Re: traversing a variable with regex instead of a file

From: James Edward Gray II (james_at_grayproductions.net)
Date: 10/10/03 

Date: Fri, 10 Oct 2003 12:15:52 -0500
To: angie ahl <angie@vertebrate.co.uk>

On Friday, October 10, 2003, at 10:51 AM, angie ahl wrote:

> on 2003-10-10 James Edward Gray II said:
>
>> Keep your replies on the list, so you can get help from all the people
>> smarter than me. ;)
>
> If there are people smarter than you out there I must be an amoeba ;)

There are some very brilliant people hiding here. Stick around and
you'll spot on eventually...

>> Okay, why put this inside an if block. If it doesn't find a match it
>> will fail and do nothing, which is what you want, right? I don't
>> think
>> you need the if.
>
> Good point.

Thanks, I thought so too, but you forgot to take my advice. You still
don't need the if block you have in the code. :D

>> $content =~ s/^([^\n]*)($kw)/substr($1, 0, 2) ne 'qz' ? "$1\n$2\n" :
>> "$1$2"/mge;
>>
>
> Ok. I had to stop to pick myself up off the floor then. WOW.

I love that silly little /e modifier myself. I use it a lot. It
wouldn't surprise me if one of the Regex Gurus could do it even better
though. I'm still reading Mastering Regular Expressions myself. ;)

> # get line breaks to make <br>'s at the end
> $content =~ s/\n/-qbr-/g;
>
> # find markup and add markers so it doesn't get processed by regex,
> # no keyword links to be made inside other tags
> $content =~ s/(\[(img|page|link|mp3)=.*?\])/\nqz$1\n/g;
>
> # find HTML so it doesn't get processed by regex,
> # no keyword links to be made inside valid HTML
> $content =~ s/(<.*?>)/\nqz$1\n/g;

There are some excellent HTML parser modules on the CPAN that could
possibly help you do this kind of thing, just FYI.

> for my $href ( @Keywords ) {
>
> # get each keyword and llok for it in content.
> for $kw ( keys %$href ) {
> if ($content =~ /\b($kw)\b/g) {

The if above is not needed, as near as I can tell, though you should
probably add the \b...\b around the $kw in my expression. There's
certainly no reason to use a global search above, since you're just
checking to see if it's present.

> # do the very clever reg with help from and thanks to
> # james@grayproductions.net
> $content =~ s/^([^\n]*)($kw)/substr($1, 0, 2) ne 'qz' ?
> "$1\nqz[link=\"$href->{$kw}\" title=\"$2\"]\n" : "$1$2"/mge;
> }
> }
> }
>
> # clean up those line breaks and markers;
> $content =~ s/\n(qz)?//g;
>
> # put in <br>'s
> $content =~ s/-qbr-/<br>\n/g;
>
> print $content;
> _________________________
>
> As you can see I've adapted your regex a little to put in the full
> markup around
> the keyword.

Making changes already. You're a natural.

> The regex itself made perfect sense, it was the
>
> "" ? "" : "" bit that I've never seen before. That's really useful.
>
> I assume it means
>
> "if statement" ? "do if true" : "do if false"

You assume right. The Ternary Operator (or Conditional Operator) is a
simple if/else shorthand. It's not too popular, probably for
readibility reasons, but I use it a lot in things like Regexes and
map(). If you want to know more:

perldoc perlop

Then type:

/Conditional Operator

> Please do correct me if I'm wrong. What do you call that? I think I'm
> going it
> be using that quite a bit ;)
>
> do I even need the if false bit in this case?

Yes, and this is important. The way I wrote the match, it will find
ALL $kw occurrences, even the ones on qz lines. So after we learn
we're on a qz line, we need to put things back with our replace, so it
looks like we never messed with it.

>> I used the /e modifier for the replacement, which allows me to use
>> Perl
>> code in there. It's pretty simple. If the line didn't start with a
>> qz, we do a normal replace.
>
> That's going in my BBEdit gold dust code snippets glossary.

A BBEdit user, and by extension a Mac user too, eh? That settles it,
we're going to be great friends you and I. ;)

Unfortunately, the /e trick doesn't work in BBEdit's Search/Replace
dialogs, though it has excellent Grep support.

>> Your right about it being inefficient, of course. It was easier to
>> read than my Regex though, eh? <laughs>
>
> Are you implying that regex isn't easy to read ;)

Me, I love a little line noise in the morning, but that might be why
I'm a Super Geek.

>> The first choice may be slow,
>> but on modern computers they may both work in the blink of an eye.
>> Save worrying about speed for when you need to and try and keep your
>> life as a programmer as easy as possible until then.
>
> Sadly then is now. That's why I joined up to this list today ;)
>
> This code will be run on every single page of a website, in one go. So
> it needs
> to be as efficient as physically possible. The site will only be a few
> hundred
> pages, and not all pages will always be processed. It's a system that
> makes it's
> own links and maintains them, so eveytime a page's keywords change
> this has to
> be done to all pages that contain that keyword.

Fair enough. Just remember that measuring the different between two
pieces of code that both run in under a second is usually pretty
pointless. Start optimizing AFTER you start waiting on executions, not
before.

> I know this is not a task for the beginner, but this is actually
> version 3 of
> the code. my old programming language started to show it's dislike for
> regex.

You'll do fine and you know where to get help if you need it.

>>> If you have any suggestions I would be most grateful to hear them.
>>
>> Those are my best shots. Hope they help.
>
> They did, thank you so much.

Happy to help. Welcome to the list.

James

Relevant Pages