hasIndices flag for regex

[roryabraham]

Problem

We were trying to use a regex with the -d / hasIndices flag, and noticed that it worked on web but not iOS/Android. I then discovered that it looks like this flag is just not yet implemented in the Hermes engine:

hermes/include/hermes/Regex/RegexTypes.h

Lines 213 to 343 in 4171902

	class SyntaxFlags {
	private:
	// Note these are encoded into bytecode files, so changing their values is a
	// breaking change.
	enum : uint8_t {
	ICASE = 1 << 0,
	GLOBAL = 1 << 1,
	MULTILINE = 1 << 2,
	UCODE = 1 << 3,
	DOTALL = 1 << 4,
	STICKY = 1 << 5
	};
	public:
	// Note: Preserving the order here and ensuring it lines up with the order of
	// the offsets above makes the generated assembly more efficient.
	// Specifically, it makes the conversion to/from a byte almost free.
	uint8_t ignoreCase : 1;
	uint8_t global : 1;
	uint8_t multiline : 1;
	uint8_t unicode : 1;
	uint8_t dotAll : 1;
	uint8_t sticky : 1;
	/// \return a byte representing the flags. Bits are set based on the offsets
	/// specified above. This is used for serialising the flags to bytecode.
	uint8_t toByte() const {
	uint8_t ret = 0;
	if (global)
	ret |= GLOBAL;
	if (ignoreCase)
	ret |= ICASE;
	if (multiline)
	ret |= MULTILINE;
	if (unicode)
	ret |= UCODE;
	if (sticky)
	ret |= STICKY;
	if (dotAll)
	ret |= DOTALL;
	return ret;
	}
	/// \return a SyntaxFlags struct generated from the given byte. Bit offsets
	/// are specified above. This is used for deserialising the flags from
	/// bytecode.
	static SyntaxFlags fromByte(uint8_t byte) {
	SyntaxFlags ret = {};
	if (byte & GLOBAL)
	ret.global = 1;
	if (byte & ICASE)
	ret.ignoreCase = 1;
	if (byte & MULTILINE)
	ret.multiline = 1;
	if (byte & UCODE)
	ret.unicode = 1;
	if (byte & STICKY)
	ret.sticky = 1;
	if (byte & DOTALL)
	ret.dotAll = 1;
	return ret;
	}
	/// \return a string representing the flags
	/// The characters are returned in the order given in ES 6 21.2.5.3
	/// (specifically global, ignoreCase, multiline, unicode, sticky)
	/// Note this may differ in order from the string passed in construction
	llvh::SmallString<6> toString() const {
	llvh::SmallString<6> result;
	if (global)
	result.push_back('g');
	if (ignoreCase)
	result.push_back('i');
	if (multiline)
	result.push_back('m');
	if (unicode)
	result.push_back('u');
	if (sticky)
	result.push_back('y');
	if (dotAll)
	result.push_back('s');
	return result;
	}
	/// Given a flags string \p str, generate the corresponding SyntaxFlags
	/// \return the flags if the string is valid, an empty optional otherwise
	/// See ES 5.1 15.10.4.1 for description of the validation
	static llvh::Optional<SyntaxFlags> fromString(
	const llvh::ArrayRef<char16_t> flags) {
	// A flags string may contain i,m,g, in any order, but at most once each
	auto error = llvh::NoneType::None;
	SyntaxFlags ret = {};
	for (auto c : flags) {
	switch (c) {
	case u'i':
	if (ret.ignoreCase)
	return error;
	ret.ignoreCase = 1;
	break;
	case u'm':
	if (ret.multiline)
	return error;
	ret.multiline = 1;
	break;
	case u'g':
	if (ret.global)
	return error;
	ret.global = 1;
	break;
	case u'u':
	if (ret.unicode)
	return error;
	ret.unicode = 1;
	break;
	case u'y':
	if (ret.sticky)
	return error;
	ret.sticky = 1;
	break;
	case u's':
	if (ret.dotAll)
	return error;
	ret.dotAll = 1;
	break;
	default:
	return error;
	}
	}
	return ret;
	}
	};

Solution

Implement the -d / hasIndices flag for Hermes regex.

Additional Context

Our use case for the -d flag is described here

[roryabraham]

This was implemented in #968 and should land in the next hermes release I think?

[roryabraham]

2afc7b0 does not appear to be included in a release yet, but hopefully soon 🙂

[fbmal7]

@roryabraham This feature will be in RN 0.73.