mruby-regexp: add the regexp form of String#[] and String#slice - #7054

Problem

String#[] and its slice twin accept an Integer, a String and a Range, but not a
Regexp. str_convert_range() in src/string.c funnels every non-String, non-Range
argument through mrb_ensure_int_type(), so the regexp forms raise a TypeError that names
Integer, which does not hint at what is actually missing. mruby-regexp never overrode
[] or slice, and the gem README listed this under Limitations.

"hello"[/l+/]            # CRuby: "ll",  mruby: TypeError (Regexp cannot be converted to Integer)
"hello".slice(/l+/)      # CRuby: "ll",  mruby: TypeError
"hello"[/(l+)(o)/, 1]    # CRuby: "ll",  mruby: TypeError
"hello"[/(?<x>l+)/, :x]  # CRuby: "ll",  mruby: TypeError
"hello"[/z/]             # CRuby: nil,   mruby: TypeError
:hello[/l+/]             # CRuby: "ll",  mruby: TypeError

Change

mrbgems/mruby-regexp/mrblib/string_regexp.rb overrides [], following the pattern
split already uses in the same file: the C-defined method is aliased as __aref, and
the override hands every non-Regexp argument list straight back to it.

The delegation happens before any argument is inspected, so the non-regexp forms keep the
arity and type errors mrb_get_args() raises. With no arguments at all, args[0] is nil,
the guard fails, and __aref() raises the same ArgumentError as before. Module#=== reads
the argument's real type, since is_a? is redefinable.

A Regexp goes through Regexp#match and not #match?: the match globals have to be
published here, including the clearing a failed match does, which is why the MatchData is
fetched even when no capture was asked for. The capture argument reaches MatchData#[]
untouched, which already gives the CRuby semantics: a negative index counts back from the
last group, an index past the last group is nil, and a name that resolves to no group
raises IndexError.

slice is a second method table entry for the same C function rather than an alias of
[], in mruby and in CRuby alike, so the override is aliased to slice as well. That is
also what makes sym[re] work, and it needs no change of its own: Symbol#[] is an alias
of Symbol#slice in mruby-symbol-ext, which delegates to String#slice.

Interaction with the inline index opcodes

vm_op_getidx() answers str[Integer], str[String] and str[Range] from C, and
vm_op_getidx0() answers str[0]. Both guard on the receiver's class only, not on whether
String#[] has been redefined, so those forms keep bypassing this override. That is
harmless and desirable: they are exactly the ones the override would have delegated back to
__aref() unchanged, so behaviour is identical and they pay nothing. A Regexp index falls
into the default arm of the opcode's type switch, leaves through getidx_fallback, and
arrives here as an ordinary send.

str[i, len] and every slice call are not opcode receivers and do reach the override,
paying a Ruby frame on their way to __aref (measured at roughly 3x the direct C call on a
tight 300k-iteration loop). This is the cost of the feature living in mrblib; moving the
regexp branch into C would avoid it, at the price of a callback into the VM from
str_convert_range().

Documentation

The gem README loses its No regexp form of String#[] limitation and gains the new
forms in the usage list, and the header comment of mrblib/symbol_regexp.rb no longer says
sym[/re/] is missing.

The write side, str[re] = repl and str.slice!(re), is deliberately out of scope: both
are destructive and belong with the other destructive-method work. The removed limitation
item was about reading only, so it is dropped rather than narrowed.

Tests

mrbgems/mruby-regexp/test/regexp.rb gains assertions for the plain match, captures by
index / name / symbol, negative and out-of-range capture indexes, failed matches, the match
globals, the untouched non-regexp forms including a String subclass receiver (which the
opcodes never answer), the argument errors, and an argument that lies about its own type.
test/symbol_regexp.rb gains the sym[re] forms, guarded by a skip because mrbtest
builds this gem's tests without mruby-symbol-ext.

rake test passes with no failures, both in a default build and in one with
MRB_UTF8_STRING. Checked against CRuby 4.0.6.

Summary by CodeRabbit

• New Features

  • Added regular expression support to String#[] and String#slice, including capture selection and named captures.
  • Added support for regular expression indexing and slicing on symbols when symbol slicing is available.
  • Existing string and symbol indexing forms remain supported.

• Bug Fixes

  • Improved handling of unmatched patterns, invalid arguments, and lookalike regexp objects.

• Documentation

  • Updated regexp documentation to reflect the newly supported operations.