error comparing strings
 

I am building a script to manually update my dovecot packages directly from dovecot repo , however this script does not do that job straight ahead .
First checks apt policy to see witch packages of dovecot are installed in system , and then write those packages names to a file .

I am posting here this part only of the code , but when i look into the logfile from apt and compare strings with IF statement i get "invalid arithmetic operator on my loop."

Eventually the problem is in the if statement , i changed the -ne with != and i did not get accurate output

If you mean this line:My understanding is that you're not supposed to quote the variables, just strings themselves.

And it would be !=

The != is the correct operator to compare strings. You might use "set -x" or "set -xv" earlier in the script and watch what it tells you about the process.

Some random comments. Making a directory and subdirectory can be done in a single move:

grep, sed, and awk need never be combined:

can be changed to
leaving the $basedir there and everything within it deleted. remove the * if you want to delete that dir too.

this here
just checks the file names against each other, so they will never match if not named the same, you got a look inside of them.
example
MOD:
as @Turbocapitalist posted ahead of me, that awk he is doing is prob best to get what info out of the file that is needed to be checked against.

Why is the cat there? Wont that read the whole file then truncate, instead of just using head to read the first line?

thanks, I never really used head to know how to input to it so I just defaulted to cat first then head it.

thanks for correcting me.

Also, regarding the mkdir/rm stuff... since -f and -p suppress errors regarding (non-)existence I think the whole if/else might not be needed.

Depending on the specifics of what should/shouldn't be deleted, perhaps one of these is suitable:

Just a comment:is much more efficient I think, but not really important.

keeping in mind you need to add the -r (for recursive) if going past parent dir.

Yep, my edit came just too slow...


It's useful to know, but my recommendation is readability first, unless/until efficiency becomes a measurable issue.

Thanks for all the replies .

thanks for some tips .

I am not comparing files , what i am doing is to see in "Installed" field , pick whats next "var2" , and if output is different than "(none)" then pick up the name of that installed package only to a file .

example :

when loop gets to dovecot-pigeonhole-dbg:
it reads the next line "Installed" , then grabs "2:2.3.9.2-1+debian10" and compare with "(none)" .
In case it is different then writes "dovecot-pigeonhole-dbg:" to a file .
I am pretty sure there is a simple way to do this , and i have faith that someone here already figured out to do it in a single line ;) .

Ah - this is why it's a good idea to explain what you're trying to achieve before explaining what you're stuck on, because you're right - this can be done in one line:

packages.txt contains the quoted text in your post.

Uses -P for Perl regex to get lookahead support, with -z to match across lines, and -o to only output the matched text.

The "(?<=\n|\A)" is start of line/content - it would ideally just be "^" but that's not supported with the -z flag.

The "dovecot.*" searches for all dovecot prefixed packages - if you need specific ones, replace it with "(package1|package2)"

The next bit looks for "\nInstalled: " that isn't followed by "(none)\n" - using a lookahead to not include it in the output.

This version does include the colon in the output, but if you don't actually want that, move the : inside the lookahead (i.e. just before the \n )

Assumes the "Installed:" line always directly follows the package name, which seems a safe assumption? If not, (and depending on exact format possibilities), I might filter the content first to remove other lines.

Here it is the problem , i want the script to check witch ones are installed without having to specify any of them .

In this file (aptinst), how to see witch packages are installed automatically without adding anything specifically to the script ?


For this specific file i need this outputs :
But these outputs may change from system to system depending on witch packages a sys admin may have installed .
This is why i cant specify a specific package to grep .
And this is why i need to look into each package "Installed" field to see if have "(none)" or is different .

The command as given will do that - I copied your latest example into the file and here's the output:
That looks like the same ten entries as your expected output.

You can even remove "dovecot" (leaving just the ".*" part) and it will continue to work even with package names not starting with dovecot (that might not be required, just pointing out the potential).

dont know why but it is not working here .
what is your grep version ?
mine is 3.3 .
I have no idea if this is a grep version , but it is strange .