HTML API: Plans for 6.8

[dmsnell]

🔝Block API | HTML issues | Refactors | ↑ Broader Roadmap | ← Plans for 6.7

Lingering work from 6.7

• Speed speed speed. Make the HTML Processor 10x faster.

  • Remove all if statements that don't execute anything (they have a comment as their body).
  • If 6.7 includes full support for all HTML tags, measure the impact of reordering the case statements in each insertion mode. Test against 100s of 1000s of websites based on web popularity.
  • Profile the parsing of 100s of 1000s of websites and see if anything surprising pops up in the results.
  • Replace '#text' === $token_type with ::STATE_TEXT_NODE === $this->parser_state
  • Eagerly set token name, type in step() where all nodes are real. Reference these values instead of calling ->get_token_name() etc…
  • Remove after_element_push() since these are all instigated from within the HTML Processor, unlike pop with pop_until() (unless we made pop_until() return a generator and we could foreach ( $state->pop_until( 'TAG' ) as $popped )
  • Flagification
    • Replace as many repetitive if checks with flags that are set on events, as is done with has_p_in_button_scope.
    • Indicate once in next_token() if a text node is only whitespace.

• Following the change to push/pop, immediately pop elements off of the stack of elements as instructed in the parsing rules, vs. letting step() perform the check and pop.

With some initial explorations I've found 16% - 40% speed improvement with some of these ideas. That's not good enough, but it's a start.

New Features and Interfaces

• Introduce safe composable HTML templating.

Blocks

• Continue developing block attribute sourcing.
• Iterate on the Server Directive Processor.
• Iterate on the Block Bindings processor.