No blocking findings
I found no actionable regression in the reviewed diff.
The broad change is justified by the measured reduction in JavaScript. SharedModule appears in 6 of 45 lazy application entries, down from 39. The routes most affected by the blanket import now load much less code.
The PR keeps component logic and templates intact. An AST comparison found that 271 changed files differ only in imports, standalone metadata, or comments. The other six files change module wiring or the monetize route entry. I checked those separately.
This is a code-review verdict, not a claim that every logged-in workflow passed. The independent checks below include compilation, bundle analysis, and limited server rendering. They do not include purchases, publishing, or an OTP login.
What the five commits change
Small pages
18 small pages remove SharedModule. Pages such as About and Pricing can load their own code without pulling in the shared module's large dependency tree.
Marketing, home, and search
MarketingPageSharedModule becomes a compatibility wrapper around standalone components. Its 57 exports stay intact. Home and search then import their own template dependencies.
Monetize
39 components become standalone. The app lazy-loads MONETIZE_ROUTES instead of MonetizeModule. Portfolio remains a nested lazy module. The text editor, AI dialog, and post-price dialog also remove SharedModule imports.
Auth, forum, resources, messages, insights, and post detail
These entries use explicit imports. PortfolioItemPreview imports the two music views it renders instead of importing all of PortfolioModule. Four music components become standalone to support that change.
What I checked in the code
Template dependencies remain available
Both versions pass Angular compilation. Of the 653 components present in both compiler outputs, 599 retain exactly the same matched directives and pipes. The other 54 lose only NativeElementInjectorDirective from ngx-intl-tel-input.
That library attaches the helper to ordinary form controls as well as phone inputs. None of these 54 templates contains ngx-intl-tel-input. I found no added or missing application directive or pipe among the common components.
Three unused store components leave the compilation graph
The base emits 656 components; the PR emits 653. The three missing outputs are the event-ticket, feed-content, and portfolio store-item components. Their classes remain in source, but no active TypeScript consumer or HTML template references them after this change. Their removal from the marketing wrapper does not remove a rendered view.
SharedModule keeps its exports and providers
All 160 exports remain. Its declaration list drops from 126 entries to 47, with the converted pieces moving into imports. The base's 126 entries include a duplicate CurrencyConversionModalComponent, so this is 78 unique declarations moved plus one duplicate removed.
The provider list and icon initialization remain unchanged. AppModule already calls provideSharedRootProviders, which supplies the shared providers and library configuration at the root. Removing a route's SharedModule import therefore does not remove those root providers.
Monetize preserves route behavior
The route array keeps the same paths, order, components, guards, data, redirects, and nested portfolio loader. Only its export and NgModule wrapper change. The app's monetize entry retains its authentication, permission, creator approval, and white-label owner guards. The redeem-ticket permission guard remains in the route array.
Independent QA bundle measurements
| Entry | Base MB | PR MB | Reduction |
|---|---|---|---|
| Home | 4.432 | 3.234 | 27.0% |
| About | 3.322 | 0.014 | 99.6% |
| Pricing | 3.316 | 0.008 | 99.8% |
| Marketing page | 4.393 | 2.161 | 50.8% |
| Auth | 3.497 | 0.996 | 71.5% |
| Forum | 3.439 | 0.809 | 76.5% |
| Resources | 3.393 | 0.898 | 73.5% |
| Messages | 5.415 | 2.713 | 49.9% |
| Insights | 4.779 | 2.480 | 48.1% |
| Post detail | 3.733 | 1.852 | 50.4% |
| Monetize | 5.365 | 3.207 | 40.2% |
The initial JavaScript grows by 4,730 bytes, from 7,135,079 to 7,139,809 bytes. Four entries that still load SharedModule grow by 5.5 to 39.8 KB. Link-in-bio and edit-preference still load it too, but their totals shrink.
The six remaining entries are profile, portfolio, dashboard, edit-preference, whitelabel-cover-page, and link-in-bio-fan-preview. This matches the PR's stated remaining work.
These are build-graph measurements, not network timings or compressed transfer sizes. The 45 entries include dynamically loaded components such as dialogs, so they should not be read as 45 distinct URLs.
Checks run for this review
| Check | Result |
|---|---|
| TopFans QA build at base and head | Pass. Client and server bundles compile. Both builds report 42 warning groups. |
| Generated ajwhite QA project at head | Pass. Reused the existing generated project and its replacement files without editing environment files. |
| Angular compiler at base and head | Pass. Compared the emitted template dependencies across both outputs. |
| AST and module comparison | Pass. Component bodies are unchanged. SharedModule and marketing-wrapper exports are preserved. |
| Eight logged-out SSR requests | Match. Both versions return HTTP 200, the same visible text, and the same ordered app-component tags. |
| Whitespace check | Pass. git diff --check reports no issues. |
The SSR requests covered /about, /pricing, /auth/login, /auth/forgot-password, /forum, /resources, /monetize, and /messages. About and Pricing rendered page content. The other six returned the common app shell in this server-only run. Those six checks do not validate the route's interactive content.
All builds used the required 10 GB Node heap limit. The builds disabled prerendering. I did not run the full prerender route list, a production configuration, or a fresh development build.
Build commands and comparison method
node --max_old_space_size=10048 node_modules/@angular/cli/bin/ng.js \
build TopFans --configuration qa --prerender=false --stats-json \
--output-path=/tmp/pr10837-astra-qaI ran the same command against base commit 16a5786e9c in a separate checkout and output directory. For white-label compilation, I replaced TopFans with the existing generated ajwhite project.
For template comparison, I ran ngc at each revision with a temporary config extending tsconfig.app.json. I normalized module aliases in the emitted component dependencies, then compared directives and pipes by source path and class name. This check measures what Angular matched in each template; it does not execute user interactions.
For bundle sizes, I read each build's stats.json, followed static JavaScript imports for each lazy entry, and subtracted the initial main, polyfill, and scripts chunks. The renamed monetize entry maps back to MonetizeModule for comparison.
What remains outside this review
The PR author reports logged-in white-label comparisons, an OTP login, dialog dependency checks, and broader SSR coverage. I have not independently rerun those sessions. Keep that evidence separate from the checks above.
Before release, retain a browser check for creating and editing a monetized item, opening post-price and text-editor dialogs, sending a message, and previewing a music portfolio. Exercise a phone input too. These are the interactions that compilation and an app-shell response cannot prove.
The author also reports an intermittent messages error when the QA API is slow and says it occurs on base. I did not reproduce it, so it is not a finding against this PR.