如何删除大括号内的代码块？

Question

I have a file that has some code block:

dimensions{
   dim1{
      //foo
   }
   dim2{
      //bar
   }
}

How can I delete the whole 'dimensions' block including everything inside? Taking into consideration that I don't know how many blocks there is inside, but I'm sure that there us same number of opening and closing curly brackets? Thanks in advance.

Answer 1

Using vim it's fairly simple to match any line that contains dimensions using :global :

:[range]g[lobal]/{pattern}/[cmd]

eg:

:g/dimensions/print

Where the command print will print all matched lines:

Changing the command to normal! one can execute normal mode commands on the matched lines, eg d elete, or y ank.

:g/dimensions/normal! ...

The command d deletes, but needs a motion following it, eg: $ to delete until the newline character, or in our case we can use % which finds the matching pair of character, { matches with } , ( matches with ) etc.

:g/dimensions/normal! d%

One can invoke vim from the command line and request it to execute commands with -c :

vim -c 'g/dimensions/normal! d%' [inputfile]...

Remember to save and close to avoid vim from staying open:

-c 'wq'

As vim can be configured to ones needs via the vimrc file and from the environment one can disable this with:

-u NONE

All together:

vim -u NONE -c 'g/dimensions/normal! d%' -c 'wq' [inputfile]...

Answer 2

If you can use vim instead of sed, you can take profit of the non-interactive mode:

vim input.txt -u NONE -c "call search('\\<dimensions{', 'w')" -c "norm maf{%d'a" -c "wq"

Be aware that this version will edit the file in place.

This method is nice because it will take into account the recursive nesting of the curly bracket blocks.

However this has some issues:

• If you have many files to process, I guess this will be really slower than using sed;
• The norm command is not easy to understand if you don't know vim

Answer 3

Using perl and a recursive regular expression:

$ cat input.txt
some text before.
dimensions{
   dim1{
      //foo
   }
   dim2{
      //bar
   }
}
some text after.
$ perl -0777 -pe 's/dimensions(\{(?:(?>[^{}]+)|(?1))*\})\n//' input.txt
some text before.
some text after.