fix(callflow): disclose diagram truncation — "18 of 1223 nodes" was only in a Mermaid comment

[Tomaskobel]

Problem

Section diagrams are capped at max_diagram_nodes / max_diagram_edges (18 / 24 by default). The truncation counts are already computed — but emitted where no reader can see them:

omitted_nodes = max(0, len(nodes) - len(selected_nodes))
omitted_edges = max(0, len(visible_edges) - included)
if omitted_nodes or omitted_edges:
    lines.append(f"    %% Omitted for readability: {omitted_nodes} nodes, {omitted_edges} edges")

%% is a Mermaid comment. It is never rendered.

What the reader does see is the section intro, which gives the totals and then hedges without a number:

This section covers 1223 nodes and 47 internal edges; the diagram shows only representative relationships to stay readable.

Directly beneath that sentence sits a diagram containing 18 nodes. Nothing on the page distinguishes "this diagram is the section" from "this diagram is 1.5% of the section", so a reader who has just been told the section has 1,223 nodes reasonably reads the picture as complete. On the graph I tested, one section drew 18 of 294 and another 18 of 1,223.

Fix

Render the coverage as visible text under each diagram:

Diagram draws 18 of 1223 nodes and 10 of 10 edges; the rest are omitted for readability.

Two supporting changes:

• select_diagram_content extracts the node/edge selection that generate_section_flowchart was doing inline, so the note and the drawing are computed from one place and cannot drift. A disclosure derived independently of the thing it describes will eventually lie.
• The line is omitted entirely when nothing was cut, so complete diagrams stay uncluttered.

Localised (zh/en) via the existing pick_text, styled with the existing --muted token.

Verification

On a 4,990-node / 10,135-edge graph: 15 sections rendered, 15 coverage notes, and every note agrees exactly with the renderer's own omitted_* counts — checked by parsing both the visible text and the Mermaid comment out of the HTML and diffing them. 0 disagreements.

Notes

• Two regression tests added; both fail on v8 as-is (the function does not exist) and pass with the fix.
• Full suite: 3,832 passing, failure set identical to v8 (30 pre-existing). test_labeling.py::test_label_communities_batches_when_over_batch_size is flaky — it failed 3 of 4 runs on unpatched v8 when I checked.
• The Mermaid %% comment is left in place; it is harmless and useful when reading the generated source.

[graphify-labs]

Graphify reviewed this change.

Looks safe to merge — no coupling regressions and no blocking issues, checked against the code graph (not a self-assessment).

---

Graphify review — findings

This PR appears to revert the unreleased 0.9.35 changelog section, folding it back and re-marking 0.9.34 as unreleased. In graphify/build.py it removes the recently-added helpers _build_prune_sets and _derive_prune_root, simplifies _infer_merge_root back to a plain grandparent heuristic, and inlines the prune-set construction in merge_raw_extraction (dropping the absolute-path root derivation and zero-match handling). The changed-symbol list also spans many extractor, serve, callflow, detect, and watch modules along with their tests, so the surface area extends across graph extraction, querying, path/direction handling, and merge/prune logic rather than a single area.

No blocking issues surfaced. 4 lower-confidence candidates did not survive cross-model review.

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 3438 functions depend on the 1573 functions this change touches.

Health — this change adds coupling hotspots:

• worse: write_callflow_html() — 7 callers, 26 callees

Verification — 3438 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 3138 function(s) in the blast radius were not formally verified this run

· 1 more finding(s) on lines outside this diff (see the check run).

[graphify-labs]

Graphify reviewed this change.

Looks safe to merge — no coupling regressions and no blocking issues, checked against the code graph (not a self-assessment).

---

Graphify review — findings

This PR adds a "diagram coverage" disclosure to the section flowcharts in the callflow HTML export. It extracts the node/edge selection logic into a shared select_diagram_content helper (used by both the flowchart renderer and the new coverage note), and adds a generate_diagram_coverage function that emits a visible <p> stating how many of a section's nodes and edges are actually drawn, returning empty when nothing is omitted. New CSS styling and two tests are included, and the coverage note is wired into write_callflow_html beneath each section diagram.

No blocking issues surfaced. 4 lower-confidence candidates did not survive cross-model review.

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 394 functions depend on the 168 functions this change touches.

Health — this change adds coupling hotspots:

• worse: write_callflow_html() — 7 callers, 26 callees

Verification — 394 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 264 function(s) in the blast radius were not formally verified this run

· 1 more finding(s) on lines outside this diff (see the check run).

[graphify-labs]

Graphify reviewed this change.

Worth a look — the grounded gate found no coupling regressions or blocking issues, but 1 advisory finding(s) below merit a look before merge.

---

Graphify review — findings

This PR adds a visible "coverage note" beneath each section flowchart in the generated call-flow HTML report, stating how many of a section's nodes and edges the capped diagram actually draws. It introduces a new select_diagram_content helper that factors out the node/edge selection logic previously inline in generate_section_flowchart, and a new generate_diagram_coverage function (plus a CSS .diagram-coverage style) that reuses that helper to render the note, emitting nothing when the diagram is complete. The surface area covers graphify/callflow_html.py (refactored selection logic, new functions, wiring into write_callflow_html, and a stylesheet rule) and two new tests in tests/test_callflow_html.py exercising the coverage output for capped and complete diagrams.

Worth a look

• Coverage note reports edge total after cap, understating omitted edges — graphify/callflow_html.py · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 404 functions depend on the 168 functions this change touches.

Health — this change adds coupling hotspots:

• worse: write_callflow_html() — 7 callers, 26 callees

Verification — 404 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 264 function(s) in the blast radius were not formally verified this run

· 1 more finding(s) on lines outside this diff (see the check run).