In 2009 the library was the thing to measure, so jquery-profile.js wraps jQuery itself instead of sampling the browser. It is a 238-line IIFE that has to be loaded after jQuery, because the first thing it does is take a snapshot: var handle = jQuery.event.handle, ready = jQuery.ready;. From then on every dispatched DOM event goes through a replacement handler that times it, and the profile is a tree — events, each holding the jQuery calls issued while it ran.

Method capture is a loop over the prototype: for (var method in jQuery.fn) wraps every method except init and stores { name, inLength: this.length, args, duration, outLength: ret.length }. An internal flag is set while a wrapper runs, so jQuery's own nested calls are not counted twice. jQuery.fn.init gets a special case that also records the selector string and the context node — the part that answers which selector was slow. The guard compares ret.selector.substr(0, test.length) != test against the literal source of the wrapper, so the profiler never records its own code as a selector.

Events nest, so the log keeps an eventStack and a curEvent pointer. Code that runs outside any handler — the usual top-level $(...).css(...) on load — lands in a synthetic bucket called inline at log[0]. A jQuery.readyList.unshift(...) entry adds a ready pseudo-event. Events that produced no methods and took ≤1 ms are discarded with log.splice(pos, 1).

The clock is (new Date).getTime() — whole milliseconds — so that drop rule follows directly: a single fast jQuery call often rounded to zero. The tool was never going to rank micro-optimisations; it ranked aggregates and outliers. Every record also carries FunctionMonitor.currentFunction as m.fn, which is how FunMon2's named application functions attach to the jQuery calls that invoked them.

The question js-meter answered — which selector, which handler, how many elements, how long — is a built-in now. Chrome DevTools' Performance panel records selector statistics: elapsed matching time per selector plus match attempts. performance.now() gives microsecond-scale timestamps where (new Date).getTime() gave integers, and the User Timing API (performance.mark / performance.measure) with PerformanceObserver lets a page publish its own named ranges without monkey-patching a library. Frame attribution moved from a hand-rolled JavaScript eventStack to the browser's own call trees.

The other half of js-meter — instrument in the page, POST JSON, aggregate server-side — survived as standard instrumentation. OpenTelemetry's browser instrumentation wraps fetch and XMLHttpRequest; Sentry's Browser SDK once shipped a dedicated jQuery AJAX integration. Both send spans to a collector, which is the same idea as $.post('/jsprofiler/jsonlogger.php', { json: …, type: 'jquery' }) — standardised, sampled, with a real backend instead of a Smarty template.

What did not survive is the hook. jQuery.event.handle was an internal of the 1.3-era event system and is gone from jQuery 1.9 onwards, so the wrapper no longer has anything to attach to. A 2026 version would instrument jQuery.event.dispatch or, better, leave the library alone and use the platform APIs. The trade — 1 ms clock, one row per jQuery call in MySQL, server-rendered reports — was reasonable for 2009 shared hosting, and is exactly what one would not build today.