Map oriented nodes to annotation rows in one place

Merged
AD
1d ago
+171-18

Add AnnotationKeyMapper, which returns the node whose annotation row an oriented node uses. In basic graphs that is the node itself. For a CanonicalDBG it is the base node, and in canonical graphs it is the canonical k-mer's node. The existing aligner's annotation buffer now uses it instead of its own mapping. Unwrapped primary graphs are rejected, as the buffer already assumed (it asserted BASIC there).

This is split out of #670, where the traversal and both aligners share the mapper, so that #670 and #672 can build on it independently. The SSHash fast path (DBGSSHash::get_base_node) is left out: that method does not exist on master yet and comes with #672, which adds the fast path when it adopts the mapper.

For every graph mode the buffer accepts, the mapping matches the code it replaces. CanonicalDBG only wraps primary graphs, so the old check on the wrapped graph's mode and the mapper's check on the wrapper's own mode agree.

A new unit test, tests/graph/test_annotation_key_mapper.cpp, covers basic graphs, the rejection of unwrapped primary graphs, and, for DNA, that both strands of a path map to the same keys in canonical and wrapped-primary graphs.

šŸ¤– Generated with Claude Code

Summary by cubic

Replaces the annotation buffer's inline node-to-annotation-key mapping with a shared AnnotationKeyMapper, so the traversal and both aligners can reuse it in upcoming work (#670, #672).

  • The mapper returns the node itself in basic graphs, the base node in CanonicalDBG, and the canonical k-mer's node in canonical graphs, spelling the path when needed.
  • Unwrapped primary graphs are now rejected explicitly instead of asserted during fetching, matching how the test helper builds primary graphs wrapped in CanonicalDBG.

Adds unit tests covering basic graphs, primary-graph rejection, and both strands of a path mapping to the same keys in canonical and wrapped-primary graphs. The primary- and canonical-graph tests run for DNA only, since building those graphs for the protein alphabet crashes.