assigning bash variable with the system command in awk

[mrz]

Hi all;

I am having all sorts of trouble trying to assign a variable within an awk script with the system command. I know there is a lot of ways around this problem, but for efficiency reasons, I would like to, within my awk script, do something like

system(x=3)

or

system(x=NR)


and, latter on the shell script which calls the awk script, use the variable $x. But nothing is passed to x.

I have already tried things like

command = "x=3"
system(command)

and also used a pipeline within the system to pipe it to /bin/sh

In fact tried a lot of stuff like that, using $(( )) etc etc etc

Funny thing is that I can create directories e write to files (yes, i could write to a file and read from there, but I dont think it is efficient, plus I am puzzled).

I guess it has something to do with a shell sub-process.

Thanks a lot in advance.

[Kloschüssel]

just guessing, but system(export X=3) may saves the day?

[mrz]

Good guess, but it also do not work. I guess the point here is export a variable from a shell sub-process to its parent process (maybe if this was the title of the thread a lot of people would be giving their shots).

From the parent to the child you do it with export, the other way around seems to be tricky.

[Kloschüssel]

thinking about it: the other way around should be impossible as that would be dangerous and would create big problems as it would be the easiest way to manipulate a system.

btw: shouldn't the awk script know what to do and output proper data that (if wanted) could be parsed by the invoker?

[mrz]

I see your point, but I guess it is not impossible. But surely I would have to make my parent process read from the child (and this is one thing I was missing at first).

In general it seems that a coprocess would be the way around this, but you cannot use it with awk. Anyway I found another way to solve my problem, but after your guess on the 10:27 post I learnt a lot of things. Thanks!

[Kloschüssel]

glad to have helped .. don't know how but glad anyway.

solve the thread once you believe it is appropriate to do so.

[mrz]

It was after your post that I realised the obvious (that awk aws run in a child process), that's how you've helped (and thanks also for the suggestions themselves).

As for the "solved", although I found another way around my particular problem, the original question is still open, and I think it is worth answering.

[Kloschüssel]

I do not believe there exists a solution for above usecase, but maybe we've got a brainy guy out there who knows.