Re: A Simple Text-Region Deletion Question . . .

Greetings Daniel,

On Sun, Oct 05, 2014 at 11:24:19PM -0700, Daniel Goldman [email protected] wrote:
> I don't know why one would want to delete that kind of block, since it's
> not symmetrical. But you must have your reasons, it's better to just
> answer the question, as Ronaldo did. And as you say, it led to a great
> example to work through. sed can do an incredible amount in one line.
> I'd be interested in how Ronaldo came up with the solution, kind of his
> thinking process, if he cares to explain.
(Maybe I can't explain very well in English yet...)

I start thinking about how to put entire block in pattern space to
simplify tests and other manipulations:

/<property name="name">/ {         # search for regexp and start a block
    :loop;
    /<\/property>.*<\/property>/ { # when pattern space had two end-tags
        s/.*//;                    # substitute for nothing (I made a
                                   # mistake here, but your code fix
                                   # this)
        b;                         # avoid the next loop (not really
                                   # needed in this case)
     };
     N;                            # append next line
     b loop;                       # jump to label ":loop"
};


> FWIW, here is a variation of Ronaldo's solution that is perhaps a little
> better and simpler:
>
> $ cat file.xml
> Stuff that comes before
> More stuff that comes before
> <property name="name">
> ...
> </property>
> </property>
> Stuff that comes after
> More stuff that comes after
> <property name="name">
> ...
> </property>
> </property>
> Stuff that comes after
> More stuff that comes after
Good sample.

> $ sed '/<property name="name">/{:loop /<\/property>.*<\/property>/d; N;
> b loop}' file.xml
[...]
A hypothetical problem with my solution and your solution is in this case:

    Stuff that comes before
    More stuff that comes before
    <property name="name">
    ...
    </property>
    1
    2
    <property name="name">
    </property>
    Stuff that comes after
    More stuff that comes after
    <property name="name">
    ...
    </property>
    </property>
    Stuff that comes after
    More stuff that comes after

Our codes removes the "symmetric" block between "asymmetric" blocks. But
this  type of  things need  attention from  Marcus. Maybe  this isn't  a
problem.

A hypothetical problem with the Marcus solution is in this case:

    Stuff that comes before
    More stuff that comes before
    <property name="name">
    ...
    Stuff that comes after
    More stuff that comes after

Only the  first two lines  of this example  will be printed  because sed
delete  the  address  regexp  '/<property  name="name">*$/'  until  find
'/<\/property>*$/' OR find end of  file. Perhaps this is an undesirable
behavior.

[]'s
--
"Não manejo bem as palavras
Mas manipulo bem as strings."
------------------------------
http://tecnoveneno.blogspot.com