Re: [9front] sed bug?

Re: [9front] sed bug?

[umbraticus]

> This works fine, it outputs test2
> {echo test; echo test1; echo test2} | sed '0,/^test1$/d'

really? I get: sed: line number 0 is illegal

for this one: 

>{echo test; echo test1; echo test2} | sed '1,/^test$/d' # has no output

it deletes line 1, then continues deleting lines until it finds ^test$
which it never does. Perhaps adding a newline at the start is the quickest
way to get what you want?

{echo; echo test; echo test1; echo test2} | sed '1,/^test$/d'

test1
test2

umbraticus

Re: [9front] sed bug?

[Joe M]

Hello,

umbraticus@prosimetrum.com wrote:
> > This works fine, it outputs test2
> > {echo test; echo test1; echo test2} | sed '0,/^test1$/d'
>
> really? I get: sed: line number 0 is illegal

yes, my bad. I pasted the wrong command. It should be a 1 instead of 0.

> for this one:
>
> >{echo test; echo test1; echo test2} | sed '1,/^test$/d' # has no output
>
> it deletes line 1, then continues deleting lines until it finds ^test$
> which it never does. Perhaps adding a newline at the start is the quickest
> way to get what you want?
>
> {echo; echo test; echo test1; echo test2} | sed '1,/^test$/d'
>
> test1
> test2
>
> umbraticus

I stumbled upon the same workaround too. On a similar note, this
command works fine also:

{echo test; echo test1; echo test2} | sed '1,1d'

Thanks again

Re: [9front] sed bug?

[Eckard Brauer]

[-- Attachment #1: Type: text/plain, Size: 1199 bytes --]

At least it look inconsistent:

% {echo test; echo test1; echo test2;echo test3} | sed '1,/^test$/d'
% {echo test; echo test1; echo test2;echo test3} | sed '1,/^test1$/d'
test2
test3
% {echo test; echo test1; echo test2;echo test3} | sed '1,/^test2$/d'
test3

seems search starts just after addressed line (1 in this case): in fist
case nothing is found, so search goes until end, then mothing remains
for output.

as the man page states:

	[address [, address] ] function [argument ...]

as valid, "1/test$/d' should work, but it seems not to be the case:

% {echo test; echo test1; echo test2;echo test3} | sed '1/^test$/d'
sed: Unrecognized command: 1/^test$/d

BTW, GNU sed (Linux) seems to behave similar.

Am Tue, 13 Nov 2018 13:04:01 -0700
schrieb Joe M <joe9mail@gmail.com>:

> Hello,
> 
> umbraticus@prosimetrum.com wrote:
>  [...]  
>  [...]  
> 
> yes, my bad. I pasted the wrong command. It should be a 1 instead of
> 0.
> 
>  [...]  
>  [...]  
>  [...]  
> 
> I stumbled upon the same workaround too. On a similar note, this
> command works fine also:
> 
> {echo test; echo test1; echo test2} | sed '1,1d'
> 
> Thanks again



-- 
:)

[-- Attachment #2: Digitale Signatur von OpenPGP --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

Re: [9front] sed bug?

[Dave MacFarlane]

"1/test$/d" doesn't match the pattern "[address [, address] ] function
[argument ...]". There's a comma that separates the two addresses in
the documentation.
On Tue, Nov 13, 2018 at 3:28 PM Eckard Brauer <eckard.brauer@gmx.de> wrote:
>
> At least it look inconsistent:
>
> % {echo test; echo test1; echo test2;echo test3} | sed '1,/^test$/d'
> % {echo test; echo test1; echo test2;echo test3} | sed '1,/^test1$/d'
> test2
> test3
> % {echo test; echo test1; echo test2;echo test3} | sed '1,/^test2$/d'
> test3
>
> seems search starts just after addressed line (1 in this case): in fist
> case nothing is found, so search goes until end, then mothing remains
> for output.
>
> as the man page states:
>
>         [address [, address] ] function [argument ...]
>
> as valid, "1/test$/d' should work, but it seems not to be the case:
>
> % {echo test; echo test1; echo test2;echo test3} | sed '1/^test$/d'
> sed: Unrecognized command: 1/^test$/d
>
> BTW, GNU sed (Linux) seems to behave similar.
>
> Am Tue, 13 Nov 2018 13:04:01 -0700
> schrieb Joe M <joe9mail@gmail.com>:
>
> > Hello,
> >
> > umbraticus@prosimetrum.com wrote:
> >  [...]
> >  [...]
> >
> > yes, my bad. I pasted the wrong command. It should be a 1 instead of
> > 0.
> >
> >  [...]
> >  [...]
> >  [...]
> >
> > I stumbled upon the same workaround too. On a similar note, this
> > command works fine also:
> >
> > {echo test; echo test1; echo test2} | sed '1,1d'
> >
> > Thanks again
>
>
>
> --
> :)



