Add String::replace_char(s) methods for performance and convenience

[AThousandShips]

Will try do some performance comparison soon, but these replace in-place essentially, avoiding any COW or other manipulation, and also avoids creating any temporaries etc.

The multi-character one is optional but added for completeness for a few cases where multiple cases are replaced at once

Can expose to scripting if desired, but generally using characters is a bit more complicated in scripting so haven't added yet

Now exposed, can drop if desired, but this should work well, the exposed methods aren't entirely useful in GDScript but are very useful in GDExtension I would say

Kept as separate commits for ease of editing until approval

See #92433 (comment)

A few more complex cases like String::validate_filename, OS::get_safe_dir_name, and get_csharp_project_name could be replaced using this, instead using a vector of characters, but for right now I keep this to the simple cases (mostly, see AnimationLibrary::validate_library_name for an exception)

See also:

• Add String::remove_char(s) methods for performance and convenience #92476

[AThousandShips]

Uploaded an alternative approach based on the avove, any benchmarking would be appreciated!

[AThousandShips]

Suggested change

	if (old_ptr[index] == p_key) {
	new_ptr[index] = p_with;
	} else {
	new_ptr[index] = old_ptr[index];
	const char32_t old_char = old_ptr[index];
	if (old_char == p_key) {
	new_ptr[index] = p_with;
	} else {
	new_ptr[index] = old_char;

Possible improvement to avoid accessing twice, but should be optimized out likely

[Ivorforce]

Looks good! I have previously shown in #100024 that a single-character replace can be up to 3x faster than string-replace, so I think this is warranted.

I have benchmarked this implementation, and it's a bit faster than the one I had (likely due to migrating the if changed outside the loop, or because of the memcpy):

	String s = "Who is Frederic?Who is Frederic?Who is Frederic?";
	auto t0 = std::chrono::high_resolution_clock::now();
	for (int i = 0; i < 1000; i ++) {
		String s1 = s.replace("e", "b");
	}
	auto t1 = std::chrono::high_resolution_clock::now();
	std::cout << std::chrono::duration_cast<std::chrono::microseconds>(t1 - t0).count() << "us\n";
	t0 = std::chrono::high_resolution_clock::now();
	for (int i = 0; i < 1000; i ++) {
		String s1 = s.replace_char('e', 'b');
	}
	t1 = std::chrono::high_resolution_clock::now();
	std::cout << std::chrono::duration_cast<std::chrono::microseconds>(t1 - t0).count() << "us\n";
	t0 = std::chrono::high_resolution_clock::now();
	for (int i = 0; i < 1000; i ++) {
		String s1 = s.replace("#", "b");
	}
	t1 = std::chrono::high_resolution_clock::now();
	std::cout << std::chrono::duration_cast<std::chrono::microseconds>(t1 - t0).count() << "us\n";
	t0 = std::chrono::high_resolution_clock::now();
	for (int i = 0; i < 1000; i ++) {
		String s1 = s.replace_char('#', 'b');
	}
	t1 = std::chrono::high_resolution_clock::now();
	std::cout << std::chrono::duration_cast<std::chrono::microseconds>(t1 - t0).count() << "us\n";

Resulting in

249us
61us
24us
18us

I just have a few small comments, but otherwise I'm fully behind a merge!

[Ivorforce]

Ah, I just about forgot: I think it would be good to check the string length for 1 (and !is_case_insensitive) in _replace_common, and if true, switch to use the replace_char implementation. This should help GDScript also benefit from the single-char implementation, since it has no char data type and cannot otherwise call it.

[AThousandShips]

Missed the earlier review but went over that, will push some changes in a few hours

Using String as an input for the keys might actually be a good idea, will look into it as an improvement and binding the method as it'd be useful there

Will investigate using variations on String for the keys and add binds for these as optional additions, as I think these would be especially useful for extensions, but will keep them separate so they can be split off if desired

[AThousandShips]

Covered the above suggestions, but leaving the change to replace_chars for the time being, it's a bit convoluted to add some of the details but should probably be worth it to do it, but keeping this from expanding too far into things

Improvements on that side can be part of a followup if desired, together with exposing this

Edit: worked out the String/char * versions so will clean this up and move to the alternative solution and push tonight or tomorrow, it works well and using that format is far more convenient in my testing

[Ivorforce]

Re-checked after the changes, looks even better now! I had approved earlier, and I'm still approving :)

[Ivorforce]

What about my earlier suggestion to add this here?

if (p_keys.size() == 1) {
    return replace_char(p_keys[0], p_with;
}

[AThousandShips]

Good point can test

[AThousandShips]

Been busy but will push a cleaned up and improved version in the next few days

Edit: Testing the new code and will push soon

[AThousandShips]

Will push parallel changes for remove_char(s) as well

[AThousandShips]

Can easily make a replace_chars(const Vector<char32_t> &, char32_t) version as well, it can use the same method, but leaving with this for now

[Ivorforce]

No need i think, there's little chance somebody even owns vector<char32_t> instead of String.

[AThousandShips]

Will move this to the individual functions for easier use, but away from the computer right now