[comp.unix.questions] Re^2: SED Question

maart@cs.vu.nl (Maarten Litmaath) (04/13/89)

jgrace@bbn.com (Joe Grace) writes:
\...
\Yes, there is a way to get sed to do this, but you have to be
\wary of sed's shortcomings.

[shell script workaround deleted]

\If sed had a way of handling an EOF without quitting, the
\[workaround would be unnecessary] [...]

My first sed attempt can easily be fixed to handle partial matches at EOF
properly:

sed -n '
        /^PROMPT>$/{
		$p
                h
                n
                H
                /^>$/{
			$b eof
                        n
                        /^<$/b
                        H
                }
		: eof
                g
        }
        p
'
-- 
 "If it isn't aesthetically pleasing, |Maarten Litmaath @ VU Amsterdam:
  it's probably wrong." (jim@bilpin). |maart@cs.vu.nl, mcvax!botter!maart

leo@philmds.UUCP (Leo de Wit) (04/14/89)

In article <2292@solo11.cs.vu.nl> maart@cs.vu.nl (Maarten Litmaath) writes:
    []
|My first sed attempt can easily be fixed to handle partial matches at EOF
|properly:
|
|sed -n '
|        /^PROMPT>$/{
|		$p
|                h
|                n
|                H
|                /^>$/{
|			$b eof
|                        n
|                        /^<$/b
|                        H
|                }
|		: eof
|                g
|        }
|        p
|'

I think you'll need a third attempt 8-). It still doesn't handle correct
a series of lines like 

      PROMPT>
      PROMPT>
      >
      <

or one like

      PROMPT>
      >
      PROMPT>
      >
      <

Alternative:

sed -n '
: start
$p
$q
N
/^PROMPT>\n>\n<$/d
/\n[^\n]*\n/{
    P
    D
}
b start'

    Leo.

rupley@arizona.edu (John Rupley) (04/14/89)

In article <2292@solo11.cs.vu.nl>, maart@cs.vu.nl (Maarten Litmaath) writes:
> jgrace@bbn.com (Joe Grace) writes:

[the problem is to delete all three-line sequences:
		PROMPT>
		>
		<
]

> \...
> \Yes, there is a way to get sed to do this, but you have to be
> \wary of sed's shortcomings.
> 
> [shell script workaround deleted]
> 
> \If sed had a way of handling an EOF without quitting, the
> \[workaround would be unnecessary] [...]
> 
> My first sed attempt can easily be fixed to handle partial matches at EOF
> properly:

[sed script deleted]

Neat script (of course), but it still fails with a partial sequence
followed by a proper sequence, such as:

		PROMPT>
		>
		PROMPT>
		>
		<

The ``restart'' loop below takes care of it:

sed -n '
	: restart
	/^PROMPT>$/{
		h
		$p
		n
		/^>$/{
			H
			${x;p;}
			n
			/^<$/d
		}
		x
		p
		x
		b restart
	}
	p
'

BTW -- the following one line Lex source does it all:

%%
PROMPT>\n>\n<\n		;

and its hard to get the logic wrong (:-)!


John Rupley
rupley!local@megaron.arizona.edu

danny@itm.UUCP (Danny) (04/14/89)

Uh,

    Why not use the sed comma operator?  Like:
    
        sed '/^PROMPT>/,/^>/d' file
        
    I got the impression that the original poster didn't care WHAT
lay between the first and last match lines.

                                        Danny
-- 
				Daniel S. Cox
				(seismo!gatech!itm!danny)

uucibg@sw1e.UUCP (3929]) (04/15/89)

In article <1003@philmds.UUCP> leo@philmds.UUCP (Leo de Wit) writes:
=I think you'll need a third attempt 8-). It still doesn't handle correct
=a series of lines like 
=
=      PROMPT>
=      PROMPT>
=      >
=      <
=
=or one like
=
=      PROMPT>
=      >
=      PROMPT>
=      >
=      <
=
=Alternative:
=
=sed -n '
=: start
=$p
=$q
=N
=/^PROMPT>\n>\n<$/d
=/\n[^\n]*\n/{
=    P
=    D
=}
=b start'
=
=    Leo.

I think I may have missed a posting or two on this topic.  Could someone
please tell me why the following does not handle the problem:

cat myfile | sed -e '/^PROMPT>$/,/^<$/d'


Many thanks,

Brian R. Gilstrap                          Southwestern Bell Telephone
One Bell Center Rm 17-G-4                  ...!ames!killer!texbell!sw1e!uucibg
St. Louis, MO 63101                        ...!bellcore!texbell!sw1e!uucibg
(314) 235-3929
#include <std_disclaimers.h>