Shell Programming and Scripting

Hi,
I have a file that I would like to translate using sed.
I can do some basic sed commands, but Im afraid this level is beyond me.
I have this file -
...alter...
..
...65536...
...
...65536...
...
...alter...
...
...65536...
etc
What I would like to do is replace the first occurrence of the 65536 after each line that has 'alter' in it (there may be serveral lines with 65536 between 'alter' lines) .
Any help would be much appreciated thanks.
Bob

If the input is always like the one you posted:

alter
one record
pattern to replace

Code:

% cat file
...alter...
..
...65536...
...
...65536...
...
...alter...
...
...65536...
% sed '/alter/{N;N;s/65536/whatever/;}' file
...alter...
..
...whatever...
...
...65536...
...
...alter...
...
...whatever...

If not (more records between the word alter and the pattern)
with Awk:

Code:

awk 'f&&/65536/{sub(/65536/,"whatever");f=0}/alter/{f=1}1' filename

Code:

awk 'f&&sub(/65536/,"whatever")&&f=0;/alter/{f=1}1' filename

Use nawk or /usr/xpg4/bin/awk on Solaris.

Thanks radoulov.
There may be any number of records between the
alter and 65536, so your awk code is on the mark.
I'll check it out first thing tomorrow (and yes, I'm on Solaris)
Do the two lines of code do the same?

Great work. I really appreciate the help.

The awk script works great thanks, but I've discovered that there may be instances where there is no line of '65536' between the 'alter' lines.
Is there any way to skip to the next 'alter', if a 65536 line does not exist?

Thanks

What do you mean by skip: you want to _ remove _ the records from alter to alter if there is not 65536 in between?

Hi,

This one should be ok for you.

code:

Code:

nawk ' 
{
if(f==1 && $0 ~ /65536/)
{
	sub(/65536/,"aaa")
	f=0
}
if ($0 ~ /alter/ )
	f=1
print
}' filename

Apologies for the confusion in the original problem statement. I should have given more detail.
The file I am processing contains SQL statements (It's from an Oracle database export).
For each table these is a block of statements (each statement ends with ';') like -
create table tablename ....
....
.... 65536 ....
.... ;
alter table tablename
............ ;
alter table tablename
.... 65536 .... ;
alter table tablename
.... 65536 .... ;
alter table tablename
............ ;
<next create statement>

For each table there will only be one 'create' statement, but any number of 'alter' statements.
Each 'alter' statement (and the 'create' statement) has a variable number of lines.
The 'alter' statements may or may not (I missed this on the original problem definition) have the '65536'.

I am trying to change the '65536' to a variable value (using a unix variable) which comes from a parameter file. The value will be something like '524288', but only for the 'alter' statements.

So the resulting file would be like -
create table tablename ....
....
.... 65536 ....
.... ;
alter table tablename
............ ;
alter table tablename
.... 524288 .... ;
alter table tablename
.... 524288 .... ;
alter table tablename
............ ;
<next create statement>

The awk code works fine until it hits the first 'alter' with no '65536' in it, when it changes the '65536' of the next occurrence, which may be in the 'create' statement.

Your efforts are much appreciated.
Regards
Bob