sed whole word search and replace but include other characters

Question

To replace whole words with sed, one does:

$ echo "bar embarassment" | sed "s/\bbar\b/no bar/g"
no bar embarassment

This is taken from another stackoverflow question. Follow on question, how do I change the definition of a word?

From linuxtopia:

GNU sed, ssed, sed16, sed15 and sedmod use certain symbols to define the boundary between a "word character" and a nonword character. A word character fits the regex "[A-Za-z0-9_]".

How does one include e.g. "-"? Or in my particular case, I want to rename variables in a R codebase, where they are littered with "." (it is often used instead of "_" in variable names, see for example google's R styleguide), so I would like to include "." in the definition of a word.

EDIT:

To be extra clear, say I want to change current.my.date <- my.date + today into current.my.date <- any.date + date, what is the sed command?

e.g. fix this command

echo "current.my.date <- my.date + today" | sed "s/\bmy.date\b/any.date/g"
current.any.date <- any.date + today

Because in its current form it also changes current.my.date

Answer 1

Try this:

$ echo "current.my.date <- my.date + today" |
    sed -r 's/(^|[^[:alnum:]_.])my\.date([^[:alnum:]_.]|$)/\1any.date\2/g'
current.my.date <- any.date + today

It assumes that a "word" is a sequence of "alpha-numeric or _ or ." characters separated by characters outside of that set or preceded by a start-of-string (^) or succeeded by an end-of-string ($).

If that's not what you want, post more sample input and expected output.

Sound like you need some variation of this:

awk '{
    head = ""
    tail = $0
    while( match( tail, /(^|[^[:alnum:]_.])my\.date([^[:alnum:]_.]|$)/ ) ) {
        head = head substr(tail,1,RSTART-1) "any.date"
        tail = substr(tail,RSTART+RLENGTH-1)
    }
    print head tail
}' file

to get what you want.

Answer 2

so I would like to include "." in the definition of a word

You can use this character class:

[A-Za-z0-9_.]

If you want to add hyphen also then use:

[A-Za-z0-9_.-]

Also remember with these additions you cannot reply on \b as word boundary since hyphen and dot are also considered word boundary. You can use negated character class for that case:

[^A-Za-z0-9_.-]

EDIT:

echo "foo-bar embarassment" | sed "s/\([A-Za-z0-9_.-]\+\)/no \1/g"
no foo-bar no embarassment