Re: [PATCH] git-remote-mediawiki: Fix a bug in a regexp

From: Jeff King <peff@peff.net>
Date: Jun 9, 2013, 02:57 UTC
Message-ID: <20130609025708.GB30393@sigill.intra.peff.net>
In-Reply-To: <vpqmwr0v45b.fsf@anie.imag.fr>

On Sat, Jun 08, 2013 at 08:38:56PM +0200, Matthieu Moy wrote:

Show 9 quoted lines

> Célestin Matte <celestin.matte@ensimag.fr> writes:
> 
> > In Perl, '\n' is not a newline, but instead a literal backslash followed by an
> > "n". As the output of "rev-list --first-parent" is line-oriented, what we want
> > here is a newline.
> 
> This is right, but the code actually worked the way it was. I'm not
> sure, but my understanding is that '\n' is the string "backslash
> followed by n", but interpreted as a regexp, it is a newline.

Yes, the relevant doc (from "perldoc -f split") is:

  The pattern "/PATTERN/" may be replaced with an expression to specify
  patterns that vary at runtime.  (To do runtime compilation only once,
  use "/$variable/o".)

So it is treating "\n" as an expression and compiling the regex each time through (though I think modern perl may be smart enough to realize it is a constant expression and compile the regex only once). You would get the same behavior with this:

  split $arg, $data;

if $arg contained '\n'. Of course, you _also_ get the same thing if you use a literal newline (either "\n" or if $arg contained a literal newline), because they function the same in a regex. In other words, it does not matter which you use because perl's interpolation of "\n" and the regex expansion of "\n" are identical: t hey both mean a newline.

A more subtle example that shows what is going on is this:

  split '.', $data;

If you feed that "foo.bar.baz", it does not split it into three words; each character is a delimiter, because the dot is compiled to a regex.

> The new code looks better than the old one, but the log message may be
> improved.

Agreed. I think the best explanation is something like:

  Perl's split function takes a regex pattern argument. You can also
  feed it an expression, which is then compiled into a regex at runtime.
  It therefore works to pass your pattern via single quotes, but it is
  much less obvious to a reader that the argument is meant to be a
  regex, not a static string. Using the traditional slash-delimiters
  makes this easier to read.

-Peff