📣 New Release: codeanalyzer-python 1.0.2

[rahlk]

codeanalyzer-python (canpy) v1.0.2

TL;DR — a patch release with one fix: the Neo4j code property is populated again. No breaking changes, no schema changes; analysis.json output is byte-for-byte unaffected.

What was fixed (#104, PR #105)

Schema v2 (since 1.0.0) stores each module's source once on PyModule.source and gives every node a byte-precise span instead of a per-node code copy. The Neo4j projection, however, was still reading the removed per-node field — so every :PyClass and :PyCallable node was written without its code property:

• the py_code_fts fulltext index (over c.code / c.docstring) had no code to index, so code search in the graph silently returned nothing;
• SDK queries that read c.code — e.g. the CLDK python-sdk Neo4j backend's get_method_bodies — returned nulls, breaking parity with the in-process backend.

The projection now derives code at projection time: the owning module's source, sliced by each node's utf-8 byte span. The Neo4j vocabulary itself is unchanged (the property and index were always declared), so there is no Neo4j SCHEMA_VERSION bump — re-running an analysis against your database refreshes the affected nodes in place.

A new regression gate (test_projected_code_property_is_the_module_source_span_slice) asserts every projected class and callable carries the exact span slice of its module source.

Upgrade

pip install -U "codeanalyzer-python==1.0.2"

Links: GitHub release · CHANGELOG