clojure-finance · Oct 20, 2026 · Oct 20, 2026 · Oct 20, 2026
diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
@@ -13,19 +13,19 @@ jobs:
       matrix:
         java: ["17", "21", "25"]
     steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-java@v5
+      - uses: actions/checkout@v7
+      - uses: actions/setup-java@v6
         with:
           distribution: temurin
           java-version: ${{ matrix.java }}
       - uses: DeLaGuardo/setup-clojure@13.7.0
         with:
+          # 2.13.0's bin/lein still says LEIN_VERSION 2.12.1-SNAPSHOT, so
+          # setup-clojure cannot install it
           lein: 2.12.0
-      - uses: actions/cache@v4
+      - uses: actions/cache@v6
         with:
           path: ~/.m2/repository
           key: m2-${{ runner.os }}-${{ hashFiles('project.clj') }}
           restore-keys: m2-${{ runner.os }}-
-      # Embedded ZooKeeper binds port 2188 and Aeron uses /dev/shm, so the
-      # suite must run in a single JVM; lein test already does that.
       - run: lein test
diff --git a/.gitignore b/.gitignore
@@ -1,5 +1,4 @@
 .DS_Store
-onyx.log*
 *.log
 resources/
 outputs/

diff --git a/README.md b/README.md
@@ -20,16 +20,14 @@ A Clojure dataframe library for **larger-than-memory** datasets. Process million
 
 **Add to project.clj:**
 ```clojure
-[com.github.clojure-finance/clojask "2.0.6"]
+[com.github.clojure-finance/clojask "2.1.0"]
 ```
 
 **Or deps.edn:**
 ```clojure
-com.github.clojure-finance/clojask {:mvn/version "2.0.6"}
+com.github.clojure-finance/clojask {:mvn/version "2.1.0"}
 ```
 
-Both also need two JVM flags, see [Requirements](#requirements).
-
 **Basic example** (using [Employees-example.csv](test/clojask/Employees-example.csv)):
 ```clojure
 (require '[clojask.dataframe :as ck])
@@ -90,44 +88,7 @@ Both also need two JVM flags, see [Requirements](#requirements).
 - **OS:** macOS or Linux
 - **JDK:** 17 or newer (tested on 17, 21, and 25)
 
-**Required JVM flags:**
-```clojure
-;; project.clj
-:jvm-opts ["--add-opens=java.base/jdk.internal.misc=ALL-UNNAMED"
-           "--enable-native-access=ALL-UNNAMED"]
-
-;; deps.edn, inside an alias (e.g. :aliases {:dev {...}}, run with -A:dev)
-:jvm-opts ["--add-opens=java.base/jdk.internal.misc=ALL-UNNAMED"
-           "--enable-native-access=ALL-UNNAMED"]
-```
-
-When running plain `java`, pass both flags on the command line.
-
-The first flag is required for Agrona (used by Onyx's messaging). Without it the first `compute` fails with `IllegalAccessError: class org.agrona.UnsafeApi ... cannot access class jdk.internal.misc.Unsafe`. The second is for lz4's native code: JDK 24+ prints a warning without it, and a future JDK will block the load.
-
-<details>
-<summary><strong>Running multiple Clojask processes</strong></summary>
-
-Each `compute` starts embedded ZooKeeper (port 2188) and Aeron (port 40200), with Aeron's files in the default media-driver directory (`/dev/shm` on Linux). To run multiple processes on one machine, give each different settings:
-
-**System properties:**
-- `clojask.zookeeper.port`
-- `clojask.aeron.port`
-- `clojask.aeron.dir`
-
-**Or environment variables:**
-- `CLOJASK_ZOOKEEPER_PORT`
-- `CLOJASK_AERON_PORT`
-- `CLOJASK_AERON_DIR`
-
-</details>
-
-<details>
-<summary><strong>Logging</strong></summary>
-
-Onyx writes warnings and errors to `.clojask/clojask.log` in the working directory, rotated at 10 MB with one backup.
-
-</details>
+Upgrading from 2.0.x: clojask no longer uses Onyx, so the JVM flags it needed (`--add-opens=java.base/jdk.internal.misc=ALL-UNNAMED`, `--enable-native-access=ALL-UNNAMED`) and the `clojask.zookeeper.*` / `clojask.aeron.*` / `CLOJASK_*` port and directory settings can be removed.
 
 ## Documentation
 
@@ -138,7 +99,7 @@ Onyx writes warnings and errors to `.clojask/clojask.log` in the working directo
 
 ## How It Works
 
-Clojask uses [Onyx](https://github.com/clojure-finance/onyx) (a maintained fork) as a single-machine thread pool. Operations are collected lazily and executed together when you call `compute`, streaming data through your transformation pipeline without loading the full dataset into memory.
+Operations are collected lazily and executed together when you call `compute`, streaming data through your transformation pipeline without loading the full dataset into memory. Each `compute` runs in your JVM: one thread reads the input in batches, a pool of worker threads (the number you pass to `compute`) applies the operations and filters, and the calling thread writes the results in input order. Group-by and joins first sort the rows into groups by key, kept in temporary files under `.clojask/` (or in memory with `:in-memory true`); group-by and outer joins write their results in group order.
 
 ## Issues & Feedback
 

diff --git a/docs/documentation.md b/docs/documentation.md
@@ -444,7 +444,7 @@ This means you cannot further apply complicated operations to a joined dataframe
 
 #### enable-debug / disable-debug
 
-Turn debug mode on or off. With debug mode on, operations are no longer checked against a preview when they are added, and logging is raised to `:debug`. Both take no arguments and print the new mode.
+Turn debug mode on or off. With debug mode on, operations are no longer checked against a preview when they are added. Both take no arguments and print the new mode.
 
