SQL that branches

Performance

These selected results come from a single-threaded comparison with DuckDB 1.4.4 running in-process through JDBC. The suite uses 10 million rows on an Intel Core Ultra 7 258V with JVM 25 and reports the median of 10 measured runs after five warmups.

Stratum wins 41 of 55 queries in the published 10-million-row, single-threaded suite. DuckDB wins on several sparse-selectivity filters, high-cardinality group-bys and global COUNT(DISTINCT). Full methodology and raw results in the benchmark docs.

Branch tables without copying them

st/fork creates an O(1) copy-on-write branch. It records a new root while leaving unchanged chunks shared. st/sync! persists a branch to storage. st/load restores any named branch. Pass column data as a table map to SQL queries, or register live storage-backed tables in the server with register-live-table!.

require('[stratum.api :as st],
  '[konserve.file-store :as fs],
  '[clojure.core.async :refer [<!!]])

;; Open storage and load the orders dataset (10M rows)
def store: <!!(fs/new-fs-store("/data/stratum"))
def orders: <!!(st/load(store, "orders"))

;; Fork in O(1); unchanged chunks remain shared
def experiment: st/fork(orders)

;; Persist the fork as a named branch that shares chunks with main
<!!(st/sync!(experiment, store, "experiment"))

;; Query both branches via SQL by passing column data as a table map
st/q("SELECT SUM(price * qty) FROM t", {"t" st/columns(orders)})
;; => {:SUM(price * qty) 4821903.40}   ← main branch unchanged

st/q("SELECT SUM(price * qty) FROM t", {"t" st/columns(experiment)})
;; => {:SUM(price * qty) 4401238.66}   ← experiment branch

;; Time-travel: load any historical branch from storage
def baseline: <!!(st/load(store, "orders-baseline"))
st/q("SELECT COUNT(*) FROM t", {"t" st/columns(baseline)})
;; => {:COUNT(*) 9847233}

(require '[stratum.api :as st]
         '[konserve.file-store :as fs]
         '[clojure.core.async :refer [<!!]])

;; Open storage and load the orders dataset (10M rows)
(def store  (<!! (fs/new-fs-store "/data/stratum")))
(def orders (<!! (st/load store "orders")))

;; Fork in O(1); unchanged chunks remain shared
(def experiment (st/fork orders))

;; Persist the fork as a named branch that shares chunks with main
(<!! (st/sync! experiment store "experiment"))

;; Query both branches via SQL by passing column data as a table map
(st/q "SELECT SUM(price * qty) FROM t" {"t" (st/columns orders)})
;; => {:SUM(price * qty) 4821903.40}   ← main branch unchanged

(st/q "SELECT SUM(price * qty) FROM t" {"t" (st/columns experiment)})
;; => {:SUM(price * qty) 4401238.66}   ← experiment branch

;; Time-travel: load any historical branch from storage
(def baseline (<!! (st/load store "orders-baseline")))
(st/q "SELECT COUNT(*) FROM t" {"t" (st/columns baseline)})
;; => {:COUNT(*) 9847233}

Yggdrasil coordinates branches across Datahike, Stratum and Proximum so an application can refer to a consistent set of SQL, Datalog and vector snapshots. Yggdrasil → · Dataset API docs →

Where Stratum fits

Datahike is the system of record for immutable transactions, Datalog and time travel. Stratum provides a columnar layout for SQL scans and analytics, including group-bys, joins and window functions.

Quick start

# Java 22+, no Clojure needed
java --add-modules jdk.incubator.vector \
     -jar stratum-standalone.jar \
     --index /data/orders.csv

# Or try the built-in demo tables
java --add-modules jdk.incubator.vector \
     -jar stratum-standalone.jar --demo

# Connect with any PostgreSQL client
psql -h localhost -p 5432 -U stratum

require('[stratum.api :as st])

;; Query with DSL
st/q(
  {:from {:price prices, :qty quantities},
   :where [[:> :price 100]],
   :group [:region],
   :agg [[:sum [:* :price :qty]] [:count]]})

;; Or with SQL
st/q(
  "SELECT region, SUM(price * qty), COUNT(*)\n       FROM orders WHERE price > 100 GROUP BY region",
  {"orders" {:price prices, :qty quantities, :region regions}})

(require '[stratum.api :as st])

;; Query with DSL
(st/q {:from {:price prices :qty quantities}
       :where [[:> :price 100]]
       :group [:region]
       :agg   [[:sum [:* :price :qty]]
               [:count]]})

;; Or with SQL
(st/q "SELECT region, SUM(price * qty), COUNT(*)
       FROM orders WHERE price > 100 GROUP BY region"
      {"orders" {:price prices :qty quantities
                 :region regions}})

How it works

SQL capabilities

• DML: SELECT, INSERT, UPDATE, DELETE, UPSERT (INSERT ON CONFLICT), UPDATE FROM, CREATE TABLE, DROP TABLE
• Aggregates: SUM, COUNT, AVG, MIN, MAX, STDDEV, VARIANCE, CORR, MEDIAN, PERCENTILE_CONT, APPROX_QUANTILE, COUNT(DISTINCT), FILTER clause
• Group-by: any number of columns, string and numeric, with HAVING
• Joins: INNER, LEFT, RIGHT, FULL with multi-column keys
• Window functions: ROW_NUMBER, RANK, DENSE_RANK, NTILE, PERCENT_RANK, CUME_DIST, LAG, LEAD, SUM/COUNT/AVG/MIN/MAX OVER with frame clauses
• Composition: CTEs (WITH), subqueries, IN/EXISTS, UNION/INTERSECT/EXCEPT
• Expressions: CASE WHEN, COALESCE, NULLIF, GREATEST, LEAST, CAST
• Date/time: DATE_TRUNC, EXTRACT, DATE_ADD, DATE_DIFF
• String: LIKE/ILIKE, UPPER/LOWER, LENGTH, SUBSTR
• Files: FROM read_csv('file.csv'), FROM read_parquet('file.parquet')
• Analytics: ANOMALY_SCORE, ANOMALY_PREDICT, ANOMALY_PROBA, ANOMALY_CONFIDENCE (isolation forest via SQL; online rotation for concept drift, docs)
• Other: EXPLAIN, SELECT DISTINCT, IS NULL/IS NOT NULL, LIMIT/OFFSET

Work with us

If you need help getting Stratum into production, we can help with integration, custom development, and support contracts.

Installation

; deps.edn (Clojure CLI)
; check https://clojars.org/org.replikativ/stratum for the latest version
{:deps {org.replikativ/stratum {:mvn/version "0.3.72"}}}

; JVM flags required (add to :jvm-opts or alias)
:jvm-opts

["--add-modules=jdk.incubator.vector" "--enable-native-access=ALL-UNNAMED"]

;; Leiningen, project.clj
;; [org.replikativ/stratum "0.3.72"]

; deps.edn (Clojure CLI)
; check https://clojars.org/org.replikativ/stratum for the latest version
{:deps {org.replikativ/stratum {:mvn/version "0.3.72"}}}

; JVM flags required (add to :jvm-opts or alias)
:jvm-opts ["--add-modules=jdk.incubator.vector"
           "--enable-native-access=ALL-UNNAMED"]

;; Leiningen, project.clj
;; [org.replikativ/stratum "0.3.72"]

Requires JDK 22+. Clojure 1.12+. Apache 2.0 license. Latest version on Clojars.