Feature #6801

String#~ for a here document

Added by kyo endo over 1 year ago. Updated over 1 year ago.

[ruby-core:46799]
Status:Assigned
Priority:Normal
Assignee:Yukihiro Matsumoto
Category:-
Target version:Next Major

Description

=begin
Let me propose a new method (({String#~})) for a here document.

次のような実装の(({String#~}))を提案します。

class String
  def ~
    margin = scan(/^ +/).map(&:size).min
    gsub(/^ {#{margin}}/, '')
  end
end

This is for removing leading margins of a here document.

これはヒアドキュメントにおける先頭マージンを除去するためのものです。

class ATool
  def self.help
    lines = ~<<-EOS
              Instruction of `#{self}`

    `#{self}` is one of a great tool in the world.
       This helps you a lot on your daily work.
      Your life will be changed with `#{self}`!!
           Everyone knows about `#{self}`.
        So, You can ask them to learn `#{self}`

                Just Use `#{self}`

                   from Today!
    EOS
    lines
  end
end

puts ATool.help

# >>           Instruction of `ATool`
# >> 
# >> `ATool` is one of a great tool in the world.
# >>    This helps you a lot on your daily work.
# >>   Your life will be changed with `ATool`!!
# >>        Everyone knows about `ATool`.
# >>     So, You can ask them to learn `ATool`
# >> 
# >>             Just Use `ATool`
# >> 
# >>                from Today!

you can put a tilde sign just before the "<<" characters to call String#~, just like putting a minus sign after "<<" to indent the terminator. This is achieved with uniqueness of tilde sign method, which takes the receiver object on the right-hand side.

If you define String#unindent for this purpose, you put .unindent after EOS. I think its less elegant than a tilde.

"<<"文字の直前にチルダを置くことで、String#~を呼び出すことができます。これはチルダ記号のメソッドが、そのレシーバオブジェクトを右側に取るという特徴によって実現されます。仮にこの目的のためにString#unindentを定義した場合は、EOSの後に.unindentを書くことになりますが、これはチルダに比べると美しくありません。

I understand that this is not general method for string but just for here document. However, I think using a tilde as a method is very restricted because it can't take parameters and is difficult to read the meanings or behavior from the sign. From this, there are few tilde methods within built-in classes of ruby inspite of its uniqness(only for Fixnum, Bignum and Regexp).

so, I think using tilde for the above purpose is the one of the few good chances.

Thank you for your consideration.

このメソッドがStringのための汎用的なものでなく、ヒアドキュメント専用であるということに問題がある点は理解しています。しかし、チルダをメソッドとして使える機会は、それが引数を取れないことと、その記号からはその意味や動作を察することは難しいことから、非常に限られていると想像します。故に、そのユニークさにも拘らず、Rubyの組み込みクラスでの使用例は少しです(Fixnum, BignumおよびRegexpでだけ)。

従って、上記目的でチルダを使うことは、数少ない良い機会であると考えます。

以上、ご検討のほどよろしくお願い致します。

This is based on the following my Japanese blog post.

((URL:http://melborne.github.com/2012/04/27/ruby-heredoc-without-leading-whitespace/))
=end

History

#1 Updated by Nobuyoshi Nakada over 1 year ago

  • Description updated (diff)

#2 Updated by Benoit Daloze over 1 year ago

I think the sequence ~<<-EOS becomes very cryptic, and this should be supported in the parser (or at least be integrated into the language). I actually prefer <<-EOS.undent as the intention is clearer.

Maybe something like <<+EOS, which would remove the common indentation?
In your example, it seems fine to have to manually undent the text,
but I think the most common use-case is declaring an heredoc in an indented code.
It does not look nice to have to lose all indentation at that place.
It seems I have similar ideas to what was said in ruby-core:39851.

#3 Updated by kyo endo over 1 year ago

Eregon (Benoit Daloze) wrote:

I think the sequence ~<<-EOS becomes very cryptic, and this should be supported in the parser (or at least be integrated into the language). I actually prefer <<-EOS.undent as the intention is clearer.

Heredoc sytax is already cryptic. so programmers looks it carefully and find the difference between <<-EOS and ~<<-EOS. Yes, this might be supported in the parser. But if so, it will be a big feature or big issue and might be difficult to find the exit. I don't know. On the other hand, I don't prefer Ruby provide <<-EOS.undent style method for this purpose. I prefer it looks a special syntax same as - or <<.

Maybe something like <<+EOS, which would remove the common indentation?
In your example, it seems fine to have to manually undent the text,
but I think the most common use-case is declaring an heredoc in an indented code.

