From: mathew <meta@...>
Date: 2005-07-29T07:46:01+09:00
Subject: Re: Sed -> Ruby : .. and ...

Keith Fahlgren wrote:
> ruby:
> #!/usr/bin/env ruby
> 
> files = ARGV
> 
> files.each do |arg|
>   f = File.open(arg)
>   puts "\nOpening file #{f}"
>   working_file = f.read
> 
>   working_file.gsub!(/MATCH/,'GLOBAL REPLACEMENT')
> 
>   puts "\nDoing ACTION in #{arg}"
>   f = File.new(arg, "w")
>   puts "\nWriting #{f} now"
>   f.print(working_file)
>   f.close
> end

One problem with this code is that if you break out of it, you might 
find that it deleted your data.

A better approach is to rename the data file to a temporary file, write 
the data, then unlink the temporary file.

The best approach is to write the new data to temporary file #1, rename 
the input file to a different temporary file #2 and then immediately 
rename #1 to the input filename, and finally delete temporary file #2. 
That minimizes the window in which the state on disk is not what you 
want it to be. It ensures that the worst possible case is that you have 
a temporary file left behind in the data directory, and the input file 
contains either the data from before processing, or the data after 
processing.

Unfortunately, implementing the best approach is non-trivial, because 
you have to worry about rename not working across filesystem 
boundaries... But you should at least make sure you don't delete the 
user's data. :-)

> Here's the one I have problems with (we had a working Perl equivalent 
> but are trying to abandon Perl). 
> 
> sed:
> /BEGIN RANGE/,/END RANGE/{
>         s/MATCH/REPLACEMENT/g
> }

working_file.gsub!(/(?=BEGIN RANGE)(.*?)(?=END RANGE)/m) {||
   $1.gsub(/MATCH/m, 'REPLACEMENT')
}

(?= ) is a zero-width assertion; the regexp engine matches the BEGIN 
RANGE and END RANGE, but then forgets about them when it comes to 
removing and replacing, so they're still left there in the final string.

(.*?) is a non-greedy match, which ensures that we get the shortest 
possible match between a BEGIN RANGE and an END RANGE; otherwise, if you had

    BEGIN RANGE
     MATCH
    END RANGE
     MATCH
    BEGIN RANGE
     MATCH
    END RANGE

all three MATCHes would be replaced.

The block just does a normal global search and replace on the character 
sequence.

Don't forget to comment what those three lines do, for the benefit of 
the person who has to maintain the code...


mathew
-- 
<URL:http://www.pobox.com/~meta/>
          WE HAVE TACOS