Very Computer

I am writing a script in Freebsd 4.9 using /bin/sh shell to add and remove
users for an ftp server running proftpd.  I have everything working fine and
can add the follwing to proftpd.conf ok.  However, the rmuser part removes
everything except the lines added to the conf file for this account.  I have
tried using sed to find / replace them but can only manage one line at a
time and some of the lines are in other users accounts.   If each accoiunt
has the same lines added with only the $user different, which is passed to
the rmuser code, is there a way of safely finding and deleting a message
block.

If I could find the line number <Anonymous ~$user> appears on and deleted
the range from this line number + 21 for example.

I iknow there are perl scripts etc.. for doing this, but I really want to
learn how to write threse for myself.
Any help appreciated.

Jess

lines added to proftpd.conf

<Anonymous ~$user>
   User   $user
   Group  nologin

  ### Limit the maximum number of anonymous logins
   MaxClients   10

  ### We want 'welcome.msg' displayed at login, and '.message' displayed
  ### in each newly chdired directory.
   DisplayLogin   welcome.msg
   DisplayFirstChdir  .message

RequireValidShell off
AnonRequirePassword on

<Directory /home/$user>
  AllowOverwrite  on
</Directory>
</Anonymous>

Quote:> Hi

> I am writing a script in Freebsd 4.9 using /bin/sh shell to
> add and remove users for an ftp server running proftpd.
>  I have everything working fine and can add the follwing
> to proftpd.conf ok.  However, the rmuser part removes
> everything except the lines added to the conf file for this
> account.  I have tried using sed to find / replace them but
> can only manage one line at a time and some of the lines
> are in other users accounts.   If each accoiunt has the same
> lines added with only the $user different, which is passed to
> the rmuser code, is there a way of safely finding and deleting
> a message block.

> If I could find the line number <Anonymous ~$user> appears
> on and deleted the range from this line number + 21 for example.

> I iknow there are perl scripts etc.. for doing this, but I really
> want to learn how to write threse for myself.
> Any help appreciated.

    <user1 ...>
     ... user1 data
     ...
    <user2 ...>
     ... user2 data
     ...
    <user3 ...>
     ... user2 data
     ...

and you wished to delete the entire <user2> block, you could create a simple
'sed' script, something like:

    /^user2/,/^user3/{/^user3/b; d;}

and execute:

    sed -f myscript olduserdata > newuserdata

If you used an 'end-of-user-block' marker [say as a comment line]:

    <user1 ...>
     ... user1 data
     ...
    #enduser1
    <user2 ...>
     ... user2 data
     ...
    #enduser2
    <user3 ...>
     ... user2 data
     ...
    #enduser3

this would be even easier, as you wouldn't need to know what the user entry
immediately following the target user was, and could instead do:

    /^user2/,/^\#enduser2/d

You could, as you say, use Perl [or awk, perhaps ?]. How about something
like the following awk script [executed as: awk -f myscript olduserdata >
newuserdata]:

  BEGIN { nlines = 21; }

  {
    if (/^user2/) { nlines = 0; }
    if (nlines >= 21) { print($0); }
    nlines++;
  }

which accomplishes the 'find and delete next 21 lines' method you mentioned.
It's a little difficult to perform this using 'sed' !

I hope this helps.

Anthony Borla

P.S.

The above examples are shown in a verbose style for clarity. You can,
certainly, shorten them by:

* Turning them into 'one-liners' placed directly on the command-line
* Use terser expressions [in awk - same applies to Perl]

This works really well.

sed "/^<Anonymous ~$user>/,/<\/Anonymous>/ d" proftpd.conf > /tmp/tmpfile &&
cat /tmp/tmpfile > proftpd.conf

I think ill make it keep a backup copy of the conf file on each deletion
just to be sure though.

Jess

Quote:> Hi

> I am writing a script in Freebsd 4.9 using /bin/sh shell to add and remove
> users for an ftp server running proftpd.  I have everything working fine
and
> can add the follwing to proftpd.conf ok.  However, the rmuser part removes
> everything except the lines added to the conf file for this account.  I
have
> tried using sed to find / replace them but can only manage one line at a
> time and some of the lines are in other users accounts.   If each accoiunt
> has the same lines added with only the $user different, which is passed to
> the rmuser code, is there a way of safely finding and deleting a message
> block.

> If I could find the line number <Anonymous ~$user> appears on and deleted
> the range from this line number + 21 for example.

