fix(export): stop rewriting unchanged wiki/Obsidian pages on every run (#3060)

[abhay-codes07]

Fixes #3060.

What

to_wiki() and to_obsidian() regenerate the full page set on every call and write every page unconditionally:

• wiki.py deletes every *.md up front (for old in out.glob("*.md"): old.unlink()), guaranteeing all of them are then rewritten from scratch even when identical;
• export.py::_owned_write calls write_text_atomic(...) without comparing against what is already on disk.

An export is triggered by any graph.json change, so on an always-on watch --semantic + export setup it runs constantly. Measured by the reporter on one vault (26.7k nodes): 31,417 pages / 137 MB rewritten per export, 51 exports/day → ~7 GB/day, while the graph moved by ~20 nodes between runs. The disk cost is the smaller half — every rewrite is a modification event, so Obsidian re-indexes the whole vault, sync clients re-upload it, and inotify pipelines re-fire, dozens of times a day, for content that did not change.

Fix

Compare before writing:

1. paths.write_text_atomic_if_changed(path, text) -> bool — writes (and reports True) only when the content differs. It compares decoded text, not bytes: write_text_atomic writes in text mode and translates \n↔\r\n on Windows, so a byte compare would report every file as changed.
2. export._owned_write uses it and still records the note in _written — that list is both the ownership manifest and the prune-exclusion set (export obsidian: never prunes notes for removed nodes, so re-export merges old and new graphs #1896), so an unchanged note stays owned and is neither pruned as stale nor dropped from the manifest.
3. to_wiki drops the up-front unlink-all (which forced the rewrite and opened a window where a reader saw an empty wiki/ mid-export) for an end-of-run sweep that removes only the *.md this run did not produce — same orphan guarantee for relabelled/removed communities.

Verification

New tests:

• The helper skips an identical write (mtime untouched), and writes on a diff or a missing file.
• An Obsidian re-run leaves an unchanged note's mtime untouched yet still owned (not pruned), and still rewrites a note whose body changed (new neighbor).
• A wiki re-run leaves unchanged articles untouched and still sweeps a relabelled community's stale page.

The wiki / export / Obsidian / canvas / CLI-export suites stay green (100 in the focused sweep; 184 including the wider export set). The ownership and #1896 prune tests are unaffected.

🤖 Generated with Claude Code

https://claude.ai/code/session_01JJbLfztSxm2tH5cJwBbe9q

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[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).

Formal verification. 1 change(s) tested, no difference found (not proven).

---

Graphify review — findings

Adds write_text_atomic_if_changed, which compares decoded UTF-8 text against what's on disk and skips the atomic replace (preserving mtime/inode) when identical, treating a missing or non-UTF-8 target as changed — the byte-level rename churn that used to fire inotify/re-index/sync on every re-export is gone (#3060). Routes Obsidian's _owned_write and every wiki article/index write through it, so unchanged pages are left alone while still being recorded as owned so they aren't pruned. Replaces to_wiki's up-front unlink of all *.md with an end-of-run sweep that removes only files this run didn't produce, killing both the full rewrite and the mid-export window where readers saw an empty wiki/, while still cleaning up pages orphaned by drifting community labels.

No blocking issues surfaced. 10 lower-confidence candidates did not survive cross-model review.

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 2344 functions depend on the 105 functions this change touches.

Health — this change adds coupling hotspots:

• new: extract() — 655 callers, 45 callees
• new: _rebuild_code() — 142 callers, 54 callees
• new: build_from_json() — 218 callers, 20 callees
• new: build_merge() — 76 callers, 14 callees
• new: to_obsidian() — 41 callers, 14 callees
• new: save_semantic_cache() — 63 callers, 9 callees
• new: save_manifest() — 40 callers, 11 callees
• new: to_json() — 58 callers, 7 callees
• …and 58 more — each is listed as a finding

Verification — 2344 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: 1437 function(s) in the blast radius were not formally verified this run

Test selection

Test selection

115 of 291 test file(s) selected (40%) via static blast radius.

• tests/test_affected_cli.py — impact
• tests/test_agents_platform.py — impact
• tests/test_analyze.py — impact
• tests/test_atomic_canvas_export.py — impact
• tests/test_atomic_version_stamp.py — impact
• tests/test_atomic_writes.py — impact
• tests/test_benchmark.py — impact
• tests/test_benchmark_raw_graph.py — impact
• tests/test_build.py — impact
• tests/test_build_merge_dedup_scope.py — impact
• tests/test_build_merge_hyperedges_and_prune.py — impact
• tests/test_build_merge_shrink_guard.py — impact
• tests/test_cache.py — impact
• tests/test_callflow_html.py — impact
• tests/test_carried_hyperedge_remap.py — impact
• tests/test_charmap_encoding.py — impact
• tests/test_chunking.py — impact
• tests/test_cli_export.py — impact
• tests/test_cluster.py — impact
• tests/test_codebuddy.py — impact
• tests/test_community_labels_skill.py — impact
• tests/test_confidence.py — impact
• tests/test_corrupt_graph_json.py — impact
• tests/test_cpp_objc_cross_file_calls.py — impact
• tests/test_cross_extension_reexport_self_cycle.py — impact
• tests/test_dedup.py — impact
• tests/test_dedup_remaps_hyperedges.py — impact
• tests/test_definition_file_portability.py — impact
• tests/test_detect.py — impact
• tests/test_devin.py — impact
• tests/test_duplicate_annotation_edges.py — impact
• tests/test_evidence_binding.py — impact
• tests/test_explain_cli.py — impact
• tests/test_export.py — impact
• tests/test_export_control_characters.py — impact
• tests/test_export_idempotent_writes.py — impact, changed-test
• tests/test_export_path_length.py — impact
• tests/test_external_stub_endpoints.py — impact
• tests/test_extract.py — impact
• tests/test_extract_cache_location.py — impact
• tests/test_extract_cli.py — impact
• tests/test_falkordb_integration.py — impact
• tests/test_file_label_disambiguation.py — impact
• tests/test_global_add_tag_inference.py — impact
• tests/test_global_graph.py — impact
• tests/test_go_qualified_resolution.py — impact
• tests/test_god_nodes_cli.py — impact
• tests/test_god_nodes_exclude_hubs.py — impact
• tests/test_hollow_chunks_arm_shrink_guard.py — impact
• tests/test_hook_out_of_project_paths.py — impact
• … and 65 more

Selection is safe under the controlled-regression assumption; always-run tests + a periodic full run are the backstops. Advisory — it never changes the check verdict.

Formal verification

No difference found (not proven): No behavior difference found in to\_obsidian (not a proof).

The verifier ran both versions of to\_obsidian on many inputs and saw identical behavior every time. Strong evidence the change is safe, but evidence, not a proof.

Guarantee: Empirical: differential testing (both versions run on many generated inputs). A divergence on an untested input remains possible, so this is 'no counterexample found', not 'proven equivalent'.

Note: An input the sampler did not try could still differ.

Could not verify: Could not verify to\_wiki.

The verifier did not have enough to check to\_wiki, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: no capturable inputs from the test suite; property tier: parameter `output_dir` is annotated `str | Path` — outside the synthesizable primitive/collection set

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

[safishamsi]

Shipped in v0.9.67 (on PyPI). Cherry-picked with authorship preserved. Thanks @abhay-codes07! The decoded-text compare (not byte compare) to avoid the CRLF false-diff, and keeping skipped-but-owned pages out of the prune set, were both the right calls.