-- 
- Dave

Re: [9front] sed bug?

[Eckard Brauer]

[-- Attachment #1: Type: text/plain, Size: 618 bytes --]

You're right, as the result of the search is an address.

But in that case, 1,/pattern/ applies the function d to the first line,
as 1,<address> evaluates to 1,1. So it points out to be a bug in sed
(even in GNU sed).

Am Tue, 13 Nov 2018 15:32:12 -0500
schrieb Dave MacFarlane <driusan@gmail.com>:

> "1/test$/d" doesn't match the pattern "[address [, address] ] function
> [argument ...]". There's a comma that separates the two addresses in
> the documentation.
> On Tue, Nov 13, 2018 at 3:28 PM Eckard Brauer <eckard.brauer@gmx.de>
> wrote: [...]  
>  [...]  
>  [...]  
> 
> 
> 



-- 
:)

[-- Attachment #2: Digitale Signatur von OpenPGP --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

Re: [9front] sed bug?

[Antons Suspans]

On Tue, Nov 13, 2018 at 09:28:06PM +0100, Eckard Brauer wrote:
> BTW, GNU sed (Linux) seems to behave similar.

The man page of GNU sed describes this behavior explicitly:

"if addr2 is a regexp, it will not be tested against the line that addr1 matched"

(Man pages of Plan 9 are typically less tormenting.)

Re: [9front] sed bug?

[Eckard Brauer]

[-- Attachment #1: Type: text/plain, Size: 742 bytes --]

What does that mean, and where does the regexp apply? This had to be
documented, as the case shows.

Either ther regexp applies beginning the line from addr1 (what I'd
treat as the expected behaviour), we have a bug in *sed, or the pattern
applies from the line /following/ addr1, then it had to be clearly
documented, what isn't the case.



Am Tue, 13 Nov 2018 22:43:54 +0200
schrieb Antons Suspans <antox@ml.lv>:

> On Tue, Nov 13, 2018 at 09:28:06PM +0100, Eckard Brauer wrote:
>  [...]  
> 
> The man page of GNU sed describes this behavior explicitly:
> 
> "if addr2 is a regexp, it will not be tested against the line that
> addr1 matched"
> 
> (Man pages of Plan 9 are typically less tormenting.)



-- 
:)

[-- Attachment #2: Digitale Signatur von OpenPGP --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

Re: [9front] sed bug?

[Eckard Brauer]

[-- Attachment #1: Type: text/plain, Size: 649 bytes --]

OK, got it just when reading again. Please ignore preceding mail.

Am Tue, 13 Nov 2018 22:07:15 +0100
schrieb Eckard Brauer <eckard.brauer@gmx.de>:

> What does that mean, and where does the regexp apply? This had to be
> documented, as the case shows.
> 
> Either ther regexp applies beginning the line from addr1 (what I'd
> treat as the expected behaviour), we have a bug in *sed, or the
> pattern applies from the line /following/ addr1, then it had to be
> clearly documented, what isn't the case.
> 
> 
> 
> Am Tue, 13 Nov 2018 22:43:54 +0200
> schrieb Antons Suspans <antox@ml.lv>:
> 
>  [...]  
> 
> 
> 



-- 
:)

[-- Attachment #2: Digitale Signatur von OpenPGP --]
[-- Type: application/pgp-signature, Size: 195 bytes --]