On: Why Learn Awk? (2016)

[cauthon]

> I don't know of any easier equivalent of "awk '{ print $2 }'" for what it does.

Does `cut -f2` not work? My complaint with cut is that you can't reorder columns (e.g. `cut -f3,2` )

Awk is really great for general text munging, more than just column extraction, highly recommend picking up some of the basics

Edit to agree with the commenters below me: If the file isn't delimited with a single character, then cut alone won't cut it. You need to use awk or preprocess with sed in that case. Sorry, didn't realize that's what the parent comment might be getting at.

[syncsynchalt → cauthon]

It does not. Compare:

    $ echo '   1     2   3' | cut -f2
       1     2   3
    $ echo '   1     2   3' | cut -f2 -d' '
    
    $ echo '   1     2   3' | awk '{print $2}'
    2

"-f [...] Output fields are separated by a single occurrence of the field delimiter character."

[timbit42 → syncsynchalt]

echo ' 1 2 3' | tr -s ' ' | cut -b 2- | cut -d' ' -f2

or

echo ' 1 2 3' | tr -s ' ' '\t' | cut -b 2- | cut -f2

[cauthon → syncsynchalt]

oh yes, totally agree. if the data isn't delimited by a single character, then you definitely need awk or sed+cut

[ratsmack → syncsynchalt]

Also, the field separator (FS) can be a regular expression.

    FS = "[0-9]"

[jerf → ratsmack]

IIRC, there is an invocation of cut that basically does what I want, but every time I try, I read the manual page for 3 or 4 minutes, craft a dozen non-functional command lines, then type "awk '{ print $6 }'" and move on.

[masklinn → jerf]

> IIRC, there is an invocation of cut that basically does what I want

I don't think there is, because cut separates fields strictly on one instance of the delimiter. Which sometimes works out, but usually doesn't.

Most of the time, you have to process the input through sed or tr in order to make it suitable for cut.

The most frustrating and asinine part of cut is its behaviour when it has a single field: it keeps printing the input as-is instead of just going off and selecting nothing, or printing a warning, or anything which would bloody well hint let alone tell you something's wrong.

Just try it: `ls -l | cut -f 1` and `ls -l | cut -f 13,25-67` show exactly the same thing, which is `ls -l`.

cut is a personal hell of mine, every time I try to use it I waste my time and end up frustrated. And now I'm realising that really cut is the one utility which should get rewritten with a working UI. exa and fd and their friends are cool, but I'd guess none of them has wasted as much time as cut.

[catalogia → syncsynchalt]

Perfect example of how "do one thing and do it well" is a lie.

[masklinn → cauthon]

> Does `cut -f2` not work?

Most utilities don't use a tab character as separator, and that's what cut operates to by default. Can't cut on whitespace in general, which is what's actually useful, and what awk does.

Only way to get cut to work is to add a tr inbetween, which is a waste of time when awk just does the right thing out of the box.

[cauthon → masklinn]

> which is a waste of time when awk just does the right thing out of the box.

Agree in general. Only exception I'd make to this is when you're selecting a range of columns, as someone else mentioned elsewhere in the thread. I typically find (for example) `| sed -e 's/ \+/\t/g' | cut -f 1,3-10,14-17` to be both easier to type and easier to debug than typing out all the columns explicitly in an awk statement.

[b215826 → cauthon]

Instead of piping to sed, I would simply use

  | tr -s ' ' '\t'

[jerf → cauthon]

"Does `cut -f2` not work?"

As others have pointed out, no. It should! (Said the guy sitting comfortably in front of his supercomputer cluster in 2020. No, I don't do HPC or anything; everything's a supercomputer by the time that cut was written's standards.) But it doesn't. Going out on a limb, it's just too old. Cut comes from a world of fixed-length fields. Arguably it's not really a "unix" tool in that sense.

"highly recommend picking up some of the basics"

I have, that's the other 10%. I've done non-trivial things with it... well... non-trivial by "what I've typed on the shell" standards, not non-trivial by "this is a program" standards.

[jolmg → cauthon]

Not if the columns are separated by variable number of spaces. By default, the delimiter is 1 tab. You can change it to 1 space, but not more and not a variable number.

In my experience, most column based output uses variable number of spaces for alignment purposes. Tabs can work for alignment, but they break when you need more than 8 spaces for alignment.