CHANGELOG.md

Change Log

All notable changes to this project will be documented in this file. This change log follows the conventions of keepachangelog.com.

[Unreleased]

Nothing yet.

[0.9.2] - 2026-04-29

Fixed

• Trace pretty-print bounds now apply to all collections, not only seqs. set-print-length! is the clearer public setter; set-print-seq-length! remains as a backward-compatible alias. Production stubs expose both names.
• Tap payloads now carry consistent timestamp metadata. :t remains a numeric millisecond timestamp, and new set-date-time-fn! lets callers customize the companion :date-time value for :form, :code, :frame-*, :fx-effect, and dbg tap payloads.
• Public docstrings for fn-traced, defn-traced, fx-traced, and defn-fx-traced now document :msg / :m consistently, including that one label is copied to every emitted body :code entry from a single invocation.
• parse-opts (the kw-style trailer parser shared by dbg / dbgn / fn-traced / defn-traced / fx-traced / defn-fx-traced / dbg-last and the tracing.runtime wrap-* macros) no longer silently terminates on an unrecognized option. The previous implementation was a (loop [...] (cond ...)) with no :else branch — passing a typo (e.g. (tracing/dbgn form :onec)) or an option from a future debux version this fork hadn't picked up caused the cond to return nil, dropping the prior accumulator and every option that came after the unknown token. The new :else clause logs a warning to *err* (clj) or console.warn (cljs) and recurs with (next opts), preserving every recognized option and surfacing the typo at macroexpansion time. Same silent-failure shape as the earlier :if → :condition mishap that motivated pinning each cond branch with a parse-opts-* deftest.
• wrap-sub! and wrap-handler! :sub now accept the full reg-sub argument surface: layer-2 computation fns, explicit signal-fn + computation-fn pairs, :<- input sugar, and 3-arity computation fns with dynamic vectors. Only the final computation fn is wrapped with fn-traced; input-signal wiring is preserved.
• Runtime-wrap path for reg-event-fx / reg-event-ctx handlers now produces :fx-effects per-key entries to match the source-edit (reg-event-fx :foo (fx-traced [...] ...)) trace surface. wrap-event-fx! previously expanded to fn-traced, which surfaces only the per-form :code payload; it now expands to fx-traced so 10x's Fx panels and re-frame-pair render the same per-effect breakdown for runtime-wrapped and source-instrumented handlers. wrap-event-ctx! gets the same treatment via a new :ctx-mode true opt on fx-traced / defn-fx-traced that points :fx-effects emission at (:effects ctx) instead of the body's full return — the per-effect breakdown the reg-event-ctx contract expects.
• Production stub for the runtime API. Two new files — tracing-stubs/src/day8/re_frame/tracing/runtime.cljc and the in-src counterpart src/day8/re_frame/tracing_stubs/runtime.cljc — close the gap that broke release builds (two-jar setup) for any caller requiring day8.re-frame.tracing.runtime. Stub wrap-handler! / wrap-event-fx! / wrap-event-ctx! / wrap-sub! / wrap-fx! macros compile to bare re-frame.core/reg-event-db (or sibling) with no fn-traced wrap; unwrap-* / wrapped? / wrapped-list / unwrap-all! become no-op runtime fns. Mirrors the fn-traced / fx-traced / dbg / dbgn stub pattern (commits e0bd687, bae1b0d, b3248af).
• Production stubs for the three runtime configuration knobs on day8.re-frame.tracing — set-tap-output!, set-print-seq-length!, reset-indent-level!. The dev-side ns re-exports these as (def x ut/x) from debux.common.util; consumers wire them at app boot. Without stubs, a release build picking up tracing-stubs/src/day8/re_frame/tracing.cljc (or the in-src src/day8/re_frame/tracing_stubs.cljc under :ns-aliases) failed at compile/load with an unbound-var error against any of the three names. Both stub files now carry no-op defns with matching arities; release callers see the same surface but no work happens. New stub-runtime-ns-has-every-public-var / stub-tracing-ns-has-runtime-config-knobs deftests in tracing_stubs_test.clj use ns-publics to pin the contract: every public var the dev-side day8.re-frame.tracing.runtime advertises must exist in the stub ns, and the three config knobs must exist on the parent stub ns.
• CLJ-side :skip-all-args-type classifier now lists use, mirroring the CLJS-side entry. Without this, (dbgn (use 'some.ns)) under JVM-CLJ fell through to the fn-call default and dbgn descended into the quoted-ns arg — extra trace noise at best, ill-formed classification for callers passing multiple symbol args. New regression test in test/day8/re_frame/debux/regression_test.clj pins the one-trace, no-descent contract.
• CLJ-side comment classification now matches CLJS and upstream debux: (dbgn (comment ...)) is :skip-all-args-type, emits one nil-valued trace for the whole form, and does not descend into the discarded comment body.
• Sink fns in common/util.cljc (send-form!, send-trace!, -send-frame-enter!, -send-frame-exit!, -emit-fx-traces!) now short-circuit on re-frame.trace/trace-enabled? BEFORE building the per-emit payload (form tidying, cond-> assembly, get-in over *current-trace*, fx-effect mapv). Previously the macroexpansion-time (when ...) gates plus merge-trace!'s own is-trace-enabled? check suppressed the trace MUTATION but left the payload-construction work running on every dispatch in dev profile / REPL / misconfigured-release builds where the live tracing namespace is loaded with trace-enabled? false. send-trace! keeps the documented set-tap-output! independence — its top gate is (or @tap-output? @#'trace/trace-enabled?), so tap> consumers still receive entries when tap is enabled even with trace-enabled? off.

