taoensso.tufte — com.taoensso/tufte 2.1.0-RC2

A simple, fast, monitoring profiler for Clojure/Script.

Usage: wrap+name interesting body exprs with the p macro. Then activate profiling of these wrapped exprs using the profiled or profile macros:

(profiled {} (p :my-fn (my-fn))) ; Returns [<body-result> <?pstats>] (profile {} (p :my-fn (my-fn))) ; Returns <body-result>, dispatches ; ?pstats to any registered handlers.

Extensive facilities are provided for compile-time elision and runtime filtering.

See the relevant docstrings for more info: p, profiled, profile, add-handler! ; Core API

(p [opts & body] [id & body]) ; e.g. (p ::my-pid (do-work)) (profiled [opts & body]) ; e.g. (profiled {:level 2} (my-fn)) (profile [opts & body]) ; e.g. (profiled {:level 2} (my-fn))

(add-handler! [handler-id ns-pattern handler-fn])

How/where to use this library: Tufte profiling is highly optimized: even without elision, you can usually leave profiling enabled in production (e.g. for sampled profiling, or to detect unusual performance behaviour). Tufte's stats maps are well suited to programmatic monitoring.

Abbreviations, etc.

• form id = pid = id given in p
• group id = gid = id given in profile

*min-level* clj/s

e/o #{0 1 2 3 4 5 6}

*ns-filter* clj/s

(fn [?ns] -> truthy).

-compile-ns-filter clj/s

-elide? clj

(-elide? level-form ns-str-form)

Returns true iff level or ns are compile-time filtered.

add-accumulating-handler! clj/s

(add-accumulating-handler! {:keys [ns-pattern handler-id]
                            :or {handler-id :accumulating}})

Experimental, subject to change!

Creates a new StatsAccumulator (and agent in clj), then registers a handler to accumulate profile output to the StatsAccumulator using the agent.

Returns the StatsAccumulator. Deref it to drain the accumulator and return {<group-id> <merged-pstats>}.

One common pattern is to deref the accumulator every n minutes/etc. to get a view of performance over the period, e.g.:

(defonce my-sacc (add-accumulating-handler! "*")) (defonce my-sacc-drainer ;; Will drain and print formatted stats every minute (future (while true (when-let [m (not-empty @my-sacc)] (println (format-grouped-pstats m)) (Thread/sleep 60000)))))

See also format-grouped-pstats,

add-basic-println-handler! clj/s

(add-basic-println-handler! {:keys [ns-pattern handler-id format-pstats-opts]
                             :or {ns-pattern "*" handler-id :basic-println}})

Adds a simple handler that logs profile stats output with println.

add-handler! clj/s

(add-handler! handler-id handler-fn)

(add-handler! handler-id ns-pattern handler-fn)

Use this to register interest in stats output produced by profile calls. Each registered handler-fn will be called as:

(handler-fn {:ns-str _ :level _ :?id _ :?data _ :pstats _ :pstats-str_ _})

Map args: :ns-str - Namespace string where profile call took place :level - Level e/o #{0 1 2 3 4 5}, given in (profile {:level _} ...) :?id - Optional group id, given in (profile {:id _} ...) :?data - Optional arb data, given in (profile {:data _} ...) :pstats - As in (second (profiled ...)). Derefable, mergeable. :pstats-str_ - `(delay (format-pstats pstats))

Error handling (NB): Handler errors will be silently swallowed. Please try/catch and appropriately deal with (e.g. log) possible errors within handler-fn.

Async/blocking: handler-fn should ideally be non-blocking, or reasonably cheap. Handler dispatch occurs through a 1-thread 1k-buffer dropping queue.

Ns filtering: Provide an optional ns-pattern arg to only call handler for matching namespaces. See compile-ns-filter docstring for details on ns-pattern.

Handler ideas: Save to a db, log, put! to an appropriate core.async channel, filter, aggregate, use for a realtime analytics dashboard, examine for outliers or unexpected output, ...

capture-time! clj/s

(capture-time! id nano-secs-elapsed)

(capture-time! pdata id nano-secs-elapsed)

Note: this is a low-level primitive for advanced users! Can be useful when tracking time across arbitrary thread boundaries or for async jobs / callbacks / etc.

