Awk or Sed ?

[butterfm]

Hi,

I'm struggling a bit with the following problem. I've tried awk and sed but can't get a solution.

I have a file containing text similar to the following (although I have cut bits out to fit it on this page)

NOT NULL, "PAGEHEIGHT" NUMBER(10, 0) NOT NULL) PCTFREE
40 INITRANS 1 MAXTRANS 255 STORAGE(INITIAL 20480 NEXT
1 MAXEXTENTS 2147483645 PCTINCREASE 0 FREELISTS 1
REM FREELIST GROUPS 1) TABLESPACE "FIN_CODAT" ;
REM ... 2 rows
CONNECT CUST_TEST;
CREATE UNIQUE INDEX "CUST_TEST"."OAS_ASM_INDEX1" ON "OAS_ASM" ("CODE" )
PCTFREE 10 INITRANS 2 MAXTRANS 255 STORAGE (INITIAL 20480 NEXT 20480
MINEXTENTS 1 MAXEXTENTS 2147483645 ;

What I need to do it replace the text, accross multiple lines, between the word "PCTFREE" and the next occuring ";" with another string, although I would want to keep semi colon ";"

e.g. Replacing the text with the word DOG would give me

NOT NULL, "PAGEHEIGHT" NUMBER(10, 0) NOT NULL) DOG ;
REM ... 2 rows
CONNECT CUST_TEST;
CREATE UNIQUE INDEX "CUST_TEST"."OAS_ASM_INDEX1" ON "OAS_ASM" ("CODE" )
DOG ;

I have managed this with sed when the text I want to replace only spans 2 lines but not when it spans more than 2 lines.

Any help with this would be greatly appreciated.

Thanks, Matt.

 

[jad]

is it all on one line? or is it broken like the example you gave above?

 

[butterfm]

I'm afraid it's broken like the example above.

Matt.

 

[jad]

you might be better using something like perl, reading the whole file into an object, and doing a match that can span multiple lines.

another alternative is that you could (using sed) join all lines together that have 'PCTFREE' in them that don't end with ';', then replace everything from 'PCTFREE' to ';'

hmm, just thought you could probably turn on a flag when you see PCTFREE, replace PCTFREE (and beyond on that line) with your text, and ignore all lines until you find a ';'

can do that in both awk and sed...

nawk '
{show=semi}
BEGIN {show=1; semi=0}
/;/ {show=1}
/PCTFREE/{
    if (index($0,";") == 0) semi=0;
    sub("PCTFREE.*$", "DOG;");
}
show==1 {print}' < filename

seems to work in awk ... sure there must be a nicer way to get around the index bit, and make it smaller, but it works

 

[jad]

actually not quite ...

nawk '
BEGIN {show=1; semi=1}
{show=semi}
/;/ {show=1}
/PCTFREE/{
    if (index($0,";") == 0) semi=0;
    sub("PCTFREE.*$", "DOG;");
}
show==1 {print}' < filename

need to set semi to 1 at the beginning, otherwise all lines leading up to the first colon are ommitted

 

[butterfm]

Thanks for your help Jad.

Much appreciated.

Matt.

 

[jad]

umm ... the '

/;/ {show=1}

' should have been '

/;/ {semi=1}

'

oops ... don't you just love cut and paste

 

[nwardez]

Hi,

Really want sed ? Try this sed script file

:jump
/PCTFREE[^;]*$/{
N
s/PCTFREE\(.*\)\n\(.*\)/PCTFREE\1 \2/
t jump
}
s/PCTFREE.*;$/DOG;/

 

[nwardez]

Still wonder why I wrote an unreadable "s/PCTFREE\(.*\)\n\(.*\)/PCTFREE\1 \2/" instead of simple "s/\n//" ...

 

[jad]

all of our examples assume that the semi-colon is the last thing on the line, and that there is never more than 1 semi-colon on a line.

since it looks like sql coding, anything after a semi-colon wouldn't be alowed, so it should never occur ...