Work String#starts_with? with regex

[makenowjust]

This feature is imported from Ruby 2.5.0.

See: https://bugs.ruby-lang.org/issues/13712

This PR does not implement String#ends_with? for Regex because Ruby does not and it is hard to implement. You should see above link and ruby's discussion.

[ysbaddaden]

What's the benefit over specifying ^ or \A in the regular expression?

"foobar" =~ /^foo/

[makenowjust]

Readability.

[RX14]

That's a fairly confusing definition, how about:

def starts_with?(re : Regex)
  match_data = $~ = re.match_at_byte_index(self, 0, Regex::Options::ANCHORED)

  match_data != nil
end

[makenowjust]

Your code has another confusability. It seems match_data is redundant and last statement can be replaced with $~ != nil, unfortunately it cannot. It is hard to explain this reason, so I cannot accept your suggestion. Please approve.

[RX14]

I disagree.

[makenowjust]

Why?

[RX14]

I just think that this way is easier to understand. Perhaps:

match_data = re.match_at_byte_index(self, 0, Regex::Options::ANCHORED)

$~ = match_data
match_data != nil

[asterite]

This is such a small method that I think the original code is fine.

[ysbaddaden]

IMHO this is redundant and we'd expect ends_with? to also accept a regular expression, but it won't :/

My vote is "not fond but do whatever".

[makenowjust]

/^foo|bar/ =~ "foobar" is not same as /^(?:foo|bar)/ =~ "foobar". A regex beginner causes such a mistaken sometimes. It is prevented when String#starts_with? accept regex. I think it is valuable enough.

[asterite]

I think the asymmetry with ends_with is a good reason no to include this. It can be easily achieved with ^ in front of a regex.

[makenowjust]

For example, there is a method requiring a prefix:

def on_prefix(prefix)
  if gets.starts_with? prefix
    # process on given prefix is here.
  end
end

To specify two or more prefixes to this, we have to fix this method. However we don't need to fix when starts_with? accepts regex.

And now, I fixed ends_with? accepts regex. Yeah, it's perfect!

[asterite]

I agree with the use case. But the implementation of ends_with?(Regex) will create a new regex each time it's called. Not very efficient.

I don't know, maybe it's not that bad just having starts_with?(Regex), specially if Ruby implemented this. Probably others should decide if this belongs in the std.

[makenowjust]

IMO starts_with?(Regex) is useful and it prevents a bug from regex, so I don't let go due to asymmetry.

It is really hard to implement ends_with?(Regex) efficiently. I think current implementation is the only way.

[RX14]

Doesn't this require /(?:#{re})\z/?

[makenowjust]

Needless, re.to_s yields enclosed string.

[RX14]

Oh, cool.

[makenowjust]

ping

[jhass]

Let's do this, @asterite's own comment pointed out a good reason for this, the common confusion between ^ and \A.