See new-pdata for more info on low-level primitives.

chance clj/s

(chance p)

Returns true with 0<p<1 probability.

compile-ns-filter clj/s

(compile-ns-filter ns-pattern)

Returns (fn [?ns]) -> truthy. Some example patterns: "foo.bar", "foo.bar.", #{"foo" "bar"}, {:whitelist ["foo.bar."] :blacklist ["baz.*"]}

defnp clj/s≠ macro

clj

(defnp name doc-string? attr-map? [params*] prepost-map? body)

(defnp name doc-string? attr-map? ([params*] prepost-map? body) + attr-map?)

cljs

(defnp & sigs)

Like defn but wraps fn bodies with p macro.

defnp- clj/s≠ macro

clj

(defnp- name doc-string? attr-map? [params*] prepost-map? body)

(defnp- name doc-string? attr-map? ([params*] prepost-map? body) + attr-map?)

cljs

(defnp- & sigs)

Like defn- but wraps fn bodies with p macro.

fnp clj/s≠ macro

clj

(fnp name? ([params*] prepost-map? body) +)

(fnp name? [params*] prepost-map? body)

cljs

(fnp & sigs)

Like fn but wraps fn bodies with p macro.

format-grouped-pstats clj/s

(format-grouped-pstats m)

(format-grouped-pstats m
                       {:keys [group-sort-fn format-pstats-opts]
                        :or {group-sort-fn (fn [m]
                                               (get-in m [:clock :total] 0))}})

Experimental, subject to change. Takes a map of {<group-id> <PStats>} and formats a combined output string using format-pstats.

format-pstats clj/s

(format-pstats ps)

(format-pstats ps opts)

Formats given pstats to a string table. Accounted < Clock => Some work was done that wasn't tracked by any p forms. Accounted > Clock => Nested p forms, and/or parallel threads.

handlers_ clj/s

{<handler-id> <handler-fn>}

HandlerVal cljs

may-profile? clj/s

(may-profile? level)

(may-profile? level ns)

Returns true iff level and ns are runtime unfiltered.

merge-pstats clj/s

(merge-pstats)

(merge-pstats ps0)

(merge-pstats ps0 ps1)

Statistics are lossless unless data to merge are very large. Accuracy of total clock time depends on merge order:

• Time is exact if pstats are merged in order of increasing completion time (earliest to latest :t1).
• Otherwise time is estimated as the maximum possible clock time, ignoring any disjoint intervals (periods when clock was proceeding without any measured activity).

new-pdata clj/s

(new-pdata)

(new-pdata {:keys [dynamic? nmax] :or {dynamic? true nmax default-nmax}})

Note: this is a low-level primitive for advanced users! Returns a new pdata object for use with with-profiling and/or capture-time!. Deref to get pstats:

(let [pd (new-pdata) t0 (System/nanoTime)] (with-profiling pd {} (p :foo (Thread/sleep 100)) (capture-time! pd :bar (- t0 (System/nanoTime)))) @pd)

Dynamic (thread-safe) by default. WARNING: don't change this default unless you're very sure the resulting pdata object will not be concurrently modified across threads. Concurrent modification will lead to bad data and/or exceptions!

p clj/s macro

(p id & body)

(p opts & body)

Profiling spy. Always executes body, and always returns <body-result>.

When [ns level] unelided and profiling is active, records execution time of body.

Compile-time opts: :id - Form id for this body in stats output (e.g. ::my-fn-call) :level - e/o #{0 1 2 3 4 5} ; Default is 5

profile clj/s macro

(profile opts & body)

Always executes body, and always returns <body-result>.

When [ns level] unelided and [ns level when] unfiltered, executes body with profiling active and dispatches stats to any registered handlers (see add-handler!).

Handy if you'd like to consume/aggregate stats output later/elsewhere. Otherwise see profiled.

