Is InlineStr worth it? #1040
Description
Out of curiosity, I made the InlineStr constructors panic. Test results:
failures:
parse::test::footnote_offsets_exclamation
parse::test::issue_313
parse::test::issue_319
strings::test_special_string::cow_char_to_cow_str
strings::test_special_string::cow_str_to_cow
strings::test_special_string::cow_str_to_string
strings::test_special_string::cowstr_char_to_string
strings::test_special_string::inlinestr_ascii
strings::test_special_string::inlinestr_fits_twentytwo
strings::test_special_string::inlinestr_unicode
strings::test_special_string::small_boxed_str_clones_to_stack
None of the spec tests fail. Out of the stuff in third_party, I think only third_party/markdown-it/inline-entity.md panics (it has numerical &entities;).
I doubt InlineStr is a net negative, but it might not be helping with performance in any practical scenario. We'd need a more diverse corpus of benchmarks to properly evaluate that.