Clojask

A Clojure dataframe library for larger-than-memory datasets. Process millions of rows with lazy evaluation and parallel execution—all with familiar dataframe operations.

Why Clojask?

Feature	Description
Larger than memory	Stream data from disk—no need to fit everything in RAM
Lazy & parallel	Operations are pipelined and executed across multiple threads
Relational operations	Filter, transform, group-by, aggregate, join, with arbitrary Clojure functions
File-to-file	Read CSV in, write CSV out, with built-in IO
Native types	Works with standard Clojure and Java types
Faster than Dask	Benchmarks show significant speedups on large datasets

Quick Start

Add to project.clj:

[com.github.clojure-finance/clojask "2.0.6"]

Or deps.edn:

com.github.clojure-finance/clojask {:mvn/version "2.0.6"}

Both also need two JVM flags, see Requirements.

Basic example (using Employees-example.csv):

(require '[clojask.dataframe :as ck])

;; Load a CSV
(def df (ck/dataframe "Employees-example.csv"))

;; Preview the data (the second row shows each column's type)
(ck/print-df df)
;; |         Employee |     EmployeeName |       Department |           Salary |       UpdateDate |
;; |------------------+------------------+------------------+------------------+------------------|
;; | java.lang.String | java.lang.String | java.lang.String | java.lang.String | java.lang.String |
;; |                1 |            Alice |               11 |              300 |       2020/12/12 |
;; |                2 |              Bob |               11 |              600 |       2020/12/01 |
;; ...

;; Set column types
(ck/set-type df "Salary" "double")
(ck/set-type df "UpdateDate" "date:yyyy/MM/dd")

;; Transform data: give Bob a raise, as a new column
;; (an operation over several columns must write to a new column)
(ck/operate df
  (fn [name salary]
    (if (= name "Bob") (+ salary 100) salary))
  ["EmployeeName" "Salary"]
  "NewSalary")

;; Compute with 8 threads, output to file
(ck/compute df 8 "results.csv"
  :select ["Employee" "EmployeeName" "Department" "NewSalary" "UpdateDate"])
;; results.csv:
;; Employee,EmployeeName,Department,NewSalary,UpdateDate
;; 1,Alice,11,300.0,2020/12/12
;; 2,Bob,11,700.0,2020/12/01
;; ...

Operations

Solid arrows show required sequence; dotted arrows show optional paths.

Available operations:

• Create: dataframe
• Types: set-type, set-parser, set-formatter
• Columns: operate, rename-col, select-col, delete-col, reorder-col, get-col-names
• Filter: filter
• Group: group-by, aggregate
• Combine: inner-join, left-join, right-join, outer-join, rolling-join-forward, rolling-join-backward
• Output: compute (with options such as :select, :exclude, :melt, :header), print-df, preview

sort stands outside this pipeline: it sorts the dataframe's source file on disk and writes the result directly to an output file.

Requirements

• OS: macOS or Linux
• JDK: 17 or newer (tested on 17, 21, and 25)

Required JVM flags:

;; 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.

Running multiple Clojask processes

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

Logging

Onyx writes warnings and errors to .clojask/clojask.log in the working directory, rotated at 10 MB with one backup.

Documentation

• API Reference — Full documentation for all functions
• Examples Repository — Real-world usage patterns
• Aggregation Functions — Built-in and custom aggregations
• Type System — Supported data types and parsing

How It Works

Clojask uses 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.

Issues & Feedback

Found a bug or have a question? Check the existing issues or open a new one.

License

MIT