[0.7.0] - 2026-04-27

The v0.7 line closes the remaining feature gaps relative to philoskim/debux's option surface and adds two re-frame-shaped extensions on top — function entry/exit markers and per-effect tracing for reg-event-fx handlers. Nine additive items, no breaking changes.

Consumers upgrading from 0.6.x: see docs/upgrading-to-v0.7.md for a per-feature recipe walkthrough and trace-stream contract notes.

Added

• :once / :o option on fn-traced / defn-traced / dbg / dbgn (commit 33225e8). Suppresses consecutive emissions whose (form, result) pair matches the previous one. Per call-site identity (gensym'd at expansion); state survives across handler invocations until the result actually changes. Useful for high-frequency dispatches where you only want to see what's NEW. Composes with :if cleanly. Public reset via day8.re-frame.tracing/reset-once-state!.
• :verbose / :show-all option on fn-traced / defn-traced / dbgn (commit 0177254). Wraps leaf literals (numbers, strings, booleans, keywords, chars, nil) that the default zipper walker skips for noise reduction. Special-form skips (recur, throw, var, quote, catch, finally — the :skip-form-itself-type set) STAY honoured because instrumenting them corrupts evaluation semantics. Resolves the :skip-classification open question from docs/v0.6-roadmap.md Pair A.
• :final / :f option on fn-traced / defn-traced / dbg / dbgn / fx-traced / defn-fx-traced (commit 8eeadf6). When set, only the outermost (indent-level 0) :code entry is emitted per top-level wrapping form — every nested per-form entry is suppressed. Useful when you only care about the final value of each handler-body expression and the per-step zipper trace is noise. Composes with :if / :once / :msg.
• :msg / :m option on dbg / dbgn / fn-traced / defn-traced / fx-traced / defn-fx-traced (commit b0a68b9). Attaches a developer-supplied label to each emitted :code entry so consumers can distinguish output from many parallel call sites. Per-call dynamic — the value is evaluated at trace time, not macroexpansion.
• dbg-last macro — thread-last-friendly counterpart to dbg (commit 304ef11). Sits in the natural ->> step slot: (->> coll (filter pred?) dbg-last (map xf)). Value-transparent; same payload schema and opts (:name / :locals / :if / :once / :msg / :tap?) as dbg.
• tap> output channel for trace records (commit 18b06dc). New day8.re-frame.tracing/set-tap-output! toggle plumbs every send-trace! payload through tap> in addition to its normal in-trace sink, so any add-tap consumer (custom 10x panels, REPL probes, eval-cljs scripts) sees the same processed :code records. Default off — preserves the existing trace-only behaviour. Companion to the :tap? per-call opt on dbg / dbg-last.
• Function entry/exit markers — :trace-frames tag emitted by fn-traced / defn-traced (commit 8ba53a8). Each invocation produces a paired {:phase :enter :frame-id …} / {:phase :exit :frame-id … :result …} marker bracketing the existing :code payload, so consumers (10x Code panel, custom inspectors) can pair the markers and bound the intermediate :code entries. Off-trace: markers are silently dropped (no tap> fallback — frames are framework-level boundary info, not user-visible data). Exception path: only :enter is guaranteed; a missing :exit is the signal that the body threw.
