Feature #8714: Non-interpolated regular expression literal

• Status: Open
• Assignee: -
• Target version: -

[phluid61 (Matthew Kerwin)]

=begin

I propose a new %string for non-interpolated regexp literals: %R

It is common to see erroneous bug reports around the use of ((%#%)) in regexp literals, for example where (({/[$#]/})) raises a syntax error "unexpected $undefined", and this confuses people. The only solution is to rearrange the regular expression (such as (({/[$#]/}))), which is not always desirable.

An non-interpolated regexp, such as (({%R/[$#]/})), would allow a much simpler resolution.

=== Known Issues

• the capitalisation is the opposite of %Q(interpolated) and %q(uninterpolated)
• %R was also proposed for literal Rationals in #8430, although I believe this has been superseded by the (({1.2r})) syntax

=end

[Eregon (Benoit Daloze)]

Why not Regexp.new('simple quoted string or any literal not interpolating')?

Also,
/[$#]/ # => /[$#]/ and
"[#$]" => "[#$]" are fine on 2.0 and later.

(Although "[#$]".inspect should probably not escape '#' but it is harmless)

[phluid61 (Matthew Kerwin)]

Eregon (Benoit Daloze) wrote:

Why not Regexp.new('simple quoted string or any literal not interpolating')?

That would have to be Regexp.new(%q/.../) to avoid having to escape single quotes and double-escape backslashes in the string literal.. which is pretty obtuse.

Also,
/[$#]/ # => /[$#]/ and
"[#$]" => "[#$]" are fine on 2.0 and later.

(Although "[#$]".inspect should probably not escape '#' but it is harmless)

Off the top of my head, I can't think of how to construct a regexp literal to match a hash character at the end of the string (i.e. /#$/), without first constructing a string.

[Eregon (Benoit Daloze)]

Off the top of my head, I can't think of how to construct a regexp literal to match a hash character at the end of the string (i.e. /#$/), without first constructing a string.

Well you can escape the "#": /#$/ =~ "#" # => 0.
%r{#$} works too.
If you want to match at the end of the String, you should use /#\z/.

But indeed simply /#$/ gives "unterminated regexp meets end of file".
After all $/ is a global variable (the input record separator), so it is only logical it interpolates it.

Also, /regexp/ literal needs escape only for #, \ and / if I am not mistaken,
which is quite restricted compared to what must be escaped in "" or %Q.

[phluid61 (Matthew Kerwin)]

Eregon (Benoit Daloze) wrote:

Off the top of my head, I can't think of how to construct a regexp literal to match a hash character at the end of the string (i.e. /#$/), without first constructing a string.

Well you can escape the "#": /#$/ =~ "#" # => 0.

Of course!

%r{#$} works too.

irb(main):004:0> %r{#$}
SyntaxError: (irb):4: syntax error, unexpected $undefined
%r{#$}
^
from /usr/local/bin/irb:12:in `'
irb(main):005:0> %r{#$}
=> /#$/

If you want to match at the end of the String, you should use /#\z/.

At the end of the line, then. ;)

But indeed simply /#$/ gives "unterminated regexp meets end of file".
After all $/ is a global variable (the input record separator), so it is only logical it interpolates it.

Even if it's not a (valid, defined) global variable, the parser still attempts to interpolate it. For example: /#$]/ (there is no $] in ruby)

Also, /regexp/ literal needs escape only for #, \ and / if I am not mistaken,
which is quite restricted compared to what must be escaped in "" or %Q.

That's only partly true. # only need be escaped when it is followed by $, @ or {. Therein lies the source of a lot of confusion. From what I can see, ruby-doc.org says "Arbitrary Ruby expressions can be embedded into patterns with the #{...} construct." which is very easy to miss, and it's not always clear that "#$x" or /#$x/ are part of the #{...} construct.

I admit that this is a standard part of ruby interpolation, but "#$x#@y" is not commonly encountered in the wild, and is much more likely to occur in a (symbol-rich) regexp than a (typically human readable) string. Thus I propose an option to construct regexps that don't treat # as special.

Note: I'd still expect other backslash-escapes (like \u{...}) to work in uninterpolated regexps, because even uninterpolated regexps should be able to do normal perly things like %R/\u{263a}\n/

[nobu (Nobuyoshi Nakada)]

For other %-literals, upper cases do interpolation, e.g., %Q, %W, %I.
I'm afraid that it would cause confusion by making %R exceptional.

[phluid61 (Matthew Kerwin)]

nobu (Nobuyoshi Nakada) wrote:

For other %-literals, upper cases do interpolation, e.g., %Q, %W, %I.
I'm afraid that it would cause confusion by making %R exceptional.

It is unfortunate that %r and %x buck the trend. I would suggest a different letter, such as %P, but I think that might be just as confusing.

[drbrain (Eric Hodel)]

phluid61 (Matthew Kerwin) wrote:

%r{#$} works too.

irb(main):004:0> %r{#$}
SyntaxError: (irb):4: syntax error, unexpected $undefined
%r{#$}
^
from /usr/local/bin/irb:12:in `'
irb(main):005:0> %r{#$}
=> /#$/

What ruby version are you using? It works for me. Without irb (since sometimes you can't trust it):

$ ruby -ve 'p %r{#$} =~ "foo#"'
ruby 2.0.0p247 (2013-06-27 revision 41674) [x86_64-darwin12.4.0]
3

With irb:

$ irb

%r{#$} =~ "foo#"
=> 3
RUBY_DESCRIPTION
=> "ruby 2.0.0p247 (2013-06-27 revision 41674) [x86_64-darwin12.4.0]"

[phluid61 (Matthew Kerwin)]

drbrain (Eric Hodel) wrote:

What ruby version are you using? It works for me.

Ah, sorry, that was ruby 2.0.0p0 (2013-02-24 revision 39474) [x86_64-linux]

After updating to -p247 it works.

[Eregon (Benoit Daloze)]

phluid61 (Matthew Kerwin) wrote:

That's only partly true. # only need be escaped when it is followed by $, @ or {. Therein lies the source of a lot of confusion. From what I can see, ruby-doc.org says "Arbitrary Ruby expressions can be embedded into patterns with the #{...} construct." which is very easy to miss, and it's not always clear that "#$x" or /#$x/ are part of the #{...} construct.

Yeah, I wish there was only #{}, I would much rather see the removal of #@... and #$... than introducing a new literal, but I doubt it would be accepted for compatibility.