Search w/offset and Replace Script

[kevinesharp]

I have the need for a script that takes in a large test file exported from a database, searches for an expression and replaces a string offset from the expression. I can do it in vi, and can even script it, but it is not efficient. I need to replace am attribute associated with a tag name for about 20 devices, and repeat the process about 100 times. That is 2000 Search and replaces over 100,000 lines.

Example:
vi {file}

{file}(only one line #83939)
00083939|Global Register 04 Bit12DPU LS13|di_glbl_r04_b12 | | |undefine |000000000000| |00000999|000000000000|00|000000000000|05620|00061 |00000|255|255|255|000|FALSE |00000|011|001|TRUE |00000|002|002|TRAVEL |00000|005|003|INVALID |00000|001|00000|1|0|0|0|0|1|0|0|0|0|0|0|0|0|0|0|008|0|0|0|0|0|000 000000000|002|00000|004| 0|05|000000081844|0|0|0|0|0|0|0|0|0|001|TRUE |00000|002|0|0|0|0|000|FALSE |00000|011|0|

:source! {script}

{script}
/di_glbl_r04_b12/s+274
r3lr1

{script needed}
start at line 1
find di_glbl_r04_b12
move forward 277 characters
replace 00000 with 00031
repeat until end of file

start at line 1
find di_glbl_r05_b00
move forward 277 characters
replace 00000 with 00031
repeat until end of file

...

{result}
00083939|Global Register 04 Bit12DPU LS13|di_glbl_r04_b12 | | |undefine |000000000000| |00000999|000000000000|00|000000000000|05620|00061 |00000|255|255|255|000|FALSE |00000|011|001|TRUE |00000|002|002|TRAVEL |00000|005|003|INVALID |00000|001|00031|1|0|0|0|0|1|0|0|0|0|0|0|0|0|0|0|008|0|0|0|0|0|000 000000000|002|00000|004| 0|05|000000081844|0|0|0|0|0|0|0|0|0|001|TRUE |00000|002|0|0|0|0|000|FALSE |00000|011|0|

I am a day and a half into this and am new to Linux. Any suggestions?
Thanks for any help I can get!

[Irithori]

Would it be possible, to do that change in the database?
Some UPDATES for that row with a WHERE = di_glbl_r05_b00, di_glbl_r04_b12, etc?

Even if a sed/perl/whatever script
a) can be developed (which Iam sure off, just cant think of a quick solution)
b) would be faster
then this script would still be hack-ish: If the data changes just a little bit, then the "move forward 277 bytes" wont work anymore..

[kevinesharp]

I can make the changes in the database, but only through a form which is very slow.

The data will not change because the column width of each cell is fixed. That is the space between each pipe (|). The system is proprietary, so I am kind of stuck with using export and import utilities.

The problem I found with sed is that I cannot offset from the pattern.

====

So, the database is a .dbd file, but not mysql. The db porcesses are dbstart, dbnam, and dbproc when the project is brought online. The export creates .asc files which, I guess is ascii flat files. Maybe this will help?

[Kloschüssel]

use sed with something like this:

's/^(|.*[^|]){n}|00000|/\1|00031|/g'

what follows is an explanation of the dark magic above.

first of all select the portion you want:
^ := beginning of line
(|[^|]*){n} := exactly n-times a "|data" with the end character [^|]
|00000| := the search string

then replace it:
\1 := backreference to the first part of the line (group 1)
|00031| := the replacement

PS: untested code snippet, shipped without warranty

[Irithori]

The problem, as I understand it, he doesnt want to change all lines.

But only the ones with a special "id", like di_glbl_r04_b12, di_glbl_r05_b00, etc
And these seem to be random, ie following no pattern..

[Kloschüssel]

Then he may just need to adapt the part with:

(|[^|]*){n}

to match his needs. It's not the final solution, but more a good hint.

[kevinesharp]

How about using awk? Each row is 86 cells wide. I could search for a match in $3 and replace $35 with 00031 or 00000 based on the $3 match. I just need to "or" the search strings and "print" $1"|"$2"|"..."|"variable"|"$36...

I am trying it now.

[Kloschüssel]

awk can handle it too. in the end awk is even more powerful than sed could ever be (as sed only accepts regular expressions and awk comes shipped with a scripting language than can match context aware grammars).

[kevinesharp]

Okay, I got awk to work. I am using

awk -F '|' -f script.sh file.acs > newfile.acs

The script has a if else statement.

if ($3 == "di_globl_r01_b13 " || $3 == "..." || $3 == "..." )
VAR = "00031"
else
VAR = $35

Now the problem is that there are a few that are getting changed that should not. I only want to change any that has DPU in $2 (Description Field) and not TPU.

Does anyone know how to include an && $2 "*DPU*" in the if statement? As far as I can tell the wild card does not work. It might look like:

if ($2 == “*DPU*” && ($3 == "di_globl_r01_b13 " || $3 == "..." || $3 == "...")

[kevinesharp]

Does anyone know how to use a wild card in awk?