Parse each package part once, and bound the total

[AbhinavMir]

Packages now cache parsed part trees, so a part referenced many times parses once instead of once per reference, with a running node total bounding the parts too large to cache.

For issue #43.

[abimaelmartell]

@cubic-dev-ai

[cubic-dev-ai]

@cubic-dev-ai

@abimaelmartell I have started the AI code review. It will take a few minutes to complete.

[cubic-dev-ai]

1 issue found across 8 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="src/package/archive.rs">

<violation number="1" location="src/package/archive.rs:172">
P2: The parsed-tree cache fills monotonically and never evicts: `cached_nodes` only grows and `trees` is never cleared, so once the cache holds `MAX_CACHED_XML_NODES` (200k) nodes, every later part is permanently served uncached — even small parts that are repeated heavily. Combined with the never-decreasing `parsed_nodes` document total, a large-but-legitimate document whose repeated parts come after other parts filled the cache will re-parse each repeat and charge its full node count, and can hit a spurious `max_document_xml_nodes` ResourceLimit even though no part is individually abusive. The cache retains whichever parts parsed first (which may never be re-referenced) while failing to cache the parts that are actually repeated. Consider evicting cached entries (e.g., LRU or dropping the least-useful part when the budget is exceeded) so that repeated parts stay parse-once, or at least re-cache a part that is demonstrably referenced more than once.</violation>
</file>

<sub>Reply with feedback, questions, or to request a fix.

Fix all with cubic | Re-trigger cubic</sub>

[cubic-dev-ai]

P2: The parsed-tree cache fills monotonically and never evicts: cached_nodes only grows and trees is never cleared, so once the cache holds MAX_CACHED_XML_NODES (200k) nodes, every later part is permanently served uncached — even small parts that are repeated heavily. Combined with the never-decreasing parsed_nodes document total, a large-but-legitimate document whose repeated parts come after other parts filled the cache will re-parse each repeat and charge its full node count, and can hit a spurious max_document_xml_nodes ResourceLimit even though no part is individually abusive. The cache retains whichever parts parsed first (which may never be re-referenced) while failing to cache the parts that are actually repeated. Consider evicting cached entries (e.g., LRU or dropping the least-useful part when the budget is exceeded) so that repeated parts stay parse-once, or at least re-cache a part that is demonstrably referenced more than once.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At src/package/archive.rs, line 172:

<comment>The parsed-tree cache fills monotonically and never evicts: `cached_nodes` only grows and `trees` is never cleared, so once the cache holds `MAX_CACHED_XML_NODES` (200k) nodes, every later part is permanently served uncached — even small parts that are repeated heavily. Combined with the never-decreasing `parsed_nodes` document total, a large-but-legitimate document whose repeated parts come after other parts filled the cache will re-parse each repeat and charge its full node count, and can hit a spurious `max_document_xml_nodes` ResourceLimit even though no part is individually abusive. The cache retains whichever parts parsed first (which may never be re-referenced) while failing to cache the parts that are actually repeated. Consider evicting cached entries (e.g., LRU or dropping the least-useful part when the budget is exceeded) so that repeated parts stay parse-once, or at least re-cache a part that is demonstrably referenced more than once.</comment>

<file context>
@@ -132,9 +152,28 @@ impl<'a> Package<'a> {
+        let before = self.parsed_nodes;
+        let tree = Rc::new(parse_xml_counted(bytes, &mut self.parsed_nodes)?);
+        let nodes = self.parsed_nodes - before;
+        if self.cached_nodes + nodes <= limits::MAX_CACHED_XML_NODES {
+            self.cached_nodes += nodes;
+            self.trees.insert(name.to_string(), Rc::clone(&tree));
</file context>