Cut part of txt file with specified terms....

[LITWINCZUK]

Normal procedure is started with: (A)
00000000.3987 2010-04-13 09:04:22 argv[0]: './encrypt'
00000000.3987 2010-04-13 09:04:22 argv[1]: './'

And end with: (B)
00000014.9556 2010-04-13 09:00:21 SendMessage: wiadomosc wyslana
00000015.1078 2010-04-13 09:00:22 wiadomosc wyslana
00000015.1078 2010-04-13 09:00:22 ~Message()


But sometime i have error (C):
00000006.3951 2010-04-13 09:04:28 Encrypt ret: 7
00000006.3951 2010-04-13 09:04:28 blad szyfrowania
00000006.3951 2010-04-13 09:04:28 ~Message()

Normal pass is: (A) ..... (B) etc....

Error: (A) ..... (C) (no B here !)

I need to get:
1. All errors from file - section A -> C (there are no B in this pass), but without any normal, sucessfull termination (section A->B)
2. Only last error pass.

Can i do this with sed command?
To cut all errors, only one A-C per error ?

[LITWINCZUK]

and one again:
A - unique in this section is
argv[0]: './encrypt'

C - unique text is
Encrypt ret: 7

[Dennis Handly]

>Can I do this with sed command?

This seems much too complex for sed. awk may be better.

>To cut all errors, only one A-C per error?

You are going to have to provide much more details, actual files. I'm not sure what is A, B or C.

Perhaps this is easier, what do you not want?

[LITWINCZUK]

There are:

1. normall pass:
start with section A, next some other lines... and end with B.

2. Pass with error: starti with A, some other lines... end with C.


Of course, as normal program function there are many A-B in log and only few with A-C.

I need to cut only sections with A-C and throw away all A-B.

When i trying something like
sed -n -e '1,/argv\[0\]: /p' -e '1,/Encrypt ret: 7/p' plik.txt

i get start with A.... (first from start of file) to first error - C section. But with all A-B in this file. I must remove them.
Best is cut on first C section and go backwards line by line untill i get first A section.
But i dont know to do it :/

[Dennis Handly]

How many lines are in A, B and C, exactly like you have except for the leading dates/time?

What varies?

>Best is cut on first C section and go backwards line by line until I get first A section.

Backwards doesn't work well unless you have arrays, with awk or perl.

Do you have some example data?

[LITWINCZUK]

>How many lines are in A, B and C, exactly like you have except for the leading dates/time?

This is different. Dont look at date.
Sections like "00000000.3987 2010-04-13 09:04:22" is always different. Number of lines between beginning and end is always different.

I want to focus on:
argv[0]: './encrypt'
Encrypt ret: 7
SendMessage: wiadomosc wyslana

Upper words is unique in logs. If i try to catch all lines between it, it will work fine.


>Do you have some example data?
There are some data, also uuencoded area. This is log from delta encryptor - so i cannot post all of them.

[James R. Ferguson]

Hi:

Dennis asked you "How many lines are in A, B and C, exactly like you have except for the leading dates/time?"

You haven't answered this question. Are there always some set number of lines to A, B and/or C? Is there some specific delimiter between A, B, and C as for example a blank line?

Regards!

...JRF...

[James R. Ferguson]

Hi (again):

You might try playing and adapting something like this:

# cat ./extract
#!/usr/bin/perl
use strict;
use warnings;
my $AVAL='a'; #...change to what you want to match...
my $BVAL='b'; #...ditto
my $CVAL='c'; #...ditto
my ( @list, $rangeop );
sub spout {
#...the list must contain $AVAL and $CVAL to be a candidate...
@list = grep( !/$BVAL/, @list );
if ( grep( /$CVAL/, @list ) ) {
print ">>> $_\n" for @list;
}
@list=();
}
while (<>) {
chomp;
if ( $rangeop = /$AVAL/../$CVAL/ ) {
@list = () if /$AVAL/;
push @list, $_;
}
if ($rangeop =~ m/E0/) { #...signifies range's end...
spout();
}
}
END {
spout();
}
1;

...which if run interactively (where lines with ">>> " are the output generated at that point in the line-by-line input) would look like:

# ./extract
a
b
a
b
c
>>> a
>>> c
aaa
xxx
yyy
zzz
bbb
bbb
ccc
>>> aaa
>>> xxx
>>> yyy
>>> zzz
>>> ccc
a
c
>>> a
>>> c
a
b
a
a
a
b
b
c
>>> a
>>> c

Regards!

...JRF...

[Dennis Handly · Accepted solution]

Assuming you start with "argv[0]:" and end with "Encrypt ret: 7" and 2 more lines or "SendMessage: wiadomosc wyslana" and 2 more lines you could use this awk script:
awk '
BEGIN { count=0; collect=0 }
function print_lines() {
for (i = 1; i <= count; ++i)
print save[i]
count = 0;
}
function save_line() {
save[++count] = $0
}
/argv\[0]:/ { # A start
collect = 1
save_line()
next
}
/Encrypt ret: 7/ { # C error
save_line() # save 3
getline; save_line()
getline; save_line()
print_lines()
collect = 0
next
}
/SendMessage: wiadomosc wyslana/ { # B normal
getline; getline # flush 3
count = 0
collect = 0
next
}
collect {
save_line()
next
}' input-file

[LITWINCZUK]

THANKS!