http://www.perlmonks.org?node_id=11141603

ibm1620 has asked for the wisdom of the Perl Monks concerning the following question:

I want to split up an ASCII text document into words (as recognized by /b{wb}), strings of the non-word characters between words, and strings of newlines.

The following code almost works, but instead of treating the newlines as separate tokens, it leaves them appended to the preceding word

#!/usr/bin/env perl use strict; use warnings; my $book = do {local $/; <DATA>}; # slurp the book # Split book into words (delimited by \b{wb}), sequences of newlines, # and sequences of anything else. while ($book =~ /( ( \W+ ) | ( \b{wb}.+?\b{wb} ) | ( \n+ ) ) /xg) { show($1); } print "\n"; # show(): make spaces and newlines visible sub show { my $str = shift; $str =~ tr/\n/$/; $str =~ tr/ /_/; print "{$str}\n"; } __DATA__ --First paragraph-- Second one's followed by only one newline. "Hello," she said, "How's t +ricks?" Third paragraph doesn't end with any punctuation ... and the splitting + works 4th one is separated by two newlines. The End.

The output is:

{--} {First} {_} {paragraph} {--$} <- The newline ('$') should be separate group {Second} {_} {one's} {_} {followed} {_} {by} {_} {only} {_} {one} {_} {newline} {._"} {Hello} {,"_} {she} {_} {said} {,_"} {How's} {_} {tricks} {?"$$} <- the two newlines should be a separate group {Third} {_} {paragraph} {_} {doesn't} {_} {end} {_} {with} {_} {any} {_} {punctuation} {_..._} {and} {_} {the} {_} {splitting} {_} {works} <- Correctly {$$} <- split {4th} {_} {one} {_} {is} {_} {separated} {_} {by} {_} {two} {_} {newlines} {.$$_________} <- should be three separate groups {The} {_} {End} {.$}

I'm wondering what I'm doing wrong, and whether there's a better solution. (Would the split function be preferable?)

Replies are listed 'Best First'.
Re: Splitting multiline string into words, the stuff between words, and newlines
by LanX (Saint) on Feb 24, 2022 at 00:55 UTC
    > I'm wondering what I'm doing wrong,

    \W is a negation of \w but is still including \n

    I negated both with [^\w\n]

    That's the result you wanted? Can't really comment on the rest, looks weird to me.

    Cheers Rolf
    (addicted to the Perl Programming Language :)
    Wikisyntax for the Monastery

      Yes, that's exactly what I wanted. Thank you.

      (I'm just playing with different ways of building Markov chains, a la dissociated-press.)

Re: Splitting multiline string into words, the stuff between words, and newlines
by salva (Canon) on Feb 24, 2022 at 09:06 UTC
    You can also use split for that in order to not require a regular expression for matching non words:
    my @fragments = grep length, split /(\b{wb}.+?\b{wb}|\n+)/, $book;
    So, you get words, sequences of new lines and then everything else.
      This looks to me like it should work, but it splits the strings of non-words into separate characters!

      "For example ...\n" -> {For}{_}{example}{_}{.}{.}{.}{$}
        That is because \b{wb} matches between those signs.

        This seems to solve the issue:

        my @fragments = grep length, split /(\b{wb}\w.*?\b{wb}|\n+)/, $book;

        But my knowledge of Unicode and the \b{wb} semantics is rather limited so that may have other issues.