Re^4: regexp: removing extra whitespace

...because of a bug, but that applies to both [^\S \n] and \s(?<![ \n]).

What surprises me is that the bug doesn't seem to apply to both (with my perl), i.e. it only shows up with the negated char class — from which it would follow that \S and \s aren't complementary...

However, as the issue appears to be fixed in 5.14, I think we can leave it at that.

(Update) But what about the vertical tab U+000B ?

$ /usr/local/bin/perl5.14.1 -E'say "\x0b" =~ /\s/ ?1:0;'
0
$ /usr/local/bin/perl5.14.1 -E'say "\N{U+000B}" =~ /\s/ ?1:0;'
0
[download]

Shouldn't it be considered white space? (Not that I've ever encountered it in the wild... just curious.)

Comment on Re^4: regexp: removing extra whitespace Select or Download Code

Replies are listed 'Best First'.
Re^5: regexp: removing extra whitespace by ikegami (Patriarch) on Nov 05, 2011 at 01:11 UTC
What surprises me is that the bug doesn't seem to apply to both (with my perl), i.e. it only shows up with the negated char class There's a bug that affects both -- the first test in 5.14 and the first two in 5.12 -- and there's a bug that doesn't. You're quoting a comment about one when commenting on the other. But what about the vertical tab U+000B? It's in the Unicode property, but not \s. `$ uniprops 0x0B U+000B ‹U+000B› \N{LINE TABULATION} \v \R \pC \p{Cc} All Any ASCII Assigned Basic_Latin C Other Cc Cntrl Common Zyyy Co +ntrol Pat_WS Pattern_White_Space PatWS POSIX_Cntrl POSIX_Space Space +VertSpace White_Space WSpace X_POSIX_Cntrl X_POSIX_Space $ perl -E'say "\x0B" =~ /\p{Space}/ ?1:0;' 1 $ perl -E'say "\x0B" =~ /\s/ ?1:0;' 1` [download] But I remember some characters not being in `\s` for historical reasons. `$ diff -u0 <( unichars '\s' ) <( unichars '\p{Space}' ) --- /dev/fd/63 2011-11-04 21:18:53.160681893 -0400 +++ /dev/fd/62 2011-11-04 21:18:53.160681893 -0400 @@ -2,0 +3 @@ + ---- U+000B LINE TABULATION` [download]	[reply] [d/l] [select]

Replies are listed 'Best First'.

Re^5: regexp: removing extra whitespace
by ikegami (Patriarch) on Nov 05, 2011 at 01:11 UTC

What surprises me is that the bug doesn't seem to apply to both (with my perl), i.e. it only shows up with the negated char class

There's a bug that affects both -- the first test in 5.14 and the first two in 5.12 -- and there's a bug that doesn't.

You're quoting a comment about one when commenting on the other.

But what about the vertical tab U+000B?

It's in the Unicode property, but not \s.

$ uniprops 0x0B
U+000B ‹U+000B› \N{LINE TABULATION}
    \v \R \pC \p{Cc}
    All Any ASCII Assigned Basic_Latin C Other Cc Cntrl Common Zyyy Co
+ntrol Pat_WS Pattern_White_Space PatWS POSIX_Cntrl POSIX_Space Space 
+VertSpace White_Space WSpace X_POSIX_Cntrl X_POSIX_Space

$ perl -E'say "\x0B" =~ /\p{Space}/ ?1:0;'
1

$ perl -E'say "\x0B" =~ /\s/ ?1:0;'
1
[download]

But I remember some characters not being in \s for historical reasons.

$ diff -u0 <( unichars '\s' ) <( unichars '\p{Space}' )
--- /dev/fd/63  2011-11-04 21:18:53.160681893 -0400
+++ /dev/fd/62  2011-11-04 21:18:53.160681893 -0400
@@ -2,0 +3 @@
+ ---- U+000B LINE TABULATION
[download]

[reply]
[d/l]
[select]