We use mutation observers to detect and load pieces of html to process with bigpipe. We can let HTML do the job with a tiny onload attribute on an <object> element.
Add
<object onload="Drupal.bigPipeLoad(this.previousElementSibling)" height="0" type="text/plain" data="data:text/plain,"></object>
To trigger the processing of a pieces of bigpipe content instead of tying to guess when it's loaded though mutation events.
Show commandsStart within a Git clone of the project using the version control instructions.
Or, if you do not have SSH keys set up on git.drupalcode.org:
changes, plain diff MR !12224
Comments
Comment #3
nod_Comment #4
godotislateI was curious about the test failures in BigPipeRegressionTest, and I think a large contributing factor is that using load events instead of the mutation observer causes the page to load much more slowly.
Not scientific, but on my local ddev with standard profile and the big_pipe_regression_test module installed:
/big_pipe_test_multiple_replacementswas roughly 1m 13 sec/big_pipe_test_multiple_replacementswas roughly 14m 06 secI think this is happening because event handling is synchronous and blocking, while the mutation observer works asynchronously.
Comment #5
nod_I profiled the change and this is a terrible, terrible idea. it's 10x slower and uses 10x the memory.
It's broken in a way that can't be saved