Define a benchmark by binding a hashmap containing a function, arguments sequence, and group to a name.

• name is a symbol that labels the benchmark.
• group is a string, shared between multiple conceptually-related benchmarks.
• f is a 1-arity function that measures some performance aspect. Its single argument is the “n” in big-O notation. f may be supplied as an S-expression or a function object. Supplying f as an S-expression has the advantage that the definition may later convey some meaning, e.g., (fn [n] (+ n n)), in the html charts and tables, whereas a function object will render less meaningfully, e.g., #function[fastester.measure/eval11540/fn--11541].
• n is a sequence of one or more arguments. During performance testing, elements of n will be individually supplied to the benchmarking utility.

Example, supplying f as an S-expression:

(defbench add-two "benchmarking addition" (fn [n] (+ n n)) [1 10 100 1000])

Example, supplying f as a function object:

(defn my-fn-obj [q] (+ q q q))

(defbench add-three "benchmarking addition" my-fn-obj [2 20 200 2000])

Both examples above share the same group label (“benchmarking addition”), as both are concerned with measuring the performance of +. Putting them in the same group signifies that they are conceptually-related, and the html document will aggregate them under a single subsection. Also notice that the n sequences need not be identical.

Note: Invoking a defbench expression, editing the name, followed by invoking defbench a second time, defines two unique benchmarks. When developing at the REPL, be aware that the namespace may become ‘stale’ with orphaned or otherwise outdated benchmarks. Fastester will run benchmarks only explicitly associated to the :benchmarks options key, defined as the namespace exists in the file on the disk.

To undefine a single, unwanted test, use clojure.core/ns-unmap.