Clojure Trove

Trove

A minimal, modern logging facade for Clojure/Script. It's a lightweight alternative to tools.logging, designed for library authors who want rich logging without forcing users to adopt a specific backend.

Trove supports both traditional string-based logging and structured data-oriented logging, with rich filtering capabilities (by namespace, id, level, data, etc.). It's tiny (1 macro, 0 deps, ~100 loc) and works with Clojure, ClojureScript, GraalVM, and Babashka.

Setup

deps.edn:

com.taoensso/trove {:mvn/version "1.1.0"}

Leiningen:

[com.taoensso/trove "1.1.0"]

See https://clojars.org/com.taoensso/trove for the latest version.

Quick Start

Library author usage (emitting logs):

(ns my-library.core
  (:require [taoensso.trove :as trove]))

;; Traditional logging (string messages)
(trove/log! {:level :info, :msg "User logged in"})
(trove/log! {:level :warn, :msg "Connection retry attempt"})

;; Structured logging (data-oriented)
(trove/log! {:level :info
             :id :auth/login
             :data {:user-id 1234, :session-id "abc123"}
             :msg "User authenticated"})

Application user setup (choosing backend):

(ns my-app.core
  (:require
   [taoensso.trove :as trove]
   [taoensso.trove.console] ; or .telemere, .timbre, .mulog, .tools-logging, .slf4j
   ))

;; Use default console backend (prints to *out* or js/console)
;; Default is already set, no action needed

;; Or switch to a different backend
(trove/set-log-fn! (taoensso.trove.telemere/get-log-fn))

;; Or disable all logging
(trove/set-log-fn! nil)

Core Concepts

Trove is a facade - it provides a single logging API (trove/log!) that library authors use. Application users choose which backend handles those logs by setting trove/*log-fn*.

Key design features:

• Map-based API (same as Telemere)
• Automatic lazy evaluation of expensive data
• Backend-agnostic filtering
• Rich structured data support
• Zero runtime dependencies

Logging API

Basic Usage

The log! macro accepts a map of options:

(trove/log!
  {:level :info        ; Required: :trace :debug :info :warn :error :fatal :report
   :id    :user/login  ; Optional: keyword identifier for this event
   :msg   "User login" ; Optional: human-readable message
   :data  {:user-id 42} ; Optional: structured data map
   :error ex})          ; Optional: exception/throwable

Log Levels

Standard levels from least to most severe:

• :trace - Very detailed diagnostic info
• :debug - Debugging information
• :info - Informational messages (default)
• :warn - Warning messages
• :error - Error conditions
• :fatal - Critical failures
• :report - Special high-priority reports

Traditional vs Structured Logging

Traditional (message-focused):

(trove/log! {:level :info, :msg "Processing order #1234"})
(trove/log! {:level :error, :msg "Database connection failed", :error ex})

Structured (data-focused):

(trove/log! {:level :info
             :id :order/process
             :data {:order-id 1234, :user-id 567, :total 99.99}})

(trove/log! {:level :error
             :id :db/connection-failed
             :data {:host "db.example.com", :port 5432}
             :error ex})

Structured logging is preferred because:

• Retains rich data types throughout pipeline
• Easier filtering and analysis
• Faster (avoid premature serialization)
• Better suited for databases and analytics tools

Event IDs

Use keyword IDs to categorize events:

;; Namespace-qualified keywords recommended
(trove/log! {:id :auth/login, :data {...}})
(trove/log! {:id :payment/success, :data {...}})
(trove/log! {:id ::order-complete, :data {...}}) ; Auto-namespaced

;; IDs enable powerful filtering
;; - Filter by ID prefix: :auth/*
;; - Track specific events
;; - Build metrics and dashboards

Lazy Evaluation

Trove automatically delays expensive data, so backends can filter before paying computation costs:

;; This expensive call only runs if the log passes filtering
(trove/log! {:level :debug
             :data (expensive-computation)})

;; Use :let for shared bindings across lazy args
(trove/log! {:level :info
             :let [result (expensive-call)]
             :msg (format-result result)
             :data (transform-result result)})

The :let bindings are only evaluated if the log passes filtering.

Backend Configuration

Available Backends

Trove includes adapters for common backends:

;; Console (default) - prints to *out* or js/console
(require '[taoensso.trove.console])
(trove/set-log-fn! (taoensso.trove.console/get-log-fn))

;; Telemere - modern structured logging
(require '[taoensso.trove.telemere])
(trove/set-log-fn! (taoensso.trove.telemere/get-log-fn))

;; Timbre - popular Clojure logging
(require '[taoensso.trove.timbre])
(trove/set-log-fn! (taoensso.trove.timbre/get-log-fn))

;; μ/log - structured events
(require '[taoensso.trove.mulog])
(trove/set-log-fn! (taoensso.trove.mulog/get-log-fn))

;; tools.logging - Java interop
(require '[taoensso.trove.tools-logging])
(trove/set-log-fn! (taoensso.trove.tools-logging/get-log-fn))

;; SLF4J - Java interop
(require '[taoensso.trove.slf4j])
(trove/set-log-fn! (taoensso.trove.slf4j/get-log-fn))

