fix(core): preserve data integrity and bound shared UI and response work

Release v0.1.46. Coordinated integrity review: GovOPlaN/govoplan-core#298.

src/govoplan_core/server/conditional_requests.py

@@ -1,12 +1,15 @@
 from __future__ import annotations
 
 import hashlib
-from collections.abc import Awaitable, Callable
+from collections.abc import AsyncIterator, Awaitable, Callable
 
 from fastapi import Request
 from starlette.responses import Response
 
 JSON_CACHE_CONTROL = "private, no-cache"
+# This bounds middleware-owned buffering, not the size of a route response.
+# Larger responses retain their streaming iterator and are never truncated.
+MAX_CONDITIONAL_JSON_BYTES = 1_048_576
 JSON_ETAG_VARY_HEADERS = (
     "Authorization",
     "Cookie",
@@ -26,15 +29,34 @@ async def conditional_json_get_middleware(
 
     The middleware deliberately works after route handling. That keeps the
     contract platform-wide without requiring every module router to learn about
-    conditional requests, while still limiting buffering to successful JSON GET
-    responses.
+    conditional requests. Only small successful JSON responses are buffered;
+    larger responses stream unchanged. Authorization still runs on every GET.
     """
 
     response = await call_next(request)
     if not _eligible_for_conditional_json_get(request, response):
         return response
 
-    body = b"".join([chunk async for chunk in response.body_iterator])
+    response.headers["cache-control"] = _conditional_cache_control(response.headers.get("cache-control"))
+    response.headers["vary"] = _merge_vary(response.headers.get("vary"), JSON_ETAG_VARY_HEADERS)
+    content_length = response.headers.get("content-length", "")
+    if content_length.isascii() and content_length.isdecimal() and int(content_length) > MAX_CONDITIONAL_JSON_BYTES:
+        return response
+
+    chunks: list[bytes] = []
+    size = 0
+    iterator = response.body_iterator
+    async for chunk in iterator:
+        if not chunk:
+            continue
+        chunks.append(chunk)
+        size += len(chunk)
+        if size > MAX_CONDITIONAL_JSON_BYTES:
+            # Include the crossing chunk exactly once, without draining the
+            # rest of the producer or copying a potentially large chunk.
+            response.body_iterator = _replay_prefix(chunks, iterator)
+            return response
+    body = b"".join(chunks)
     etag = response.headers.get("etag") or json_response_etag(body)
     headers = dict(response.headers)
     headers["etag"] = etag
@@ -48,6 +70,14 @@ async def conditional_json_get_middleware(
     return Response(content=body, status_code=response.status_code, headers=headers, background=response.background)
 
 
+async def _replay_prefix(chunks: list[bytes], iterator: AsyncIterator[bytes]) -> AsyncIterator[bytes]:
+    for chunk in chunks:
+        yield chunk
+    chunks.clear()
+    async for chunk in iterator:
+        yield chunk
+
+
 def json_response_etag(body: bytes) -> str:
     digest = hashlib.sha256(body).hexdigest()
     return f'W/"sha256-{digest}"'