/\b/ is equivalent to /(?<=\w)(?!\w)|(?<!\w)(?=\w)/. Feel free to replace \w with a character class.

It would be tedious to have to write \\$keyword([^a-zA-Z]) and then have to substitute back $1 (because I do not want it eaten).

Don't eat it if you don't want add it back. Equivalent without eating:

\\$keyword(?=[^a-zA-Z])
[download]

But you surely meant

\\$keyword(?![a-zA-Z])
[download]

In general, it's easier to extract the keyword, then check if it's the one you want.

\\([a-zA-Z]+)
[download]

In general, it's easier to extract the keyword, then check if it's the one you want.

I agree wholeheartedly. Since the LaTeX name constraint is exact and well-understood (the characters 'a' through 'z' and the characters 'A' through 'Z'), you simply need to match just those characters. Explicitly matching the right-hand boundary isn't necessary.

thanks. this is what I needed to learn. I did not know the extended regex expressions in the camel book (i.e., (?...) sequences), chapter 5, table 5.6. is there a meaningful difference between (?!a-z) and (?=^a-z)? is the former recommended? /iaw

Compare

'ab' =~ /a(?!a)/
'a' =~ /a(?!a)/
[download]

and

'ab' =~ /a(?=[^a])/
'a' =~ /a(?=[^a])/
[download]

is there a meaningful difference between (?![a-z]) and (?=[^a-z])? is the former recommended?

Yes, they're different regular expression patterns that match different things. (?![a-z]) asserts "not followed by any of the characters from 'a' through 'z', which includes not being followed by any character." (?=[^a-z]) asserts "followed by a single character that is not any of the characters from 'a' through 'z'." The former is a negative assertion; the latter is a positive assertion.

In your case, (?![a-z]) is what you would want to use.

[PerlMonks posting tip: Enclose Perl code in <code></code> tags, even code within paragraphs.]

UPDATE: Removed color.

is it possible to define the characters that '\b' matches?

I believe so, though I've never tried to do it myself. See Creating Custom RE Engines in perlre.

Unless I'm understanding your intentions wrong, that's the purpose of \b.

  Example: m/\bChi_2\b/

I don't think the underscore will cause you problems within the defined \b \b but I'm sure I'll be corrected shortly if I'm wrong. :)

UPDATE: 1.20.2011 1:30PM
• Through help in the chatterbox and "Mastering Regular Expressions" pg. 89, I learned that \w has included _ since Perl 2. So /\bChi_2\b/ will not match.


"...the adversities born of well-placed thoughts should be considered mercies rather than misfortunes." — Don Quixote

I believe he's saying that "_2" isn't part of the macro, so $keyword = 'Chi'; '...\\Chi_2...' =~ /\\$keyword\b/ should match.

Ugh! — Apologies to the OP. I misread that.

"...the adversities born of well-placed thoughts should be considered mercies rather than misfortunes." — Don Quixote

I'm not sure why you added the update. I don't see what it adds, and it's not true. /\bChi_2\b/ will match plenty of strings.

'Chi_2'   =~ /\bChi_2\b/  # Match
'!Chi_2!' =~ /\bChi_2\b/  # Match
[download]

Maybe you had a specific string in mind, but I don't see how this relates to the OP. He would not use Chi_2 in the regex pattern.

In a world where an identifier matches /^\w+\z/, you might do something like

($_ = '\\Chi+3'  ) =~ s/\\Ch\b/$ch/g;      # Won't replace
($_ = '\\Chi+3'  ) =~ s/\\Chi\b/$chi/g;    # Will replace
($_ = '\\Chi_2+3') =~ s/\\Chi\b/$chi/g;    # Won't replace
[download]

But what if identifiers match /^[a-zA-Z]\z/? You'd want the following behaviour:

($_ = '\\Chi+3'  ) =~ s/\\Ch???/$ch/g;    # Won't replace
($_ = '\\Chi+3'  ) =~ s/\\Chi???/$chi/g;  # Will replace
($_ = '\\Chi_2+3') =~ s/\\Chi???/$chi/g;  # Will replace
[download]

That's the OP's question.

As I've already mentioned, I recommend extracting the identifier, then checking if it's one of interest. This can be as simple as the following:

/\\([a-zA-Z]+)/ exists($vars{$1}) ? $vars{$1} : "\\$1" /eg
[download]

The technique scales well, and it avoids the problem of matching something you've previously replaced.

   

"Maybe you had a specific string in mind, but I don't see how this relates to the OP. He would not use Chi_2 in the regex pattern.

I did have a very similar string in mind.    

In a world where an identifier matches /^\w+\z/, you might do something like"
   ($_ = '\\Chi_2+3') =~ s/\\Chi\b/$chi/g; # Won't replace

I understand my update isn't contributing to the OP's original question. I'm not trying to distract from his post or the thread, simply attempting to correct what I said regarding the underscore having no effect on the RegEx's success (again the one I had in mind).

In my original reply I was referring to matching 'Chi' within 'Chi_2' using \b. I previously said that I didn't think the underscore would be a problem. However after some help in the CB from erix and Tanktalus it was shown that an underscore would interfere with this particular match:

 say (("Chi_2" =~ /\bChi\b/) ? "match" : "no match");
 returns: "no match"

  * Thanks again to Tantalus for this control structure.

Again, apologies for any confusion caused.



"...the adversities born of well-placed thoughts should be considered mercies rather than misfortunes." — Don Quixote