Behaviour change of String#gsub(pattern) {|m| ... } for ruby 1.9/ruby2?

ruby-talk
1

String#gsub(pattern) {|m| … }

It really would be nice to get match data in ‘m’, but this would surely
break a lot of scripts. How about String#regsub, #regsub!, #resub,
#resub! ? What do others think?

···

–
dave

2

Hi,

···

In message “behaviour change of String#gsub(pattern) {|m| … } for ruby 1.9/ruby2?” on 03/11/20, David Garamond lists@zara.6.isreserved.com writes:

String#gsub(pattern) {|m| … }

It really would be nice to get match data in ‘m’, but this would surely
break a lot of scripts. How about String#regsub, #regsub!, #resub,
#resub! ? What do others think?

Sounds nice. The only reason for the current behavior is that sub
predates MatchData. But we have to define migration path.

						matz.
3

Is there an easy way for String#gsub (et al.) to check the arity of
the provided block?

String#gsub(pattern) { |m, md| … }
yields match string and matchdata

String#gsub(pattern) { |m| … }
yields match string only

That would break the least amount of code.

-austin

···

On Thu, 20 Nov 2003 20:57:58 +0900, Yukihiro Matsumoto wrote:

In message "behaviour change of String#gsub(pattern) {|m| … }

String#gsub(pattern) {|m| … }

It really would be nice to get match data in ‘m’, but this would
surely break a lot of scripts. How about String#regsub,
#regsub!, #resub, #resub! ? What do others think?
Sounds nice. The only reason for the current behavior is that sub
predates MatchData. But we have to define migration path.

–
austin ziegler * austin@halostatue.ca * Toronto, ON, Canada
software designer * pragmatic programmer * 2003.11.20
* 10.18.43

4

Yukihiro Matsumoto wrote:

Hi,

Moin!

String#gsub(pattern) {|m| … }

It really would be nice to get match data in ‘m’, but this would surely
break a lot of scripts. How about String#regsub, #regsub!, #resub,
#resub! ? What do others think?

Sounds nice. The only reason for the current behavior is that sub
predates MatchData. But we have to define migration path.

I like this, because matz told me I shouldn’t be using $1 in the
gsub-block and there’s no way to not do so right now. :slight_smile:

Add a MatchData#to_str and it should work for most cases. We could give
a warning for cases where scripts would still be broken by this chance.

Regards,
Florian Gross

5

Is there an easy way for String#gsub (et al.) to check the arity of
the provided block?

svg% cat b.rb
#!/usr/bin/ruby
def a(&block)
   p block.arity
end

a {|x| }
a {|x, y| }
svg%

svg% b.rb
1
2
svg%

Guy Decoux

6

Hi –

Yukihiro Matsumoto wrote:

Hi,

Moin!

String#gsub(pattern) {|m| … }

It really would be nice to get match data in ‘m’, but this would surely
break a lot of scripts. How about String#regsub, #regsub!, #resub,
#resub! ? What do others think?

Sounds nice. The only reason for the current behavior is that sub
predates MatchData. But we have to define migration path.

I like this, because matz told me I shouldn’t be using $1 in the
gsub-block and there’s no way to not do so right now. :slight_smile:

I’ve also just noticed this:

irb(main):001:0> “abc”.gsub(/((x?)abc)/) {|n,m| p n, m}
“abc”
nil

as opposed to:

irb(main):002:0> /((x?)abc)/.match(“abc”).captures
=> [“abc”, “”]

which makes it hard to do, say, downcasing of a capture in #gsub,
without checking it for nil-ness or, as you say, using $1 and friends.

David

···

On Fri, 21 Nov 2003, Florian Gross wrote:

–
David A. Black
dblack@wobblini.net

7

Lightly tested, seems to work:

class String
alias _sub sub
alias _sub! sub!

def sub(*a, &b)
  b ? _sub(*a) {|s| b.arity == 2 ? b[s, $~] : b[s]} : _sub(*a)
end

def sub!(*a, &b)
  s = sub(*a, &b)
  s == self ? nil : replace(s)
end

end

Not portable, fragile, sometimes works:

module Kernel
alias _sub sub
alias _sub! sub!

def sub(*a, &b)
  $a, $b, $c = a, b, ""

  set_trace_func(proc {|*x|
    if x[0] == 'return'
      eval("$c.replace($_.sub(*$a, &$b))", x[4])
      set_trace_func(nil)
    end
  })

  $c
end

def sub!(*a, &b)
  $a, $b = a, b
  r, w = IO.pipe

  fork or return set_trace_func(proc {|*x|
    w.puts(eval("$_.sub!(*$a, &$b).inspect", x[4]))
    exit!
  })

  Process.wait
  $c = eval(r.readline) or return

  set_trace_func(proc {|*x|
    if x[0] == 'return'
      eval("$_.replace($c)", x[4])
      set_trace_func(nil)
    end
  })

  $c
end

end

···

Austin Ziegler austin@halostatue.ca wrote:

String#gsub(pattern) { |m, md| … }
yields match string and matchdata

String#gsub(pattern) { |m| … }
yields match string only

That would break the least amount of code.

8

I had actually done what you did, Guy, but I don’t know the C side of Ruby,
so I don’t know how easy it is.

I think that this is probably a good stab at a solution, though.

-austin

···

On Fri, 21 Nov 2003 00:31:05 +0900, ts wrote:

Is there an easy way for String#gsub (et al.) to check the arity of A>
the provided block?

–
austin ziegler * austin@halostatue.ca * Toronto, ON, Canada
software designer * pragmatic programmer * 2003.11.20
* 11.31.12