> I iknow there are perl scripts etc.. for doing this, but I really want to
> learn how to write threse for myself.
> Any help appreciated.

> Jess

> lines added to proftpd.conf

> <Anonymous ~$user>
>    User   $user
>    Group  nologin

>   ### Limit the maximum number of anonymous logins
>    MaxClients   10

>   ### We want 'welcome.msg' displayed at login, and '.message' displayed
>   ### in each newly chdired directory.
>    DisplayLogin   welcome.msg
>    DisplayFirstChdir  .message

> RequireValidShell off
> AnonRequirePassword on

> <Directory /home/$user>
>   AllowOverwrite  on
> </Directory>
> </Anonymous>

Quote:

> Am I missing something?

I can honestly say I haven't seen an example of *relative* line addressing
in any of the 'sed' examples I've accessed recently. Take a look at some of
the entries at:

    http://sed.sourceforge.net/grabbag/tutorials/

I've seen mention of:

* Absolute Line Address [e.g. 1,10]
* Context Addresses [/re1/,/re2/]

but absolutuely nothing of *relative* line address.

I was, therefore, under the impression any such line counts had to be
performed manually [i.e. by storing and updating values in the hold space].
Given this potential complexity, a solution requiring this approach is much
more easily implemented in 'awk', and, therefore, was the solution
suggested.

I can only assume, given the age of some of the links, and the non-use of
relative line addressing in the examples I saw, that relative line
addressing in 'sed' is a relatively [no pun intended :)] new feature ?

Cheers,

Anthony Borla

Got it working with the following

sed "/^<Anonymous ~$user>/,/<\/Anonymous>/ d" proftpd.conf > /tmp/tmpfile &&
cat /tmp/tmpfile > proftpd.conf

Jess

#sed "/<Anonymous ~test>/,+21d" filename
sed: 1: "/<Anonymous ~test>/,+21d": expected conrext address
#

#sed "/<Anonymous ~test>/,21d" filename
Removes only the line <Anonymous ~test> and not the following 21, I have
tried different amounts with same result
#

If you can spot what I am doing wrong I would appreciate the advice.

Thanks Again

Jess

Quote:

> I seem to be unable to manage the same result.

> #sed "/<Anonymous ~test>/,+21d" filename
> sed: 1: "/<Anonymous ~test>/,+21d": expected conrext address
> #

> #sed "/<Anonymous ~test>/,21d" filename
> Removes only the line <Anonymous ~test> and not the
> following 21, I have tried different amounts with same result
> #

> If you can spot what I am doing wrong I would appreciate
> the advice.

// -----------------
user1
sss
sss
ggg
ffff
ddd
#enduser
user2
sss2
sss2
ggg2
ffff2
ddd2
#enduser
user3
sss3
sss3
ggg3
ffff3
ddd3
#enduser

// -----------------

Executed this command line:

    sed "/^user2/,+6d" datafile

Obtained this result:

user1
sss
sss
ggg
ffff
ddd
#enduser
** user2
** sss2
** sss2
** ggg2
** ffff2
** ddd2
** #enduser
user3
sss3
sss3
ggg3
ffff3
ddd3
#enduser

All the lines marked with '**' were excluded from the output result - they
were, effectively, deleted.

I would say what is happening is that we are each using a different version
of 'sed'. The one I used to perform the above test, and that used by Icarus
must either be later versions, or simply be versions that contain [possibly]
non-standard extensions. This would certainly explain why I had not seen the
use of relative line addresses in any of the 'sed' code samples I had looked
at.

I suggest reading through the 'sed' FAQ [see links I earlier posted] for
more information about this.

So, to solve your problem:

* Use one of my earlier suggestions [i.e. awk, or alter data file]
* Download a [later ?] version of 'sed' that incorporates these
   extensions

I would ordinarily spend more time solving this 'mystery' but am currently a
little timestrapped :) !

I hope this helps.

Anthony Borla

I read throiugh the links you posted and have saved all the one liners for
future reference.

This definitely does not work on my system, I assume compatability as this
method is also not mentioned anywhere in the links posted - most likely for
this reason.

As I said, this works fine on my system and is a method used in the links
pages so hopefully has more widespread support.

sed "/^<Anonymous ~$user>/,/<\/Anonymous>/ d" filename

Jess

#!/bin/sh
for user in $*
do
 sed "/^<Anonymous ~$user>/,/^<\/Anonymous>/ d"
done

--
Michael Tosch
IT Specialist
HP Managed Services Germany
Phone +49 2407 575 313