Repository navigation
[Feature]: don't erase weaker-confidence evidence when edges or paths merge #4881
Description
Repro for all (graphify 0.9.97, v8 @ 0a8cd4e), Report.cs:
public interface IReport { string Build(); }
public class PdfReport : IReport { public string Build() { return "pdf"; } }
public class CsvReport : IReport { public string Build() { return "csv"; } }
public class Exporter {
private readonly IReport _report;
public Exporter(IReport report) { _report = report; }
public string Run() { return _report.Build(); }
}graphify update .
Today: a same-relation edge collision keeps the higher-confidence edge (#3711, graphify/build.py:1910–1911), so an AMBIGUOUS/INFERRED duplicate disappears. Paths are single shortest paths, so two routes to the same symbol are never compared.
Expected: on collision, keep the winner but record the other provenance (e.g. also_seen: [{confidence, source_location}]), and when reporting multiple paths to one target, surface the weakest path confidence instead of the best. Display already unions tags for parallel edges (serve.py:2144–2145, [EXTRACTED/INFERRED]); this would make the stored graph match.
Refs: graphify/build.py:191, 1910–1911; graphify/serve.py:2144–2145.
Activity
github-actions commentedon Oct 30, 2026
Thanks for opening this issue, @SyedFahad7. A maintainer will take a look soon.
If you would like to discuss it in real time, come say hi on our Discord server. For longer-form questions and ideas there is also GitHub Discussions.
To help us triage, please make sure the report includes what you expected, what actually happened, and the steps (and a small sample) to reproduce it.
Merge pull request #4891 from SyedFahad7/fix/4881-keep-weaker-evidence
feat(path): report the weakest of several equal-length routes (#4881)
test: reconcile confidence-score output tests with #4881/#4889 (for #…
safishamsi commentedon Oct 30, 2026
The second half shipped in v0.9.99:
graphify path/shortest_pathnow report the weakest of several equal-length routes instead of an arbitrary one, so the confidence never overstates the connection. Both halves of this issue are now resolved (first half,also_seen, shipped in 0.9.98 via #4891).fix(build): keep a collapsed edge's weaker evidence in also_seen (Gra…