Sorry, I don't understand your point. In the example, the target string is shifted to the left for 4 whitespaces with String#~.

#4 Updated by Benoit Daloze over 1 year ago

merborne (kyo endo) wrote:

Maybe something like <<+EOS, which would remove the common indentation?
In your example, it seems fine to have to manually undent the text,
but I think the most common use-case is declaring an heredoc in an indented code.

Sorry, I don't understand your point. In the example, the target string is shifted to the left for 4 whitespaces with String#~.

I meant your example is probably not the most typical, with its "centered" nature. And I'm fine with having to manually remove the spaces or flush it to the left for such a usage. I would be less fine, with, for example:

module MyApp
module Commands
class Show
def error_message(err)
<<+EOS
Error: #{err.class}
#{err.message}
EOS
end
end
end
end
(found in )

Sorry, I did not remember http://blade.nagaokaut.ac.jp/cgi-bin/scat.rb/ruby/ruby-core/39959.
You might be right String#~ (or String#undent) to have more chances to ever end.

#5 Updated by kyo endo over 1 year ago

Eregon (Benoit Daloze) wrote:

merborne (kyo endo) wrote:

Maybe something like <<+EOS, which would remove the common indentation?
In your example, it seems fine to have to manually undent the text,
but I think the most common use-case is declaring an heredoc in an indented code.

Sorry, I don't understand your point. In the example, the target string is shifted to the left for 4 whitespaces with String#~.

I meant your example is probably not the most typical, with its "centered" nature. And I'm fine with having to manually remove the spaces or flush it to the left for such a usage. I would be less fine, with, for example:

module MyApp
module Commands
class Show
def error_message(err)
<<+EOS
Error: #{err.class}
#{err.message}
EOS
end
end
end
end
(found in )

Sorry, I did not remember http://blade.nagaokaut.ac.jp/cgi-bin/scat.rb/ruby/ruby-core/39959.
You might be right String#~ (or String#undent) to have more chances to ever end.

ok. thank you.

#6 Updated by Thomas Sawyer over 1 year ago

=begin
Alas %L was rejected :(

str = %L|This
| is
| margin
| controlled.

=end

#7 Updated by Yusuke Endoh over 1 year ago

  • Status changed from Open to Assigned
  • Assignee set to Yukihiro Matsumoto
  • Target version set to Next Major

#8 Updated by Thomas Sawyer over 1 year ago

=begin
This is really an area in which Ruby is sorely missing some needed String methods. Some of the most commonly used methods in Ruby Facets deal with indention, mainly:

# Preserves relative tabbing such that the line with the least amount
# of white space ends up with n number of spaces before non-space and
# all other lines move relative to it.

def tab(n)
str = gsub("\t", " ") # option?

min = []
str.each_line do |line|
  next if line.strip.empty?
  min << line.index(/\S/)
end
min = min.min

str.indent(n - min)

end

# Preserves relative tabbing, such that the first non-empty line ends up with
# n number of spaces before non-space, and all subsequent lines move relative
# to the first.
#
# CREDIT: Gavin Sinclair

def tabto(n)
if self =~ / *\S/
indent(n - $1.length)
else
self
end
end

# Indent left or right by n spaces, or n number of c string.
#
# CREDIT: Gavin Sinclair, Trans, Tyler Rick

def indent(n, c=' ')
if n >= 0
gsub(//, c * n)
else
gsub(/#{Regexp.escape(c)}{0,#{-n}}/, "")
end
end

# Equivalent to String#indent, but modifies the receiver in place.

def indent!(n, c=' ')
replace(indent(n,c))
end

# Remove excessive indentation. Useful for multi-line strings embeded in
# already indented code.
#
# <<-END.unindent
# ohaie
# wurld
# END
#
# Outputs ...
#
# ohaie
# wurld
#
# Instead of ...
#
# ohaie
# wurld
#
# CREDIT: Noah Gibbs, mynyml

def unindent(size=nil)
if size
indent(-size)
else
char = ' '
self.scan(/[\ \t]*\S/) do |m|
if size.nil? || m.size < size
size = m.size
char = m[0...-1]
end
end
size -= 1
indent(-size, char)
end
end

# Equivalent to String#unindent, but modifies the receiver in place.
#
# CREDIT: mynyml

def unindent!
self.replace(self.unindent)
end

While I understand the intention of using (({~})) so as to make it part of HERE document syntax, I don't think it outweighs the advantages of having more flexible methods such as these.

=end

Also available in: Atom PDF