• fx-traced / defn-fx-traced macros for reg-event-fx handlers (commit bae1b0d). Inherits all of fn-traced's per-form :code emission, frame markers, and :locals / :if / :once / :verbose opts; ALSO emits one :fx-effects entry per key in the returned effect-map ({:fx-key k :value v :t ms}). Resolves the dbgn.clj:353 "trace inside maps, especially for fx" TODO without modifying the zipper walker. Production stubs in both tracing_stubs namespaces strip opts and compile to bare fn / defn.
• Classifier parity with upstream debux — closes the four confirmed gaps (commit b471026). New :skip-all-args-type category: forms whose args carry compile-time semantics (protocol impls, reify method bodies, comment, var / quote / import, etc.) get one trace at the top level with no descent into the args — strictly more informative than :skip-form-itself-type. condp's test/expr pairs are now correctly walked, extend / extend-protocol / extend-type are recognised, and the previously-missing forms in :skip-fn-ret-type / :skip-fn-args-type got added.

Internal

• +debux-trace-id+ runtime binding alongside the existing +debux-dbg-opts+ / +debux-dbg-locals+ (commit 33225e8). Each dbgn / dbgn-forms / mini-dbgn expansion bakes in a unique gensym'd string used as the per-call-site identity for :once dedup.
• mini-dbgn (test-only) uses a fixed "mini-dbgn" trace-id instead of a gensym so the macroexpansion-shape assertions in dbgn_test.clj stay byte-stable.
• Runtime API coverage — wrap-event-fx! / wrap-event-ctx! / wrap-sub! and wrap-handler! :event each get a dedicated deftest. Two unit-level tests pin the chain-builder choice (reg-event-db) and the :sub-kind reaction-deref → :code emission path; two integration tests dispatch through reg-event-fx and reg-event-ctx chains end-to-end. Closes the gap that previously left three of the four runtime-API macros without automated guard.

[0.6.3] - 2026-04-27

Fixed

• Issue #40 — fn-traced / defn-traced no longer hang on loop/recur. Two-part fix: recur carve-out in :skip-form-itself-type (commit db9b7de) and o-skip? predicate now compares against the correct fully-qualified symbol (commit 48de2e8).
• remove-skip's o-skip? handler now re-evaluates the new loc instead of descending into the replaced form (commit 8d4e331).
• Documentation images restored: doc/img/cursive-{1,2,3}.png re #38 (commit ff5122a).
• Test-selector parity between lein test and cognitect.test-runner — ^:failing and ^:integration tests are excluded by default in both runners (commit 4eee8a3).
• defn-traced* now propagates :docstring and :meta to the resulting var (commit a527e6a). ::ms/defn-args had been conforming both fields all along, but defn-traced* only spliced :name and :bs, silently dropping any leading docstring or attr-map. Result was a traced handler whose var had no :doc / :added / :tag meta even though the source looked correct.

