Shell Programming and Scripting

Code:

awk -F'~' '
  NR==FNR {a[$1]=$6;}
  NR!=FNR && a[$1]!=$6' file1 file2

(I do not testing, but used code is not good (besides bad written) and your understanding is not exactly correct.)

Code:

 'NR==FNR{p=$1;$1=x;A[p]=$0;next}

- what the 'p' and '$1=x' do here?!
Apsolutely useless! That is enough:

Code:

 'NR==FNR{A[$1]=$0;next}

... then append FS (a comma) ...
- No! FS is set to '[ \t]*' - spaces

... $2=A[$1] FS $2;print then append FS (a comma) followed by A[$1] ...
- ??? did you try to say ' a comma (in 'print') appends a FS? That would be correct!
- that statement do not print anything, but reassign the $2 (the seconf field) by the A[$1], followed by FS (whatever it is; in your code it a space), folloved by the original value of the $2.

Also, I think the 'FS=~' should be with the '\': 'FS=\~' - the '~' by shell going to be interpreted and substituted as a home-directory.

And, the used 'awk' (should it be 'nawk' to use the FNR and '?:'-operator?) code is really weird!
It reads first file with delimiter '~' into array A, indexing by first '~' field;
check the second file,having delimitered on spaces;
having the same first field in the second file as any line in the first file, it applys to the second space-delimitered field the line from the first file (with ~-delimiters) after the '~': so, the first field is just repeated.

I could hardly guess how after all the output could be used!

alex

you did not give alternative solution to the code.
Do you have one?

The awk solution did not work wither

What's wrong with yazu solution?! -(to your second part of your replay!)

(To your first statement: )
Actualy, you did not sound like you are looking for final line of code to do your job: that I most likely would not even start to responde!
I (seems wrongly) guess that you are looking for understanding how to go your stuff and for that bring up your work on (seems to you) simular piece of code.
I was trying to help you in that (in understanding).

To do your stuff after all explanation you should not have any problem, IMHO.

Speaking about the task: The answer is reflection of the question. In other words: 50% of an answer is a good question and in the question!
You did not ask enough clear to have a reasonable answer.
By discution I see that the delimiter in your file is the '~'.
Do you have any other restictionons?
Maybe files are big to be a problem to have the 1-st+6-th column be kept in memory (that I could see as a possible problem in yazu's solution).. maybe something else?

Alright, assume both file are delimitered by the same '~' delimiter.
Assume by 'comparing column 1 and column 6' means (as used later) file-2's '~'-delimitered columns 1 and 6 are not the same as any record's the same columns in the file-1.
Assume the 'do not match ' means (as, again, understand later) never been found in the file-1 in the same positions in any record simultaneosly.

On all these requirements the first 'yazu' solution completely fine: build a hash of (1-st + 6-th) columns of the first file and print the second file's record when the same combination (not by value, but the combination: 1-st + 6-th columns of the second file's record) hash member is not exist!

What the problem do you have with that solution?
Not clear? (Probably, already not, but) here are few words:
- Only on first (by the NR==FNR) file the hash 'a' is built (the '++' in not nessesary, just shortest to be sure that the such index hash member is any how referensed.)
- On the second file the check of the 'a' hash is in the condition of selection for an action that is not specified; so, default is used: print a full (current) record.
File's delimiter is defined as command line parameter by the -F option. That is done in single quotes; so the special meaning of the '~' is not processed by the shell.

What else?!
Perfect solution to the described by my assumptions' task!
Something is not good in assumptions? - Explain!
Something does not work? - show that; but, better think and try out by yourself. When you stuck - bring it here.
That way help would be helpfull!
Just to solve a task helps just to cheaters or liers, who pretends to know and has no any interest to the subject.

I do not believe anyone could be not agree with my statements!
That why I have no interest to help people who do not think on task, unable to present a question reasonably and do not care answers without complete and final solution of the task.

Good luck!


How it is not work?
Should it be 'nawk'?
Show some result of test!

the solution by yazu was correct.
my system could not use awk properly, it is setup to use nawk

this worked

Code:

nawk -F~ 'NR==FNR{A[NR]=$1$16;next}$1$16!=A[FNR]' f1 f2

thanks