Capstone: A Complete, Working Flexible-Routing Site

Sam — the site's own admin — logs in and does one real, ordinary session's worth of work. Every step below draws on a specific earlier chapter of this course, closing with a chapter-attribution table and an honest scope note. Sam is the same recurring persona every one of this course's own five already-complete siblings reuses for its own capstone — this telling of the story is, chronologically, the most recently written of the six, even though it's the very first in the series' own numbering.

Step 1 — Opening the Site

Sam visits the site's own domain. nginx terminates HTTPS (Chapter 11's own certbot certificate) and proxies to the Next.js production server, kept running unattended under PM2.

Step 2 — Logging In

Sam enters the same admin credential the old PHP site always used. NextAuth's authorize() callback runs, and bcrypt.compare() verifies it against the stored $2y$-tagged hash with no special configuration — nothing was ever reset. A signed JWT session is established.

Step 3 — Building the Chain, Live

Through the hand-built admin interface — no free admin panel, no scaffold generator — Sam creates five real pages, one level at a time:

1. programming (top-level)
2. general-purpose-languages (under programming)
3. java (under general-purpose-languages)
4. fundamentals (under java)
5. chapter-1 (under fundamentals)

Five real levels deep — no longer the illustrative example this course opened with in Chapter 1.

Step 4 — Visiting It Live

Sam opens /programming/general-purpose-languages/java/fundamentals/chapter-1 directly. The [...path] catch-all route joins the five URL segments into one string and resolves the page in a single fullPath lookup — no five-level chase. PageShell renders the breadcrumb through Chapter 6's own include-based fix, one query for the whole ancestor chain rather than five separate ones, and the dark theme applies through the single global stylesheet import.

Step 5 — The Kanji Check

Sam creates one more page — 水 — under programming. It saves, routes, and renders through the exact same pipeline as every other page: no special-casing, no depth exception, no separate rendering path. Chapter 7's own verified finding holds: nothing in this pipeline ever slices a path string by index, so it never mattered whether the character needed one UTF-16 code unit or a surrogate pair.

Step 6 — A Safe Edit

Sam edits chapter-1's own title through the now-protected updatePageTitle — Chapter 8's mutation, Chapter 9's auth() check standing in front of it. Loading the page again immediately afterward shows the new title — Chapter 8's own updateTag call guarantees this, not just usually true.

Step 7 — Reorganizing: Paying Chapter 2's Own Deferred Cost

Sam decides java belongs under a new archive section instead. Through the searchable flat-list parent picker, Sam reparents it. movePage's own recursive CTE finds both real descendants — fundamentals and chapter-1 — in one query, and each one's own fullPath is recomputed correctly, in dependency order. The cost Chapter 2 named and deliberately deferred gets paid here, for real, against a real five-level chain — not a toy example.

Step 8 — The Delete Refusal

Sam tries to delete fundamentals, forgetting it still has chapter-1 beneath it. The database's own onDelete: Restrict constraint refuses the raw delete; Prisma surfaces it as P2003; Chapter 10's own try/catch converts it into a real, readable message — not a crashed request.

Chapter Attribution

Hands-On Exercises