Uppercasing First Letter of Words Using SED

Use the following sed command for capitalizing the first letter of the each word.

echo -e "Trouble me \nGold rush brides" | sed -r 's/\<./\U&/g'

output

Trouble Me
Gold Rush Brides

The -r switch tells sed to use extended regular expressions. The instructions to sed then tell it to "search and replace" (the s at the beginning) the pattern \<. with the pattern \U& globally, i.e. all instances in every line (that's the g modifier at the end). The pattern we're searching for is \<. which is looking for a word boundary (\<) followed by any character (.). The replacement pattern is \U&, where \U instructs sed to make the following text uppercase and & is a synonym for \0, which refers to "everything that was matched". In this case, "everything that was matched" is just what the . matched, as word boundaries are not included in the matches (instead, they are anchors). What . matched is just one character, so this is what is upper cased.

I know you said sed, but for shell scripting one of the easiest and more flexible for me was using Python which is available in most systems:

$ echo "HELLO WORLD" | python3 -c "import sys; print(sys.stdin.read().title())"
Hello World

For example:

$ lorem | python3 -c "import sys; print(sys.stdin.read().title())"
Officia Est Omnis Quia. Nihil Et Voluptatem Dolor Blanditiis Sit Harum. Dolore Minima Suscipit Quaerat. Soluta Autem Explicabo Saepe. Recusandae Molestias Et Et Est Impedit Consequuntur. Voluptatum Architecto Enim Nostrum Ut Corrupti Nobis.

You can also use things like strip() to remove spaces, or capitalize():

$ echo "  This iS mY USER ${USER}   " | python3 -c "import sys; print(sys.stdin.read().strip().lower().capitalize())"
This is my user jenkins

Using sed with tr:

name="athens"
echo $name | sed -r "s/^[[:alpha:]]/$(echo ${name:0:1} | tr [[:lower:]] [[:upper:]])/"

Using bash (without sed, so a little off topic):

msg="this is a message"
for word in $msg
do
   echo -n ${word^} ""
done

This Is A Message

Capitalize the first letter of each word (first lowercase everything, then uppercase the first letter).

echo "tugann sí spreagadh" | sed -E 's/.*/\L&/g' | sed -E 's/(\b.)/\U\1/g'
> Tugann Sí Spreagadh

Proposed sed solutions until now will only work if the original text is in lowercase. Although one could use tr '[[:upper:]]' '[[:lower:]]' to normalize the input to lowercase, it may be convenient to have an all-in-one sed solution :

sed 's/\w\+/\L\u&/g'

This will match words (\w means word character, and \+ at least one and until the next non-word character), and lowercase until the end (\L) but uppercase the next (i.e. first) character (\u) on each (g) matched expression (&).

[Credits]

Late post, but using ex (or vi) is POSIX (like the awk suggestion above) so it should work on any system that ships a proper "vi" implementation. FWIW, Busybox does not make the grade here.

POSIX sed only supports BREs and I personally do not advocate relying on GNU extensions if you have a choice.

However, POSIX ex/vi supports a few ERE features, including the word boundary < and the \U switch. If your platform does not have a compliant ex/vi then you have bigger problems than EREs, IMO.

As a bonus, a wq! will optionally perform an "in-place" edit.

cat <<EOF > casing.txt
Trouble me
Gold rush brides
EOF

ex -s casing.txt <<'EOF'
%s/\<./\U&/g
%p
wq!
EOF