v0.77.0

Fixed

• Serializer Roundtrip Whitespace Corruption (IMPL-SERIALIZER-ROUNDTRIP-CORRUPTION-001): Fixed data corruption during roundtrip serialization of patterns with embedded leading whitespace:

  • Previous: Patterns with TextElements where leading whitespace follows a newline (e.g., "Line 1\n Line 2") would lose the semantic whitespace on roundtrip
  • Root cause: Serializer's per-element indentation logic didn't account for inter-element whitespace semantics; parser's common_indent calculation would strip all indentation
  • Now: Added _pattern_needs_separate_line() helper to detect problematic patterns; serializer emits such patterns on separate lines with continuation indent to preserve whitespace through roundtrip
  • Impact: Multi-line messages with indented continuation lines (code examples, formatted text) now preserve all whitespace correctly
  • Location: syntax/serializer.py lines 541-552, 557-559

• Cache Tuple Argument Handling (IMPL-CACHE-TUPLE-REJECTION-001): Fixed FormatCache silently bypassing cache for arguments containing tuples:

  • Previous: _make_hashable() lacked case tuple(): handler, causing tuples to fall through to default case and raise TypeError, which was caught and treated as unhashable
  • Now: Added tuple handler that recursively processes elements, consistent with list handling
  • Impact: Format calls with tuple arguments now correctly cache results instead of recomputing every time
  • Location: runtime/cache.py lines 343-346

• Validation CRLF Line Position Accuracy (IMPL-VALIDATION-OFFSET-MISMATCH-001): Fixed validation error line/column positions for files with CRLF line endings:

  • Previous: LineOffsetCache used raw source positions while AST spans used normalized (LF-only) positions, causing cumulative drift in error locations for CRLF files
  • Root cause: Parser normalizes CRLF/CR to LF internally, but validation was building offset cache from raw source
  • Now: Validation normalizes source using same regex pattern as parser before building LineOffsetCache
  • Impact: Windows-originated FTL files now report correct error locations without position drift
  • Location: validation/resource.py lines 678-682

• Parser Attribute Blank Line Handling (IMPL-PARSER-ATTRIBUTE-BLANK-LINE-001): Fixed attribute parsing to support blank lines between attributes per Fluent specification:

  • Previous: Parser terminated attribute parsing on blank lines, silently dropping subsequent attributes
  • Root cause: Loop consumed single newline then checked for ., breaking on blank lines (second newline)
  • Now: Added inner loop to skip consecutive newlines (blank lines) before checking for attribute marker
  • Spec compliance: Fluent EBNF Attribute ::= line_end blank? "." ... where blank ::= (blank_inline | line_end)+
  • Impact: FTL files with blank lines between attributes for readability now parse correctly
  • Location: syntax/parser/rules.py lines 1719-1722

Added

• Added _pattern_needs_separate_line() helper method to serializer for detecting patterns requiring separate-line serialization
• Added _CONT_INDENT constant in serializer for continuation line indentation