Class: TreeHaver::Backends::FFI::Parser

Inherits:

Object

• Object
• TreeHaver::Backends::FFI::Parser

Defined in:

lib/tree_haver/backends/ffi.rb

Overview

FFI-based tree-sitter parser

Wraps a TSParser pointer and manages its lifecycle with a finalizer.

Instance Method Summary

• #initialize ⇒ Parser constructor

  Create a new parser instance.

• #language=(lang) ⇒ Language

  Set the language for this parser.

• #parse(source) ⇒ Tree

  Parse source code into a syntax tree.

Constructor Details

#initialize ⇒ Parser

Create a new parser instance

Raises:

• (TreeHaver::NotAvailable) —

  if FFI not available or parser creation fails

# File 'lib/tree_haver/backends/ffi.rb', line 415

def initialize
  raise TreeHaver::NotAvailable, "FFI not available" unless Backends::FFI.available?

  Native.try_load!
  @parser = Native.ts_parser_new
  raise TreeHaver::NotAvailable, "Failed to create ts_parser" if @parser.null?

  # Note: We intentionally do NOT register a finalizer here because:
  # 1. ts_parser_delete can segfault if called during certain GC scenarios
  # 2. The native library may be unloaded before finalizers run
  # 3. Parser cleanup happens automatically on process exit
  # 4. Long-running processes should explicitly manage parser lifecycle
  #
  # If you need explicit cleanup in long-running processes, store the
  # parser in an instance variable and call a cleanup method explicitly
  # when done, rather than relying on GC finalizers.
end

Instance Method Details

#language=(lang) ⇒ Language

Set the language for this parser

Note: FFI backend is special - it receives the wrapped Language object because it needs to call to_ptr to get the FFI pointer. TreeHaver::Parser detects FFI Language wrappers (respond_to?(:to_ptr)) and passes them through.

Parameters:

• lang (Language) —

  the FFI language wrapper (not unwrapped)

Returns:

• (Language) —

  the language that was set

Raises:

• (TreeHaver::NotAvailable) —

  if setting the language fails

# File 'lib/tree_haver/backends/ffi.rb', line 442

def language=(lang)
  # Defensive check: ensure we received an FFI Language wrapper
  unless lang.is_a?(Language)
    raise TreeHaver::NotAvailable,
      "FFI backend expected FFI::Language wrapper, got #{lang.class}. " \
        "This usually means TreeHaver::Parser#unwrap_language passed the wrong type. " \
        "Check that language caching respects backend boundaries."
  end

  # Additional check: verify the language is actually for FFI backend
  if lang.respond_to?(:backend) && lang.backend != :ffi
    raise TreeHaver::NotAvailable,
      "FFI backend received Language for wrong backend: #{lang.backend}. " \
        "Expected :ffi backend. Class: #{lang.class}. " \
        "Path: #{lang.path.inspect}, Symbol: #{lang.symbol.inspect}"
  end

  # Verify the DynamicLibrary is still valid (not GC'd)
  # The Language stores @library to prevent this, but let's verify
  lib = lang.instance_variable_get(:@library)
  if lib.nil?
    raise TreeHaver::NotAvailable,
      "FFI Language has no library reference. The dynamic library may have been unloaded. " \
        "Path: #{lang.path.inspect}, Symbol: #{lang.symbol.inspect}"
  end

  # Verify the language has a valid pointer
  ptr = lang.to_ptr

  # Check ptr is actually an FFI::Pointer
  unless ptr.is_a?(::FFI::Pointer)
    raise TreeHaver::NotAvailable,
      "FFI Language#to_ptr returned #{ptr.class}, expected FFI::Pointer. " \
        "Language class: #{lang.class}. " \
        "Path: #{lang.path.inspect}, Symbol: #{lang.symbol.inspect}"
  end

  ptr_address = ptr.address

  # Check for NULL (0x0)
  if ptr.nil? || ptr_address.zero?
    raise TreeHaver::NotAvailable,
      "FFI Language has NULL pointer. Language may not have loaded correctly. " \
        "Path: #{lang.path.inspect}, Symbol: #{lang.symbol.inspect}"
  end

  # Check for small invalid addresses (< 4KB are typically unmapped memory)
  # Common invalid addresses like 0x40 (64) indicate corrupted or uninitialized pointers
  if ptr_address < 4096
    raise TreeHaver::NotAvailable,
      "FFI Language has invalid pointer (address 0x#{ptr_address.to_s(16)}). " \
        "This usually indicates the language library was unloaded or never loaded correctly. " \
        "Path: #{lang.path.inspect}, Symbol: #{lang.symbol.inspect}"
  end

  # Note: MRI backend conflict is now handled by TreeHaver::BackendConflict
  # at a higher level (in TreeHaver.resolve_backend_module)

  # lang is a wrapped FFI::Language that has to_ptr method
  ok = Native.ts_parser_set_language(@parser, ptr)
  raise TreeHaver::NotAvailable, "Failed to set language on parser" unless ok

  lang # rubocop:disable Lint/Void (intentional return value)
end

#parse(source) ⇒ Tree

Parse source code into a syntax tree

Parameters:

• source (String) —

  the source code to parse (should be UTF-8)

Returns:

• (Tree) —

  raw backend tree (wrapping happens in TreeHaver::Parser)

Raises:

• (TreeHaver::NotAvailable) —

  if parsing fails

# File 'lib/tree_haver/backends/ffi.rb', line 512

def parse(source)
  src = String(source)
  tree_ptr = Native.ts_parser_parse_string(@parser, ::FFI::Pointer::NULL, src, src.bytesize)
  raise TreeHaver::NotAvailable, "Parse returned NULL" if tree_ptr.null?

  # Return raw FFI::Tree - TreeHaver::Parser will wrap it
  Tree.new(tree_ptr)
end