[sed] "Advanced" sed question(s)

G00fy · 03-19-2006, 04:31 PM

Hi,

Basically what I try to do here is convert a slice-file (www.zeroc.com) to a C++ header file so DoxyGen can parse it without a glitch

.
So I don't care about layouting in sed... As it will be interpreted by a machine anyway

.

I have 3 questions for sed...

Let me first show the input file:

Code:

#ifndef TEST1
#define TEST1


module main
{
  local interface test1
        extends   ext1,
                  ext2
  {
                bool function( );
    idempotent  bool function( );

    nonmutating bool function( );
    nonmutating bool function( ) throws RangeError;
    nonmutating bool function( ) throws RangeError, NormalError;
    nonmutating bool function( ) throws RangeError,
                                        NormalError;
    nonmutating bool function( ) throws RangeError
                                        , NormalError;
  };
};


#endif // TEST1

Now what it has to look like:

Code:

#ifndef TEST1
#define TEST1


namespace main
{
  class test1
        : public ext1, public
                  ext2
  {
                bool function( );
    bool function( );

    bool function( ) const;
    bool function( ) const throw(RangeError);
    bool function( ) const throw(RangeError, NormalError);
    bool function( ) const throw (RangeError,
                                        NormalError);
    bool function( ) const throw (RangeError
                                        , NormalError);
  };
};


#endif // TEST1

I am trying to do this with 'sed'... Now I managed to get this script:

Code:

s/local//g
s/interface/class/g
s/extends/: public/g
s/module/namespace/g
s/sequence/vector/g
s/dictionary/map/g
/class/,/{/s/{/{public:/g
s/idempotent//g
s/nonmutating\(.*\);/\1 const;/g

now the problems I have are those:

throws ...; -> throw(...); with support for multiple errors on multiple lines
extends ..., ..., ... -> : public ..., public ..., public ...
nonmutating ... () throws... -> ... () const throw...;

I have been looking around for something called 'N', but that seems like only running once, and what I need is something that even runs on multiple lines...

For (1) I need everything from 'throws' -> ';'. And then change the keyword 'throws' -> 'throw(', and the keyword ';' -> ');'. That would do this trick, but I don't know how to tell in sed.

For (2) I need to convert the extends <class>[, <class>] into : public <class>[, public <class]...

(3) is rather simple if I know how to do (1) in fact... So never mind about that one

Could someone help me with this?

Thanks!

Frodak · 03-19-2006, 07:19 PM

you need to use multiple line addressing and stuff... this is kind of clunky, but I'm not a sed guru

Code:

#!/usr/bin/sed -f

s/module/namespace/

#parse local interface
/local interface/ , /{/    {
    /local interface/ {
        #change local interface to class
        s/local interface\(.*\)/class \1:/
        #branch to the end of script - do nothing else for local interface
        b
    }
#delete extends
    s/extends//
#add public before every keyword, but not {
    s/\([^ {]\+\)/public \1/g

}

s/nonmutating\(.*)\)/\1 const/

/throws.*;/ {
    s/throws/throw(/
    s/;/);/
    #print the line
    p
    #delete pattern space, start next cycle
    #don't process multi-line throws
    d
} 

/throws/,/;/ {
    s/throws/throw(/
    s/;/);/
}

$ cat temp.lst 
#ifndef TEST1
#define TEST1


module main
{
  local interface test1
        extends   ext1,
                  ext2
  {
                bool function( );
    idempotent  bool function( );

    nonmutating bool function( );
    nonmutating bool function( ) throws RangeError;
    nonmutating bool function( ) throws RangeError, NormalError;
    nonmutating bool function( ) throws RangeError,
                                        NormalError;
    nonmutating bool function( ) throws RangeError
                                        , NormalError;
  };
};


#endif // TEST1

e728075@NM75DTJYT4M81 ~
$ ./parse_file temp.lst 
#ifndef TEST1
#define TEST1


namespace main
{
  class  test1:
           public ext1,
                  public ext2
  {
                bool function( );
    idempotent  bool function( );

     bool function( ) const;
     bool function( ) const throw( RangeError);
     bool function( ) const throw( RangeError, NormalError);
     bool function( ) const throw( RangeError,
                                        NormalError);
     bool function( ) const throw( RangeError
                                        , NormalError);
  };
};


#endif // TEST1

Frodak

G00fy · 03-20-2006, 12:34 AM

Thanks a lot! To me you're a Guru anyway

.

What I came up with so far (after your great help):

Code:

#!/usr/bin/sed -f

# Skip C-style commenting
/\/\*/ , /\*\//b
/\/\/\(.*\)/b

s/module/namespace/
s/local//
s/idempotent//
s/dictionary/map/
s/sequence/vector/

#parse local interface
/interface/ , /{/    {
    /interface/ {
        #change interface to class
        s/interface\(.*\)/class \1/
    }
#change extends
    s/extends/: public/
#add public before every keyword, but not {
    /{/!s/,\(.*\)/, public \1/g
    s/{/{public:/
}

s/nonmutating\(.*)\)/\1 const/

/throws/,/;/ {
    s/throws/throw(/
    s/;/);/
}

This changes this file:

Code:

#ifndef TEST1
#define TEST1

/**
  module
  interface
  nonmutating
*/

// local interface extends
module main
{
  interface test2 extends ext2
  {
  };
  
  local interface test1
        extends   ext1,
                  ext2
  {
                bool function( );
    idempotent  bool function( );

    nonmutating bool function( );
    nonmutating bool function( ) throws RangeError;
    nonmutating bool function( ) throws RangeError, NormalError;
    nonmutating bool function( ) throws RangeError,
                                        NormalError;
    nonmutating bool function( ) throws RangeError
                                        , NormalError;
  };
};


#endif // TEST1

Into this output:

Code:

#ifndef TEST1
#define TEST1

/**
  module
  interface
  nonmutating
*/

// local interface extends
namespace main
{
  class  test2 : public ext2
  {public:
  };

   class  test1
        : public   ext1, public
                  ext2
  {public:
                bool function( );
      bool function( );

     bool function( ) const;
     bool function( ) const throw( RangeError);
     bool function( ) const throw( RangeError, NormalError);
     bool function( ) const throw( RangeError,
                                        NormalError);
     bool function( ) const throw( RangeError
                                        , NormalError);
  };
};


#endif // TEST1