Added

• dbg single-form tracing macro for inline expression debugging (commit 992fd28).
• Runtime API: day8.re-frame.tracing.runtime/wrap-handler! and unwrap-handler! for ephemeral fn-traced wrap-and-restore at the REPL (commit 4ed07c9).
• runtime-api? feature-detection predicate for downstream consumers (commit 6b04e6b).
• :locals and :if options to fn-traced / defn-traced (commit 4d6e507).
• Production-mode loud-fail check — macro-time assertion plus console.warn when goog.DEBUG=false (commit 10d27fd).
• :code payload schema documented in send-trace! docstring (commit 6b4f042).
• End-to-end integration-test scaffolding (4 deftests in test/.../integration_test.clj, commits 6e4f5d1 + 8d4e331).

Changed

• Toolchain modernised: deps.edn + bb.edn + shadow-cljs.edn at the project root; CI/CD workflows ported to bb tasks (commits c940967 + a1fc73e + acea1d0 + 4109a0c).
• Core dependency pins bumped to current versions (commit ee768f2).
• project.clj retained as a transitional shim through the v0.6.x line; deletion targeted for v0.7.

Internal

• docs/v0.6-roadmap.md — milestone scope as a living document (commit 9d9fc3c).
• docs/improvement-plan.md — long-term roadmap synthesis (commit 20ac8f4).
• docs/v0.6-cd-dry-run-report.md — release-pipeline validation findings (commit f35e071).

[0.6.1] / [0.6.2] - 2021-03-11

Both tags resolve to the same commit (ca70cb9); the "Unreleased" section was never promoted at release time.

Fixed

• Fix deployment of tracing-stubs.

Changed

• Upgrade Clojure to 1.10.3.
• Upgrade shadow-cljs to 2.11.22.
• Upgrade lein-tach to 1.1.0.
• Upgrade Karma to 6.2.0.

[0.6.0] - 2020-05-16

• Changes.

[0.5.5] - 2020-05-01

Fixed

• Fix require-macros in tracing-stubs ns for shadow-cljs compatibility. Thanks to @thheller for spotting the bug.

[0.5.4] - 2020-04-30

Fixed

• Fix shadow-cljs compatibility by making the day8.re-frame.tracing-stubs ns available in the main package. Thanks to @thheller. See #30.

Changed

• Upgrade ClojureScript to 1.10.748
• Upgrade shadow-cljs to 2.8.109
• Upgrade lein-shadow to 0.1.7
• Upgrade karma to 4.4.1
• Upgrade zprint to 0.5.1
• Upgrade eftest to 0.5.9
• Upgrade re-frame to 0.12.0

[0.5.3] - 2019-07-18

Breaking Changes

• Deletes obsolete lazy-seq? fn in debux.common.util. If for some reason you are using this function in your own code, it is no longer available so you will need to copy it to your own source. Re PR 26

Fixed

• shadow-cljs warning 'Use of undeclared Var debux.common.util/clojure' re PR 26

[0.5.2] - 2019-07-18

Fixed

• figwheel.main compatibility fix re #28 @danielcompton

Changed

• Upgraded Clojure to 1.10.1
• Upgraded ClojureScript to 1.10.520
• Upgraded clojure-future-spec to 1.9.0
• Upgraded re-frame to 0.10.8
• Upgraded zprint to 0.4.16
• Upgraded eftest to 0.5.8
• Upgraded io.aviso/pretty to 0.1.37

Improved

• Documentation fixes and improvements. @danielcompton

[0.5.1] - 2018-04-05

Added

• Captured traces now include indenting information. This lets you better see the order and nesting of your function calls.

Changed

• Don't rename symbol gensyms with three numbers or less, (e.g. s3 is no longer renamed to s#).

[0.5.0]

Initial release after fork from debux.

Added

• Closure-define to enable/disable tracing. Defaults to disabled, and is removed by Closure dead code elimination. To enable/disable tracing, set your compiler options to :closure-defines {"debux.cs.core.trace_enabled_QMARK_" true}