More practical caching strategy for importHref #4200
Conversation
cdata
force-pushed
the
refactor-import-href
branch
from
b67685a
to
1b65a8d
Compare
src/standard/utils.html
Outdated
| document.head.appendChild(link); | ||
| } else if (link.__dynamicImportLoaded || !link.__dynamicImport) { |
There was a problem hiding this comment.
Rationale here is that imports that were loaded dynamically and finished, or weren't loaded dynamically to begin with, should fire a synthetic load event. This might fall apart for imports that were loaded outside of importHref, but aren't fully loaded yet by the time we get here. Thoughts?
src/standard/utils.html
Outdated
| imprt.addEventListener('error', errorListener); | ||
|
|
||
| if (optAsync) { | ||
| link.setAttribute('async', ''); |
There was a problem hiding this comment.
This conditional branch exists on its own to satisfy the constraints of a test. The test asserts that a link is imported async by checking for an async attribute on a link. If a link is imported synchronously at first, and then later imported asynchronously (as is the case in the test), then we must always set the async attribute even if we are relying on a "cached" (that is, already-done-loading) link tag for state.
Previously, importHref would cache failed results. This caused failures to be cached indefinitely, even when the failure was the result of conditions that are subject to change over time (such as poor network connectivity). This implementation no longer caches failures. Additionally, it relies on the link tag's presence in the <head> of the document as an indication that it is loading or has loaded, rather than keeping an object map of URLs to link tags.
cdata
force-pushed
the
refactor-import-href
branch
from
1b65a8d
to
48fb58a
Compare
* delay callbacks until HTMLImports.whenReady if it exists * only avoid re-creating an import if there is an existing import created via importHref (detected via `import-href` attr)
…e CustomElements v0 polyfill.
|
These changes have been incorporated in #4305, but this PR was not properly closed. |
Reference Issue
Fixes #3908
Previously, importHref would cache failed results. This caused failures
to be cached indefinitely, even when the failure was the result of
conditions that are subject to change over time (such as poor network
connectivity).
This implementation no longer caches failures. Additionally, it relies
on the link tag's presence in the of the document as an
indication that it is loading or has loaded, rather than keeping an
object map of URLs to link tags.