- Enable newNextLinkBehavior. See #36436
- Run next/link codemod on test suite
Note that from when this lands on apps trying canary will need to run
the new-link codemod in order to upgrade.
Ideally we have to detect `<a>` while rendering the new link and warn
for it.
Co-authored-by: Steven <steven@ceriously.com>
<!--
Thanks for opening a PR! Your contribution is much appreciated.
To make sure your PR is handled as smoothly as possible we request that
you follow the checklist sections below.
Choose the right checklist for the change that you're making:
-->
For context, making a test change to allow testing CI changes in
https://github.com/vercel/next.js/pull/40910 without approval each run.
@ijjk How's this? The string's not being snapshotted so it shouldn't
break any tests.
Such bots typically navigate websites using hard navigations (as they
crawl one URL at a time). Respectively, they do not benefit from
prefetches at all, while increasing the cost of both the crawl and
operating the site.
<!--
Thanks for opening a PR! Your contribution is much appreciated.
In order to make sure your PR is handled as smoothly as possible we
request that you follow the checklist sections below.
Choose the right checklist for the change that you're making:
-->
## Bug
- [ ] Related issues linked using `fixes #number`
- [x] Integration tests added
- [ ] Errors have helpful link attached, see `contributing.md`
This ensures we de-dupe inflight `getStaticProps` requests the same way we do `getServerSideProps`, this also removes the `_getStaticData` and `_getServerData` methods as they were repeating the same logic so could be combined into `fetchNextData`.
This PR proposes a fix for https://github.com/vercel/next.js/issues/11107 (JS modules are loaded twice). A more detailed explanation of the investigation that led to this PR can be found in the issue's comments (https://github.com/vercel/next.js/issues/11107#issuecomment-791780168).
## Replicability
To identify that the issue replicates on any given project, you need to
1. look at the network tab (first/clean load of site, so preferably ⌘+⇧+R on an incognito tab),
2. sort by "name", and filter requests by `mime-type:application/javascript` (selecting "JS" in the devtools filters will actually show all "script" types, but ignore all "javascript" types)
3. look for pairs of identical calls with one originating from initial HTML (`preload` of priority "high" originating from "(index)" or "([page name])") and another one from a script (`prefetch` of priority "lowest" originating from a .js file), where neither of the files is served from the cache.
Here's a screenshot of an example of what to look for:
<img width="601" alt="Screen Shot 2021-03-07 at 09 59 18" src="https://user-images.githubusercontent.com/1325721/110234627-cf1c6d00-7f2b-11eb-9cd7-749bf881ba56.png">
The issue was reproduced easily on the following projects:
- On [nextjs.org](https://nextjs.org/) where duplicates add up to ~70kB of transferred javascript out of 470kB (14.9%).
- On [vercel.com](https://vercel.com/) where duplicates add up to ~105kB of transferred javascript out of 557kB (18.8%).
- On [tiktok.com](https://tiktok.com/en) where duplicates add up to ~514kB of transferred javascript out of 1556kB (33%).
- In my own project using `"next": "^10.0.1"` (private repo) where duplicates add up to about 5% of total transferred javascript.
- In the issue's comments, a developer reported a replication using `"^10.0.7"` on a [public repo](https://github.com/SidOfc/sidneyliebrand.io).
## Some information about the fix
- Both `preload` and `prefetch` values for `<link rel="x">` behave similarly, with the difference being in network priority level (preload is high priority, prefetch is lowest priority).
- Next.js uses `<link rel="preload">` in its initial HTML but then *only* uses `<link rel="prefetch">` for the rest of the lifetime of the page.
- However, when Next.js detects that a script should be requested in advance, it only checks for matching `<link rel="prefetch">` and not `<link rel="preload">` (which have higher priority and are present earlier in the DOM, thus have a greater likelihood of being already loaded).
This PR aims to fix that oversight.
## Potential issues (none AFAIK)
As far as I can tell by looking through the codebase, **there is no downside** not to add a `prefetch` when a `preload` is already in the DOM. No other script looks for a `<link>` based on its `rel` attribute.
This applies the fix from the awesome investigation done in https://github.com/vercel/next.js/issues/28797 by @jayphelps and adds a test to ensure this is working as expected. It seems that the `route-loader` has a race condition while prefetching and if a script is executed before we have created a current "future" entry to resolve the entry stays in a pending state causing routes to hang so this handles the condition by ensuring pending/errored entries do not stay around.
## Bug
- [x] Related issues linked using `fixes #number`
- [x] Integration tests added
- [x] Errors have helpful link attached, see `contributing.md`
Fixes: https://github.com/vercel/next.js/issues/28797
Fixes: https://github.com/vercel/next.js/issues/27783
This ensures we handle resolve rewrites during prefetching the same way we do during a client-transition. Previously if a rewritten source was used in an `href` neither the page bundle or SSG data if needed would be prefetched although would work correctly on a client transition.
Fixes: https://github.com/vercel/next.js/issues/22441
This pull request speeds up Next.js' rendering pipeline by fetching data, parsing it, and loading it into memory instead of only doing the network request.
This will mainly result in improved Firefox/Safari performance since they handled prefetch incorrectly—only Chrome did it right. This also gets us closer to being able to use `no-store` in our caching headers!
---
Fixes#18639
x-ref #18802
Updates the way filenames are generated for browser compilation.
Notably:
- All entry bundles now have hashes in production, this includes pages (previously pages used a buildId in the path)
- The AmpFiles no longer depends on hardcoded bundle names, it uses the buildManifest instead (internals)
- All cases where we match the page name from the chunk/entrypoint name now use the same function `getRouteFromEntrypoint` (internals)
- In development we no longer include the "faked" `buildId` set to `development` for page files, instead we just use the `/_next/static/pages` path (was `/_next/static/development/pages`). This was changed as it caused unneeded complexity and makes generating the bundles easier (internals)
- Updated tons of tests to be more resilient to these changes by relying on the buildManifest instead of hardcoded paths (internals)
Follow up of these PRs:
https://github.com/vercel/next.js/pull/13759https://github.com/vercel/next.js/pull/13870https://github.com/vercel/next.js/pull/13937https://github.com/vercel/next.js/pull/14130https://github.com/vercel/next.js/pull/14176https://github.com/vercel/next.js/pull/14268Fixes#6303Fixes#12087Fixes#1948Fixes#4368Fixes#4255Fixes#2548
As discussed, this streamlines the handling for `basePath` to not automatically strip and add the `basePath` when provided to `next/link` or `router.push/replace` and only automatically adds the `basePath` and when it is manually provided it will cause a 404 which ensures `href` still matches to the pages directory 1-to-1.
This also adds additional test cases that we discussed to ensure this behavior is working as intended
---
Fixes#13902
* Optimize Prefetching
* fix css client nav test
* fix preload viewport test
* fix production test
* patch tests more
* Make page loader wait on prefetch
* no unhandled rejection
* Save some bytes
* Use intersection observer to preload
links that are in viewport
* Clean up names
* Check for window before returning observer
* update check for IntersectionObserver
* Remove duplicate IntersectionObserver checks
* Add test for prefetching links in viewport