Compile-time opts: :level - e/o #{0 1 2 3 4 5} ; Default is 5 :dynamic? - Use multi-threaded profiling? ; Default is false :nmax - ~Max captures per id before compaction ; Default is 8e5 :when - Optional arbitrary conditional form (e.g. boolean expr) :id - Optional group id provided to handlers (e.g. ::my-stats-1) :data - Optional arbitrary data provided to handlers

Note on laziness: Please note that lazy seqs and other forms of laziness (e.g. delays) will only contribute to profiling results if/when evaluation actually occurs. This is intentional and a useful property. Compare:

(profile {} (delay (Thread/sleep 2000))) ; Doesn't count sleep (profile {} @(delay (Thread/sleep 2000))) ; Does count sleep

profiled clj/s macro

(profiled opts & body)

Always executes body, and always returns [<body-result> <?pstats>].

When [ns level] unelided and [ns level when] unfiltered, executes body with profiling active.

Handy if you'd like to consume stats output directly. Otherwise see profile.

pstats objects are derefable and mergeable:

• @pstats -> {:stats {:n _ :min _ ...} :clock {:t0 _ :t1 _ :total _}}
• @(merge-pstats ps1 ps2) -> {:stats {:n _ :min _ ...} :clock {:t0 _ :t1 _ :total _}}

Full set of :stats keys: :n :min :max :mean :mad :sum :p50 :p90 :p95 :p99

Compile-time opts: :level - e/o #{0 1 2 3 4 5} ; Default is 5 :dynamic? - Use multi-threaded profiling? ; Default is false :nmax - ~Max captures per id before compaction ; Default is 8e5 :when - Optional arbitrary conditional form (e.g. boolean expr)

Note on laziness: Please note that lazy seqs and other forms of laziness (e.g. delays) will only contribute to profiling results if/when evaluation actually occurs. This is intentional and a useful property. Compare:

(profiled {} (delay (Thread/sleep 2000))) ; Doesn't count sleep (profiled {} @(delay (Thread/sleep 2000))) ; Does count sleep

profiling? clj/s

(profiling?)

Returns e/o #{nil :thread :dynamic}.

pspy clj/s macro

(pspy & args)

p alias

refer-tufte clj

(refer-tufte)

(require '[taoensso.tufte :as tufte :refer [defnp p profiled profile]])

remove-handler! clj/s

(remove-handler! handler-id)

set-min-level! clj/s

(set-min-level! level)

Sets root binding of minimum profiling level, e/o #{0 1 2 3 4 5 6}. 0 => Enable all profiling. 6 => Disable all profiling.

set-ns-pattern! clj/s

(set-ns-pattern! ns-pattern)

Sets root binding of namespace filter. See compile-ns-filter docstring for details on ns-pattern arg.

stats-accumulator clj/s

(stats-accumulator)

Experimental, subject to change! Small util to help merge pstats from multiple runs or threads.

Returns a stateful StatsAccumulator (sacc) with:

• (sacc <group-id> <pstats>) ; Merges given pstats under given group id
• @sacc ; Drains accumulator and returns {<group-id> <merged-pstats>}

Note that you may want some kind of async/buffer/serialization mechanism in front of merge calls (e.g. an agent).

See also add-accumulating-handler!.

StatsAccumulator cljs

valid-min-level clj/s

(valid-min-level x)

valid-min-level? clj/s

(valid-min-level? x)

valid-run-level clj/s

(valid-run-level x)

valid-run-level? clj/s

(valid-run-level? x)

with-min-level clj/s macro

(with-min-level level & body)

Executes body with dynamic minimum profiling level, e/o #{0 1 2 3 4 5 6}. 0 => Enable all profiling. 6 => Disable all profiling.

with-ns-pattern clj/s macro

(with-ns-pattern ns-pattern & body)

Executes body with dynamic namespace filter. See compile-ns-filter docstring for details on ns-pattern arg.

with-profiling clj/s macro

(with-profiling pdata {:keys [dynamic? nmax] :or {nmax default-nmax}} & body)

Note: this is a low-level primitive for advanced users! Enables p forms in body and returns body's result.

See new-pdata for more info on low-level primitives.