What the equivalent of "grep | cut" using sed or awk?

Question

Say I had a config file /etc/emails.conf

email1 = [email protected] 
email2 = [email protected]
email3 = [email protected]

and I wanted to get email2

I could do a:

grep email2 /etc/emails.conf | cut -d'=' -f2 

to get the email2, but how do I do it "cooler" with one sed or awk command and remove the whitespace that the cut command would leave?

Solution 1

How about using awk?

awk -F = '/email2/ { print $2}' /etc/emails.conf

• -F = Fields are separated by '='

• '/email2/ { print $2}' On lines that match "email2", print the second field

Solution 2

The exact equivalent would be something like:

sed -n '/email2/{s/^[^=]*=\([^=]*\).*/\1/;p;}' < file

But you'd probably want instead:

sed -n 's/^[^=]*email2[^=]*=[[:blank:]]*//p' < file

(that is match email2 only on the part before the first = and return everything on the right of the first = (skipping leading blanks), not only the part up to the second = if any).

Solution 3

perl -nlE 's/email2\s*=\s*// and say'    file

Where:

• perl -nl is a for each line do...
• s/email2 = // removes the searched email id and if you could do it ...
• say prints the current input input line
• \s* zero or more spaces (equivalent to [ \t\n]*)

Comments

[cuonglm]

Is space required between { s?

[Stéphane Chazelas]

@cuonglm, no. AFAICT, ;} above is not POSIX but I don't know of any modern sed implementation where that fails.

[cuonglm]

Yes, that's exactly I have stucked when working with this. After re-reading POSIX sed documentation, I see ;} is an accepted extension, but it's not required. It's strange that GNU sed with --posix option still allow {command} form. Do you think it's a bug?

[Peter Turner]

can it trim the leading space too?

[Stéphane Chazelas]

@cuonglm {command} is not required to report an error by POSIX. It's just that a POSIX script must not use it as the behaviour is not specified there. Where GNU sed is not conformant even with --posix is that it doesn't accept ; in label names.

[Peter Turner]

is it preferable to us "<" vs just using the file as the argument (I'm putting this into a bash script)?

[Peter Turner]

I'll give you a +1 for golfing, but where I'm running this (some sort of stripped down Cisco appliance), I'm not sure I'm going to have perl, especially the version of perl that has say.

[JJoao]

perl 5.10 appeared 7 year ago but sometimes we don't have it... perl -nle 's/email2\s*=\s*// and print'

[Stéphane Chazelas]

@PeterTurner, see When should I use input redirection?.

[Wildcard]

I can't find where ;} is mentioned in the Sed POSIX specs as an accepted extension. It seems to be required by The <right-brace> shall be preceded by a <newline> or <semicolon>. @cuonglm or Stephane, could you clarify?

[cuonglm]

@Wildcard Yes, ;} is not POSIX, that what Stephane said in his comment above. But I remember it will be in next POSIX version, there's an issue in austin bug group.

[Wildcard]

@cuonglm, no, I mean I don't see it as just an extension. "The right brace shall be preceded by a newline or semicolon...." So if you precede it with a semicolon in your Sed script, haven't you followed the explicit instructions of POSIX?

[Wildcard]

@cuonglm, huh, I still don't see it in the specs. But if it will be changed in the next POSIX edition I suppose it's not critical. :)

[cuonglm]

@Wildcard: The austinbug group link austingroupbugs.net/view.php?id=961

[cuonglm]

@Wildcard Ops, look like the spec was changed follow that issue.

[Wildcard]

Let us continue this discussion in chat.

[cbarrick]

Unfortunately, this answer does little to explain why the command works. I've posted an alternative sed answer with that explanation.