Feature #15231: Remove `Object#=~`

• Status: Closed
• Assignee: -
• Target version: -

[mame (Yusuke Endoh)]

Object#=~ receives (and just discards) an argument, and always returns nil. What purpose is this method for?

The following behavior that Object#=~ caused was confusing to me.

["foo"] =~ /foo/ #=> nil

More precisely: the actual example that I encountered was to parse coverage data from output of coverage measurement tool by using Open3.capture2:

out = Open3.capture2("gcov", ...) # BUG: `out, =` was intended
if out =~ /lines\.*: *(\d+\.\d+)%/
  ...
end

Obviously, I forgot a comma to receive the return value of Open3.capture2. The method returns a two-element array, and out =~ calls Object#=~, which hid the bug. (Worse, I took several tens of minutes to debug it because I first thought that this is a bug of regexp, and spent tweaking the regexp.)

I guess Object#=~ was intended for general pattern matching, but presently the role was taken over by Object#===.

So. How about removing Object#=~?

Concerns:

• @usa (Usaku NAKAMURA) said NilClass#=~ should be newly introduced because of: if gets =~ /re/
• Object#!~ is difficult to remove: some classes define only #=~, and expect Object#!~ to delegate to #=~.

[naruse (Yui NARUSE)]

rubocop should alert obj =~ re to fix as re =~ obj.

[znz (Kazuhiro NISHIYAMA)]

• Description updated (diff)

[nobu (Nobuyoshi Nakada)]

• Description updated (diff)

[sawa (Tsuyoshi Sawada)]

I agree. At the same time, the String=~ behaviour:

str =~ obj → integer or nil click to toggle source
Match—If obj is a Regexp, use it as a pattern to match against str,and returns the position the match starts, or nil if there is no match. Otherwise, invokes obj.=~, passing str as an argument. The default =~ in Object returns nil.
(from https://ruby-doc.org/core-2.4.0/String.html#method-i-3D~)

should be changed to raise a TypeError when obj is not a Regexp, on a par with Regexp#=~ given a non-String argument. Afterall, the following unnatural asymmetry in the current behaviour should be resolved:

// =~ Object.new # >> TypeError: no implicit conversion of Object into String
"" =~ Object.new # => nil

[znz (Kazuhiro NISHIYAMA)]

"" =~ "" #=> TypeError (type mismatch: String given)

[matz (Yukihiro Matsumoto)]

I vote for the proposed change. Let's give deprecation warning first. I don't think we need to remove !~ from Kernel.

Matz.

[mame (Yusuke Endoh)]

• Status changed from Open to Closed

Applied in changeset trunk|r65984.

---

lib/webrick: explicitly convert header values to a string

The values of @header are expected to be all strings;
WEBrick::HTTPResponse::[]=(key, val) explicitly converts the second
argument to a string and assigns it to @header hash.
However, there were some points in WEBrick internal code that assigns
non-String to @header. This change fixes the issues.

The values are checked by header_value =~ /\r\n/ in check_header.
The type confusion caused conflict with removal of Object#=~
[Feature #15231].