 ```clojure
 (enable-debug)
@@ -460,10 +460,10 @@ Compute the result. The pre-defined lazy operations will be executed in pipeline
 | Argument            | Type                                                         | Function                                                     | Remarks                                                      |
 | ------------------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ |
 | `dataframe`         | clojask.classes.DataFrame.DataFrame / clojask.classes.JoinedDataFrame.JoinedDataFrame | The operated object                                          |                                                              |
-| `num of workers`    | int                                                          | The number of worker instances (except the input and output nodes) | Uses [Onyx](https://github.com/clojure-finance/onyx) as the distributed platform. Group-by and aggregate computes accept at most 8 workers. |
+| `num of workers`    | int                                                          | The number of worker threads (besides the thread that reads the input and the calling thread, which writes the output) | An integer from 1 to 8, for every kind of compute. |
 | `output path`       | String / `nil`                                               | The path of the output csv file                              | If the path already exists, will overwrite the file.<br>If `nil`, will store the output in memory as a vector of vectors, which represent each row, with the header row first unless `:header false`. See [example](https://github.com/clojure-finance/clojask-examples/blob/main/src/clojask_examples/in_memory.clj). |
 | [`exception`]       | Boolean                                                      | Whether an exception during calculation will cause termination | By default `false`. Is useful for debugging or detecting empty fields |
-| [`order`]           | Boolean                                                      | If enforce the order of rows in the output to be the same as input | By default `false`. If set to `true`, will sacrifice the performance. |
+| [`order`]           | Boolean                                                      | Has no effect since 2.1                                      | Accepted so that existing calls keep working. Output rows are always in input order, except that group-by and outer joins write theirs in group order. |
 | [`output`]          | Function                                                     | Specify how to output a row vector to the output file        | Takes two arguments.<br />`writer` java.io.BufferedWriter<br />`rows` clojure.lang.PersistentVector (rows) of clojure.lang.PersistentVector (each row) |
 | [`select`]          | String / Collection of strings                               | Chooses columns to select for the operation                  | Can only specify either of select and exclude                |
 | [`exclude`]         | String / Collection of strings                               | Chooses columns to be excluded for the operation             | Can only specify either of select and exclude                |
@@ -484,9 +484,6 @@ A `clojask.classes.DataFrame.DataFrame` read from the output file (read without
 (compute x 8 "output.csv" :select "col a")
 ;; only select column a
 
-(compute x 8 "output.csv" :order true)
-;; make sure the order of the output is the same of the input
-
 (compute x 8 nil :in-memory true)
 ;; compute the dataframe in memory and store the dataframe also in memory
 

diff --git a/project.clj b/project.clj
@@ -1,17 +1,13 @@
-(defproject com.github.clojure-finance/clojask "2.0.6"
+(defproject com.github.clojure-finance/clojask "2.1.0"
   :description "Data analysis and manipulation library with parallel computing for larger-than-memory datasets"
   :url "https://github.com/clojure-finance/clojask"
   :license {:name "MIT"
             :url "https://github.com/clojure-finance/clojask/blob/1.x.x/LICENSE"}
-  :dependencies [[org.clojure/clojure "1.11.1"]
+  :dependencies [[org.clojure/clojure "1.12.6"]
                 ;;  [org.clojure/math.numeric-tower "0.0.4"]
-                 [org.clojure/data.csv "1.0.0"]
-                 [com.github.clojure-finance/onyx "0.16.0"]
-                 [com.taoensso/timbre "5.2.1"]
                 ;;  [techascent/tech.ml.dataset "5.17" :exclusions [[ch.qos.logback/logback-classic][org.slf4j/slf4j-api]]]
-                 [com.google.code.externalsortinginjava/externalsortinginjava "0.6.0"]
-                 [com.github.clojure-finance/clojask-io "1.0.8"]
-                 [com.github.clojure-finance/clojure-heap "1.0.3"]]
+                 [com.google.code.externalsortinginjava/externalsortinginjava "0.6.6"]
+                 [com.github.clojure-finance/clojask-io "1.0.8"]]
   ;; lein deploy clojars, credentials from CLOJARS_USERNAME / CLOJARS_PASSWORD
   :deploy-repositories [["clojars" {:url "https://clojars.org/repo"
                                     :username :env/clojars_username
@@ -21,11 +17,7 @@
   :source-paths      ["src/main/clojure"]
   :java-source-paths ["src/main/java"]
   :javac-options ["--release" "17"]
-  ;; Onyx's Aeron messaging (via Agrona) reads jdk.internal.misc.Unsafe.
-  ;; lz4 (via Onyx's nippy serialization) calls System.load; JDK 24+ warns without
-  ;; native access enabled.
-  :jvm-opts ["-XX:+UseG1GC" "-server" "--add-opens=java.base/jdk.internal.misc=ALL-UNNAMED"
-             "--enable-native-access=ALL-UNNAMED"]
+  :jvm-opts ["-XX:+UseG1GC" "-server"]
   :test-paths        ["test/clojask"]
   ;:java-test-paths   ["test/java"]
   ;;:injections [(.. System (setProperty "clojure.core.async.pool-size" "8"))]

diff --git a/src/main/clojure/clojask/aggregate/aggre_input.clj b/src/main/clojure/clojask/aggregate/aggre_input.clj