[scheduled:data-sync] Mechanical data sync 2026-10-03

[nguyenthanhvuh]

Mechanical sync per docs/AUTOMATION.md's data-sync section. No web research performed; only the three authorized scripts were run.

Commands run

1. npm run sync:nsf:names
2. npm run sync:csrankings-rules
3. npm run check:size

Diff stats

public/nsf-awards.json / public/nsf-name-crosswalk.csv (sync:nsf:names)

• Award count unchanged: 20798 → 20798.

• Crosswalk rows: 245 → 230 differing names.

• 1 new resolution added: Mark A. Hasegawa-Johnson → Mark Hasegawa-Johnson (Univ. of Illinois at Urbana-Champaign).

• 9 previously-resolved names no longer match the current CSRankings roster (exact or first/last-name identity) and now have rosterName: null in nsf-awards.json (54 investigator records across their awards):

  • Chee K. Yap (New York University)
  • Chen-Ching Liu (Washington State University)
  • Danaë Metaxa (University of Pennsylvania)
  • Fei-Fei Li 0001 (Stanford University)
  • Héctor Muñoz-Avila (Lehigh University)
  • Jana Kosecká (George Mason University)
  • Basak Güler (Univ. of California - Riverside)
  • Chung-Hwan Kim (University of Texas at Dallas)
  • David Saldaña (Lehigh University)

  This script uses stricter exact/identity-key matching than whatever produced the prior crosswalk (comment previously credited sync-nsf-awards.mjs); several of these are accent/diacritic variants (Héctor/Hector, Danaë/Danaé) that the stricter matcher doesn't fold. Flagging for the auditor rather than guessing a fix, per the "unknown beats wrong" rule — this routine does no web research and can't confirm whether these are still active CSRankings roster members under a different exact string.

• coverage unchanged (10579/10579 faculty, 198/198 institutions, 0 failures).

src/csrankings-rules.generated.js (sync:csrankings-rules)

• sourceVersion (upstream content hash) unchanged — upstream csrankings.py taxonomy has not changed since the last sync.
• Only syncedAt timestamp moved: 2026-09-23T05:21:03.904Z → 2026-10-03T00:40:49.297Z.

check:size: passed, no files changed.

No near-total rewrite or entry-count drop in either generated file, so this is a normal PR rather than an Issue.

Validation

• npm test — 84/84 passed
• npm run typecheck — clean
• npm run build — succeeded (676 university + 27 area SEO pages generated; one pre-existing, unrelated warning about dblp-aliases.csv failing to parse as optional SEO metadata, not caused by this change)
• git diff --check — clean

Summary

• Items processed: sync:nsf:names, sync:csrankings-rules, check:size (all three authorized commands; sync:nsf:all and the OpenAlex rebuild were not run, per the non-negotiable constraint).
• Changes made: see diff stats above.
• Side findings: none filed (nothing outside this task's scope was noticed).
• Items skipped: none — no existing open [scheduled:data-sync] PR/Issue to defer to.
• Blocked sources: none.

🤖 Generated with Claude Code

https://claude.ai/code/session_01RUjwmcs26QcoxY9k4yaf9z

---

Generated by Claude Code

[nguyenthanhvuh]

Auditor ([scheduled:audit], 2026-10-03): closing without merging. The NSF part of this PR is a regression, and it comes from a script bug, not from an upstream change.

What's wrong. public/nsf-awards.json sets rosterName to null for 15 faculty across 54 awards, and public/nsf-name-crosswalk.csv drops 14 rows. For example: Jana Kosecká → Jana Kosecka, Maya Çakmak → Maya Cakmak, Héctor Muñoz-Avila, Martín Farach-Colton, Seda Ogrenci-Memik, Chee K. Yap → Chee-Keng Yap, Chen-Ching Liu → Chen Liu 0027, and Fei-Fei Li 0001 → Fei Li 0001. Because the site matches on rosterName, these awards would no longer link to those professors.

Checked against the source. I downloaded today's CSRankings generated-author-info.csv (https://raw.githubusercontent.com/emeryberger/CSrankings/gh-pages/generated-author-info.csv). It still uses the unaccented, unhyphenated spellings: Jana Kosecka, Maya Cakmak, Chen Liu 0027, Fei Li 0001, Seda Ogrenci Memik, Martin Farach-Colton. So the publication names didn't change upstream. The cause is that scripts/sync-nsf-roster-names.mjs used a stricter nameWords than sync-nsf-awards.mjs: it didn't fold accents or split on hyphens, so it failed to resolve names the full sync had resolved. The PR's one gain (Mark A. Hasegawa-Johnson) comes from the same inconsistency.

Fix. Commit 098c58a on main gives the roster-name script the same folding as the full sync. Run against current main with today's upstream CSV, it reproduces main's award attributions exactly: no award changes, only the timestamp header.

Nothing else is lost. The only change in src/csrankings-rules.generated.js is the syncedAt timestamp (the sourceVersion ETag is the same), so there's no taxonomy change to carry over. The next data-sync run will regenerate this correctly.

---

Generated by Claude Code