How do i get "sed" to delete everything else but email address. - sed

Question

how do i get "sed" to delete everything else but email address.
db dump: someusername ::: kRW...0fPc ::: $2a$10$...aucvkDt86 ::: joesmith#gmail.com

Answer 1

Does it have to be sed? What about grep? Here's how to use it with the regex you gave:
$ cat dbdump.txt
db dump: someusername ::: kRW...0fPc ::: $2a$10$...aucvkDt86 ::: joesmith#gmail.com
another line with two e-mail addresses <test#example.com> on it -- bob#example.org
$ grep -EiEio '\b[A-Z0-9._%+-]+#[A-Z0-9.-]+\.[A-Z]{2,4}\b' dbdump.txt
joesmith#gmail.com
test#example.com
bob#example.org
The -o flag prints only the matching portions, i.e. just the e-mail addresses. -i makes the matching case insensitive. It even finds multiple e-mail addresses on the same line.
Edit: I couldn't resist the -EiEio. I suppose grep -Eio or egrep -io would also work...

Answer 2

This requires GNU sed:
sed -r 's/\b[A-Z0-9._%+-]+#[A-Z0-9.-]+\.[A-Z]{2,4}/\n&\n/ig;s/(^|\n)[^#]*(\n|$)/\n/g;s/^\n|\n$//g;/^$/d' inputfile
split input lines so email addresses and other strings are separated by newlines
erase sequences that consist of only non-# characters delimited by newlines or the beginning or end of the input line
erase extra newlines and blank lines

Answer 3

The following will work no matter where the email address is in the line but only if there is one email address per line. If there are more than one it's going to only show the last one in the line. It also won't touch lines that don't have valid email addresses in them
sed 's/^.* \([^# ]\+#[^ ]\+\) \?.*$/\1/'
Input
$ cat dbdump
this line with no valid # email address is untouched
::: a0$...aucvkDt86 ::: joesmith#gmail.com
::: a0$... foo#example.com db dump: someusername :::
Output
$ sed 's/^.* \([^# ]\+#[^ ]\+\) \?.*$/\1/' ./dbdump
this line with no valid # email address is untouched
joesmith#gmail.com
foo#example.com

Answer 4

With sed:
$ echo "db dump: someusername ::: kRW...0fPc ::: $2a$10$...aucvkDt86 ::: joesmith#gmail.com"|sed 's/.*::: //' joesmith#gmail.com
With awk:
$ echo "db dump: someusername ::: kRW...0fPc ::: $2a$10$...aucvkDt86 ::: joesmith#gmail.com"|awk '{print $NF}'
EDIT: Given the new info in your comment - it's quite hard to do what you ask without any regularity. Check the Syntax section here:
http://en.wikipedia.org/wiki/Email_address
Standard says that e.g. 1$%3{C}#example.com is a valid email address (believe it or not). You can even quote it (the example given in the article is John Doe#example.com). So, by following the standard it's almost impossible to recognize the valid e-mail.
If you restrict your search, you can e.g. extract the lines containing # by first doing:
cat your-file.txt|grep #
then do some of the above. You can even do something like this:
$ echo "garbage John.Doe123#example.com garbage"|sed 's/[^#]* \([a-zA-Z0-9.]*#[^ ]*\).*/\1/'
John.Doe123#example.com
Note that the above works under the following assumptions:
There's a space before an email address
There are no spaces in an email address itself
There is one email address in the line (it will actually get only the first one, so it can work with more then one)
The local-part (all before #) contains only letters (lower- or upper-case), digits and a dot
Expand the character set ([a-zA-Z0-9.]) as you wish to make it less restrictive - e.g. you can do [a-zA-Z0-9.-_] to include - and _.