ \hfill sentence\ \ldots with\ some\ more\ words.

How can I replace all spaces, except those appearing after any command taking the form of \sometext?

Answer 1

Since sed doesn’t support look-behind, I think this will be a lot easier using Perl.

$ perl -pe 's/\b(?<!\\)(\w+)\b /$1\\ /g' texfile
\noindent This\ is\ a\ sentence\ {\em which\ has\ some\ words}.
This\ is\ another\ \hfill sentence\ \ldots with\ some\ more\ words.

To make the changes permanent to the file in-place:

perl -pi -e 's/\b(?<!\\)(\w+)\b /$1\\ /g' texfile

Explanation:

The regex matches a word which does not start with a backslash which is followed by a space.

• \b – word boundary
• (?<! – begin a non-capturing negative look-behind (don’t match)
• \\ – escaped backslash
• ) – close the look-behind
• ( – begin a capture group
• \w+ – match one or more word characters (alphanumeric plus underscore)
• ) – close the capture group
• $1 – copy the capture group into the replacement
• \\ – add a backslash
• g – do the substitution globally

I left a couple things out of the list which should be self-evident.