Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 7 Nov 2017 21:53:38 +0300
From:      Yuri Pankov <yuripv@gmx.com>
To:        byrnejb@harte-lyne.ca
Cc:        freebsd-questions@freebsd.org
Subject:   Re: sed - remove nul lines from file
Message-ID:  <7fa1cb30-9a6f-a799-682b-1b352f1a1a58@gmx.com>
In-Reply-To: <f1f862f2f9cf02c5040c7c1acfa77d89.squirrel@webmail.harte-lyne.ca>
References:  <b21bf201363c34a90ab55c4a05ff8fd7.squirrel@webmail.harte-lyne.ca> <88a59a82-2902-9f63-0a94-bd23b910e7ad@gmx.com> <c2b1ffce6933bcb8f47c856a40d29b16.squirrel@webmail.harte-lyne.ca> <f51d6c8a-c91c-dc7c-6134-e276ec60b179@gmx.com> <88b1870184a8810072fe503917cd86be.squirrel@webmail.harte-lyne.ca> <c00693b7-ae1b-4aa8-49e2-81296cfb281f@gmx.com> <656b08b1-b79d-8e7e-27ec-e39d883f2355@gmx.com> <f1f862f2f9cf02c5040c7c1acfa77d89.squirrel@webmail.harte-lyne.ca>

next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 7 Nov 2017 13:46:44 -0500, James B  Byrne Via Freebsd-questions 
wrote:
> 
> On Tue, November 7, 2017 13:28, Yuri Pankov wrote:
>> On Tue, 7 Nov 2017 21:20:40 +0300, Yuri Pankov wrote:
>>> On Tue, 7 Nov 2017 13:14:08 -0500, James B  Byrne Via
>>> Freebsd-questions
>>> wrote:
>>>>
>>>> On Tue, November 7, 2017 13:03, Yuri Pankov wrote:
>>>>
>>>>>
>>>>> You want /d, not /g, to delete the *lines* which contain NUL
>>>>> symbols
>>>>> (that's what your subject line said).
>>>>>
>>>>
>>>> Sigh.  Thank you.  That works.  However, it also deletes any line
>>>> that
>>>> has even one NUL in it regardless of the presence of other non-nul
>>>> characters on the line.
>>>>
>>>> What I wish to accomplish is to delete only the lines that are
>>>> completely nul.  I thought that this could be accomplished by
>>>> prefacing the match sting with the start of line anchor ^ and
>>>> ending
>>>> it with the end of line anchor $ but this does not work as I
>>>> expect.
>>>
>>> "[[.NUL.]]" is just a character specified by its collation name, so
>>> treat as any other ordinary character:
>>>
>>> sed -E '/^[[.NUL.]]+$/d' INFILE > OUTFILE
>>>
>>> Need extended regexp here for '+' to work.
>>
>>
>> Or, after looking at re_format(7), it could be written using BREs,
>> your
>> choice :-)
>>
>> sed '/^[[.NUL.]]\{1,\}$/d'
>>
> 
> sed '/^[[.NUL.]]\{0,\}$/d' INFILE > OUTFILE
> 
> Which has no effect.  OUTFILE and INFILE remain identical.  I get the
> exact same result from the first invocation as well.  Likewise:
> 
> sed '/^[[.NUL.]]\{1,\}$/d' INFILE > OUTFILE
> 
> and
> 
> sed -E '/^[[.NUL.]]+$/d' INFILE > OUTFILE
> 
> # diff INFILE OUTFILE
> #
> # ll INFILE OUTFILE
> -rw-r--r--  1 root  wheel  61480 Nov  7 13:09 INFILE
> -rw-r--r--  1 root  wheel  61480 Nov  7 13:38 OUTFILE
> 
> 
> I had actually tried these combinations, or at least I believe that I
> tried these, before I wrote.  Given the complexity and arcane nature
> of whatever flavour of RE one is working with I may have transgressed
> and written them slightly differently.  But the examples you provided
> me with give the results I obtained exactly.
> 

Then there's probably something else on those lines with \0's as it 
works for me with a simple example created using the following command:

printf "abc\0\n\0\0\nd\0ef\n" > 1



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?7fa1cb30-9a6f-a799-682b-1b352f1a1a58>