Strings

Interned strings

How do I represent interned strings?

• Gecko has nsIAtom, can't pass directly
• WebKit has a subclass of String called InternedString
  • not clear how comparison works.
• bz likes the idea of having interned strings be a subclass of normal strings
• Want comparison between a regular string and an interned string to not require interning the regular string first.
• This can be done by just following the pointers and comparing the contents.
• In the JS engine JSAtom is just a subclass of JSString.

Mutability

Should our strings be mutable?

• Gecko has mutable strings and this is bad for performance
• Shows up in microbenchmarks
• Immutable strings and string builders

Interned strings again

How about sharing?

• CSS parser and DOM code

What are interned?

• Tag names on DOM nodes
• Attribute names on DOM nodes
• Attribute values (XUL only?)
• IDs on nodes
• Classes
• Tag selectors
• Class selectors
• ID selectors
• During selector matching, we do not need to intern anything
• We must prevent the layout task from taking strong references to interned strings. If layout needs to hold onto the string, it must copy it. This is usually not in performance-critical code, so this is OK.
• 3 kinds of strings:
  • HeapString
  • StaticInternedString
  • RefCountedInternedString
• Nice thing about interned strings: Precomputed hash codes. This is used for the Bloom filter.

Cost of creating string objects

• Constructors and especially destructors are expensive
• No static typing
• JS string comes in, want to create a Gecko DependentString, constructor was expensive because it had to check whether it was a DependentString
• Would be nice to avoid hacks like that
• 3 cases that Gecko has: ref counted versus owned versus dependent string versus null-terminated versus stack buffer
• Stay as simple as possible, don't add new string types unless they're really necessary!

Encodings: UTF-8 versus UTF-16

• We use UTF-16 in Gecko but UTF-8 in Servo
• Gecko (and WK) uses a ASCII encoding for text runs
• Could do something DOM-specific to avoid UTF-16 -> UTF-8 conversions in pure DOM->JS microbenchmarky calls