Console Backend Options

The default console backend supports filtering:

;; Only log :warn and above
(trove/set-log-fn!
  (taoensso.trove.console/get-log-fn
    {:min-level :warn}))

Dynamic Backend Switching

Use binding for temporary backend changes:

;; Disable logging in tests
(binding [trove/*log-fn* nil]
  (run-tests))

;; Use custom backend in specific context
(binding [trove/*log-fn* my-custom-log-fn]
  (perform-operation))

Writing Custom Backends

Implement a function matching the *log-fn* signature:

(defn my-log-fn
  [ns coords level id lazy_]
  ;; ns     - String namespace, e.g. "my-app.utils"
  ;; coords - [line column] or nil
  ;; level  - Keyword: :trace :debug :info :warn :error :fatal :report
  ;; id     - Keyword or nil, e.g. :auth/login
  ;; lazy_  - Map or delayed map: {:keys [msg data error kvs]}

  ;; Force lazy_ to get the actual values
  (let [{:keys [msg data error kvs]} (force lazy_)]
    ;; Implement filtering
    (when (should-log? level id)
      ;; Perform logging side effects
      (send-to-backend {:level level :id id :msg msg :data data}))))

;; Configure it
(trove/set-log-fn! my-log-fn)

Key implementation notes:

• Force lazy_ to access :msg, :data, :error, :kvs
• Implement filtering before forcing to avoid expensive computation
• The log-fn is called synchronously - use async/threading for expensive work
• Return value is ignored

Advanced Options

Custom Namespace and Coordinates

Override the defaults:

(trove/log! {:level :info
             :ns "custom.namespace"
             :coords [100 50]
             :msg "Override defaults"})

Custom Log Function Per Call

Use a different backend for specific logs:

(trove/log! {:level :info
             :log-fn my-special-log-fn
             :msg "Uses custom backend"})

Custom Key-Value Pairs

Pass additional data to your custom log-fn:

(trove/log! {:level :info
             :msg "Custom event"
             :my-custom-key "value"
             :another-key 123})

;; Your log-fn receives these in the :kvs key
(defn my-log-fn [ns coords level id lazy_]
  (let [{:keys [kvs]} (force lazy_)]
    (println "Custom keys:" (:my-custom-key kvs))))

Common Patterns

Library Usage Pattern

As a library author, just use trove/log!:

(ns my-library.api
  (:require [taoensso.trove :as trove]))

(defn process-data [data]
  (trove/log! {:level :debug
               :id ::process-start
               :data {:count (count data)}})
  (try
    (let [result (do-processing data)]
      (trove/log! {:level :info
                   :id ::process-complete
                   :data {:processed (count result)}})
      result)
    (catch Exception ex
      (trove/log! {:level :error
                   :id ::process-failed
                   :error ex
                   :data {:count (count data)}})
      (throw ex))))

Your users control the backend without changing your code.

Application Setup Pattern

In your application entry point:

(ns my-app.main
  (:require
   [taoensso.trove :as trove]
   [taoensso.trove.telemere]
   [my-library.api :as lib]))

(defn -main []
  ;; Configure logging backend once
  (trove/set-log-fn! (taoensso.trove.telemere/get-log-fn))

  ;; All libraries using Trove now log to Telemere
  (lib/process-data [...]))

Conditional Logging

Use when expressions for conditional logic:

(when (dev-mode?)
  (trove/log! {:level :debug
               :data (expensive-debug-info)}))

;; Or use level filtering in the backend
(trove/set-log-fn!
  (taoensso.trove.console/get-log-fn {:min-level :info}))

Key Gotchas

1. Log-fn is synchronous: The *log-fn* runs on the calling thread. Implement async/backpressure for expensive operations in your log-fn.

2. Lazy evaluation: Values like :data and :msg may be wrapped in delay. Always force the lazy_ argument in custom log-fns.

3. Backend setup timing: Set *log-fn* before any logging occurs. Do this early in application startup.

4. Nil log-fn: When *log-fn* is nil, all logging noops. This is intentional - useful for disabling logs.

5. Map required: The log! macro requires a compile-time map. Variables won't work:

   ;; This works
   (trove/log! {:level :info, :msg "ok"})
   
   ;; This fails
   (let [opts {:level :info}]
     (trove/log! opts)) ; Compile error!
   

6. ClojureScript console: In ClojureScript, the console backend checks for js/console existence before logging.

When to Use Trove

Use Trove when:

• Writing libraries that need logging
• You want structured logging without committing to a backend
• You need rich filtering capabilities
• You want ClojureScript compatibility
• You prefer a data-oriented logging API

Don't use Trove when:

• Writing an application (use Telemere, Timbre, etc. directly)
• You need advanced features like log appenders, formatting, rotation (use a full backend)
• You have no logging needs (obviously)

Trove is specifically designed for library authors. Application developers should typically use a full-featured backend directly.

References

• GitHub: https://github.com/taoensso/trove
• API Docs: https://cljdoc.org/d/com.taoensso/trove/
• Clojars: https://clojars.org/com.taoensso/trove
• Slack: #trove on Clojurians Slack
• Related: Telemere (full-featured backend using same API)