[SOLVED] Weird behaviour of a simple awk program

colucix · 10-15-2013, 01:55 PM

I have a question about awk. I made a simple one-liner to print out multiple lines in a file as a colon separated list of items. Input file:

Code:

123634.netserver
123635.netserver
123647.netserver

Code:

$ awk '{j ? j=j":"$0 : j=$0} END{print j}' file
123634.netserver:123635.netserver:123647.netserver

This works flawlessly, but if I change the variable name to _ it doesn't print the entire list but only the last item preceded by a colon:

Code:

$ awk '{_ ? _=_":"$0 : _=$0}END{print _}' file
:123647.netserver

To avoid this I have to put a space before the first double quotes:

Code:

$ awk '{_ ? _=_ ":"$0 : _=$0}END{print _}' file
123634.netserver:123635.netserver:123647.netserver

or add the --posix option:

Code:

$ awk --posix '{_ ? _=_":"$0 : _=$0}END{print _}' file
123634.netserver:123635.netserver:123647.netserver

I cannot explain the reason of this behaviour and why the underscore triggers it if not in posix compatibility mode. I tried this with awk 3.1.7 and awk 4.0.1 on different machines and the result is always the same. Why?

ntubski · 10-15-2013, 02:07 PM

13.3 Internationalizing awk Programs:

Quote:

_"your message here"

String constants marked with a leading underscore are candidates for translation at runtime. String constants without a leading underscore are not translated.

colucix · 10-15-2013, 02:09 PM

Well, I think I have the solution after having read a part of the GNU awk manual I've always ignored in the past. From chapter 13.4 (Translating awk Programs) it looks like the underscore is used as a marker for strings that has to be translated in other languages. Something that the strict POSIX standard doesn't allow. Hence the string

Code:

_":"

is a string marked for translation whereas with the blank space in the middle

Code:

_ ":"

is just the concatenation between the content of the variable _ and the string ":", exactly as it's supposed to be.

colucix · 10-15-2013, 02:10 PM

Thanks ntubski! There is always something new to learn about GNU awk!

grail · 10-16-2013, 07:06 AM

I realise the original question was about the unusual behaviour of _, but I just thought I would mention the alternative to provide the same output

Code:

awk '$1=$1' RS="" FS="\n" OFS=":" file

colucix · 10-16-2013, 07:20 AM

Nice one, grail! Thanks

danielbmartin · 10-16-2013, 08:39 AM

With this InFile ...

Code:

123634.netserver
123635.netserver
123647.netserver

... this paste ...

Code:

paste -sd: $InFile >$OutFile

... produced this OutFile ...

Code:

123634.netserver:123635.netserver:123647.netserver

Fewer keystrokes and (IMHO) better readability than other solutions.

Daniel B. Martin

colucix · 10-16-2013, 10:36 AM

Thanks Daniel! I have multiple choices at this point!