1
Fork 0
mirror of https://github.com/RGBCube/serenity synced 2025-07-25 20:37:35 +00:00

LibRegex: Implement and use a REPEAT operation for bytecode repetition

Currently, when we need to repeat an instruction N times, we simply add
that instruction N times in a for-loop. This doesn't scale well with
extremely large values of N, and ECMA-262 allows up to N = 2^53 - 1.

Instead, add a new REPEAT bytecode operation to defer this loop from the
parser to the runtime executor. This allows the parser to complete sans
any loops (for this instruction), and allows the executor to bail early
if the repeated bytecode fails.

Note: The templated ByteCode methods are to allow the Posix parsers to
continue using u32 because they are limited to N = 2^20.
This commit is contained in:
Timothy Flynn 2021-08-12 11:02:46 -04:00 committed by Linus Groh
parent a0b72f5ad3
commit 9509433e25
7 changed files with 103 additions and 16 deletions

View file

@ -40,6 +40,7 @@ using ByteCodeValueType = u64;
__ENUMERATE_OPCODE(Restore) \
__ENUMERATE_OPCODE(GoBack) \
__ENUMERATE_OPCODE(ClearCaptureGroup) \
__ENUMERATE_OPCODE(Repeat) \
__ENUMERATE_OPCODE(Exit)
// clang-format off
@ -331,10 +332,11 @@ public:
// LABEL _END = alterantive_bytecode.size
}
static void transform_bytecode_repetition_min_max(ByteCode& bytecode_to_repeat, size_t minimum, Optional<size_t> maximum, bool greedy = true)
template<typename T>
static void transform_bytecode_repetition_min_max(ByteCode& bytecode_to_repeat, T minimum, Optional<T> maximum, size_t repetition_mark_id, bool greedy = true) requires(IsIntegral<T>)
{
ByteCode new_bytecode;
new_bytecode.insert_bytecode_repetition_n(bytecode_to_repeat, minimum);
new_bytecode.insert_bytecode_repetition_n(bytecode_to_repeat, minimum, repetition_mark_id);
if (maximum.has_value()) {
auto jump_kind = static_cast<ByteCodeValueType>(greedy ? OpCodeId::ForkStay : OpCodeId::ForkJump);
@ -343,7 +345,7 @@ public:
new_bytecode.empend(jump_kind);
new_bytecode.empend(diff * (bytecode_to_repeat.size() + 2)); // Jump to the _END label
for (size_t i = 0; i < diff; ++i) {
for (T i = 0; i < diff; ++i) {
new_bytecode.extend(bytecode_to_repeat);
new_bytecode.empend(jump_kind);
new_bytecode.empend((diff - i - 1) * (bytecode_to_repeat.size() + 2)); // Jump to the _END label
@ -359,10 +361,28 @@ public:
bytecode_to_repeat = move(new_bytecode);
}
void insert_bytecode_repetition_n(ByteCode& bytecode_to_repeat, size_t n)
template<typename T>
void insert_bytecode_repetition_n(ByteCode& bytecode_to_repeat, T n, size_t repetition_mark_id) requires(IsIntegral<T>)
{
for (size_t i = 0; i < n; ++i)
// LABEL _LOOP
// REGEXP
// REPEAT _LOOP N-1
// REGEXP
if (n == 0)
return;
// Note: this bytecode layout allows callers to repeat the last REGEXP instruction without the
// REPEAT instruction forcing another loop.
extend(bytecode_to_repeat);
if (n > 1) {
empend(static_cast<ByteCodeValueType>(OpCodeId::Repeat));
empend(bytecode_to_repeat.size());
empend(static_cast<ByteCodeValueType>(n - 1));
empend(repetition_mark_id);
extend(bytecode_to_repeat);
}
}
static void transform_bytecode_repetition_min_one(ByteCode& bytecode_to_repeat, bool greedy)
@ -672,6 +692,21 @@ private:
ALWAYS_INLINE static void compare_script_extension(MatchInput const& input, MatchState& state, Unicode::Script script, bool inverse, bool& inverse_matched);
};
class OpCode_Repeat : public OpCode {
public:
ExecutionResult execute(MatchInput const& input, MatchState& state) const override;
ALWAYS_INLINE OpCodeId opcode_id() const override { return OpCodeId::Repeat; }
ALWAYS_INLINE size_t size() const override { return 4; }
ALWAYS_INLINE size_t offset() const { return argument(0); }
ALWAYS_INLINE u64 count() const { return argument(1); }
ALWAYS_INLINE size_t id() const { return argument(2); }
String const arguments_string() const override
{
auto reps = id() < state().repetition_marks.size() ? state().repetition_marks.at(id()) : 0;
return String::formatted("offset={} count={} id={} rep={}, sp: {}", offset(), count() + 1, id(), reps + 1, state().string_position);
}
};
template<typename T>
bool is(OpCode const&);