i have a file where i need to remove text between the delimiters.

say...

delimiter1 = start_block
delmiter2 = no_block

so...

start_block

data....

no_block

i am new to using sed, but i think its something like this, with some gump dumped into the curly brackets at the end of the expression. there are also multiple instances of this in a file, i want to delete these globally.

sed -e '/start_block/,/no_block/{p;}'

you almost got it:

sed '/start_block/,/no_block/d' myfile

actually i tried your posting before you posted, it doesnt work.

heres the data i want to clean up:

start_block
aml_daily al_crossref al_general itm
--------- ---

0 record(s) selected.
no_block

start_block
aml_monthly al_crossref al_general itm
--------- ----
- 43060. 43060.
- 43060. 43060.
- 43060. 43060.
3 record(s) selected.
no_block

as you can see ive put in the delimiters as specified in my first posting.

when i use your idea, it just removes the space between the first appearnce of no_block to the next start_block - all blank lines??

im sure the fact that the first thing to appear in the file is a start_block delimiter is throwing this off from working.

anyone or other ideas would be most welcome, its been driving me nuts all day!

nails's command is correct.

/home/oracle/luke/tmp$ cat file
===>1
start_block
aml_daily al_crossref al_general itm
--------- ---

0 record(s) selected.
no_block

===>3

start_block
aml_monthly al_crossref al_general itm
--------- ----
- 43060. 43060.
- 43060. 43060.
- 43060. 43060.
3 record(s) selected.
no_block

===>2

/home/oracle/luke/tmp$ sed '/start_block/,/no_block/d' file
===>1

===>3

===>2

Luke Chi

I'm with Luke, I don't see your problem. According to the data file you posted, all that should print is blank lines. Perhaps I'm missing something. Also, I think this awk script emulates what you are trying to do:

awk ' BEGIN { skip=0 }

/start_block/ { skip=1; continue }

/no_block/ { skip=0; continue }

skip == 0 { print $0 }
' myfile

The "algorithm" for such task is typically like this

flag = 1
for lines in the_file:
... if lines is equal "start_block"
...............print lines
...............flag = 1
... if flag is 0:
...............print lines
... if lines is equal "no_block":
...............print lines
...............flag = 0

OK, let me clarify this one more time, im sure Lukes command is fantastic, but as I say its not doing what i require:

My file looks like this:

start_block
aml_daily al_crossref al_general itm
--------- ---

0 record(s) selected.
no_block

===>3

start_block
aml_monthly al_crossref al_general itm
--------- ----
- 43060. 43060.
- 43060. 43060.
- 43060. 43060.
3 record(s) selected.
end_block

start_block
aml_annual al_crossref al_general itm
--------- ---

0 record(s) selected.
no_block

i want to remove all text between the start_block and no_block delimiters ive set,
because i want to retain the data between the start_block and end_block.

as i said earlier, after running the delete command above this is what happens:

start_block
aml_daily al_crossref al_general itm
--------- ---

0 record(s) selected.
no_block
start_block
aml_monthly al_crossref al_general itm
--------- ----
- 43060. 43060.
- 43060. 43060.
- 43060. 43060.
3 record(s) selected.
end_block

start_block
aml_annual al_crossref al_general itm
--------- ---

0 record(s) selected.
no_block
start_block
aml_testr al_crossref al_general itm
--------- ---

0 record(s) selected.
no_block

blank lines between occurences of no_block and start_block are removed??

1. nails offered the command, not me. "im sure Lukes command is fantastic" is not correct.

2. sed '/start_block/,/no_block/x' file will create the wrong result you created and described.

3. make sure using:
sed '/start_block/,/no_block/d'
not:
sed '/start_block/,/no_block/x'

Luke Chi

The delimiters you use make things complicated. You shouldn't have used start_block for both kinds of the blocks - want to keep block and wnat to remove block.

For example, the following delimiter names will make things much easiler:

start_keep, end_keep
start_remove, end_remove

Luke Chi

what he wants is conditionnal testing of which "end" delimiter you have.

If the end delimiter is "no block" you trash all data in between. If it's "end_block" you keep all data in between the two delimiters.

It's trickier than I know since you need to test all lines past the "start_block" command...till one of the two end delimiters is found...

This awk solution will delete each block of lines that starts with start_block and terminates with no_block rather than end_block.

The logic is that it will print each line as it comes to it UNTIL it begins a block of lines that starts with start_block.  At that point, it will buffer the lines until the block terminates with either no_block or end_block, at which point it will either discard or print the held lines, and so on.

awk '{
if ($0=="start_block")
   flag=1
if (flag!=1)
  {print
   next}
datablock=datablock newline $0
newline="\n"
if ($0=="no_block")
  {datablock=""
   flag=0
   newline=""}
if ($0=="end_block")
  {print datablock
   datablock=""
   flag=0
   newline=""}
}' myfile

I also coded this as a sed one-liner, which uses the very same approach (and I think it took me longer to code this one-liner).

sed -e '/start_block/!b' -e :a -e '$!N' -e /no_block/d -e /end_block/b -e '$!ba' myfile

Report Offensive Follow Up For Removal