extract functions

README.md

Extract an indented block from a file, starting from a given header line.

Usage: extract.awk HEADER file

e.g.

$ ./extract.awk a t
   a() {

      printf("lololol\n");
   }
$ ./extract.awk b t
   b() {


     call_the_thing();
     for(i = 1; i<a; i++) {
        i--;
     }
   }

Installation

git clone https://gist.github.com/kousu/d6a945cfc6ffa9f93e82a630098acc90
mkdir -p ~/.local/bin
ln d6a945cfc6ffa9f93e82a630098acc90/extract-block ~/.local/bin
chmod +x ~/.local/bin/extract-block

and if you're not on a system where this is pre-configured, also do

echo 'export PATH=~/.local/bin:$PATH` >> ~/.profile # or equivalent

extract-block

#!/usr/bin/env -S awk -f
# extract an indented block from a file
#
# usage: extract-block search-pattern file1 [file2 ...]

BEGIN {
    pattern = ARGV[1]
    delete ARGV[1]
    extracting=0
    current_indentation=""
    target_indentation=""
}

# detect the current line's indentatation level
{
  match($0, "^([[:space:]]*)", G)
  current_indentation=G[1]
}

# find requested header line
extracting==0 && $0 ~ pattern { #     https://www.gnu.org/software/gawk/manual/html_node/Computed-Regexps.html#Computed-Regexps # XXX is this safe? is there a code-injection attack here?
  extracting=1
  target_indentation=current_indentation

  # hack:
  #  the stopping condition would notice that target_indentation == current_indentation and stop immediately, meaning each block only prints one line
  #  circumvent that by just inlining the print step but skipping the stop step after it.
  print; next
}

# print every line with the detected indent level
extracting == 1 {
  #print "Current line |" $0 "|'s indentation level is: |" current_indentation "| which is " length(current_indentation) " long" " and |" target_indentation "| which is " length(target_indentation) " long" #>/dev/stderr
  print
}

# stop on the last line of the block
# - skip blank lines, because most editors cull trailing whitespace, so a fully blank line shouldn't count as being dedented
# - skip if there's *no* indentation; in this case, we should print the entire rest of the file
extracting == 1 && length() > 0 && length(target_indentation) > 0 && length(current_indentation) <= length(target_indentation) {
  # problem: this matches the first line (because length(current_indentation) == length(target_indentation)
  #
  # solution 1: put this line *before* the command that detects the first line
  #   -> problem: then this misses the *last*
  exit
}

t

# test file

   a() {

      printf("lololol\n");
   }


   b() {


     call_the_thing();
     for(i = 1; i<a; i++) {
        i--;
     }
   }

 ofa sdf oisd 

c

t2

allphabetitt 
dsfosup
spoufd

The idea is this would be useful in CI to detect when particular code blocks have changed.

To do this correctly requires a full AST parser. This is only an approximation. But most projects are using some kind of reasonable indentation rules, so it will work for them; and if you do use it to detect changes, it will detect when someone breaks the indentation rules too.