🏷️ Issue: Support Parent-Aware Deferred Packing for Large Objects

[quickwritereader]

🏷️ Issue: Support Parent-Aware Deferred Packing for Large Objects

Summary:
PAOSP currently defers packing of large objects (e.g. nested maps, slices, containers) that exceed the 8KB container limit. This issue formalizes the behavior where such objects are logically retained by the parent frame via placeholder headers and offset patching, enabling scalable, recursive encoding without breaking offset provenance.

---

📌 Goals

• Preserve logical ownership of large deferred objects within their parent container
• Maintain offset-indexed framing and reproducible layout across multi-container structures
• Ensure teardown-safe patching and GC neutrality during deferred packing and decoding

---

🔄 Packing Strategy

• Placeholder Header Emission
  When a large object is encountered, emit a TypeContainer header in the parent with a reserved offset (e.g. 0xFFFF) signaling deferred packing.

• Deferred Queue Management
  Add the object and its logical parent offset to a packing queue for post-container processing.

• Child Container Encoding
  Encode the deferred object into a new PAOSP container, preserving offset domains and type tags.

• Offset Patching
  After encoding, patch the parent’s placeholder header with the final offset of the child container.

---

🧪 Teardown & GC Implications

• Validate offset patching logic to prevent cross-container contamination
• Ensure GC neutrality by avoiding retained slices during queue processing
• Benchmark allocation discipline across parent-child container chains
• Surface risks of offset chain corruption or decoder misalignment

---

✅ Acceptance Criteria

• Large objects (>8KB) are deferred and encoded into child containers
• Parent headers are correctly patched with child offsets
• Decoder traverses offset chains recursively and teardown-safely
• Binary layout remains reproducible across runs and packing variants

---

Issue rewritten by Copilot

[quickwritereader]

Note: Patching does not require a direct offset. In extended packing paths, the placeholder may resolve to an index into a meta offsets table, which maps logical indices to physical container offsets. This enables scalable multi-frame encoding while preserving offset provenance and decoder compatibility.

[quickwritereader]

Side note: For deferred int array blobs, a 3-byte placeholder signals deferred packing. The first byte is a type tag (e.g. int8, int16, int32, int64), and the remaining two bytes form an index into the meta offsets table. This enables scalable encoding of large typed arrays while preserving offset provenance and decoder compatibility.

[quickwritereader]

Some chunking for maps with huge amount of keys

[quickwritereader]

nullable bool as the first key argument would mean its chunkde map. because of huge amound key entries

[quickwritereader]

we could also merge Float32 and Float64 into one type, this way we will have one empty tag

[quickwritereader]

#8 ADR 001: Addressing the 13-bit Limit via Extended Containers)
#9 ADR 002: Length-Agnostic Implicit Array Encoding