Skip to main content

ddxdb

Calculus in SQL: the derivative of a query with respect to a table's columns, as JAX takes the derivative of a function, computed by the engine.

from ddxdb import Context

ctx = Context()
ctx.sql("CREATE TABLE w (i BIGINT, val DOUBLE) AS VALUES (0, 1.0), (1, 2.0)")
ctx.sql("""
WITH f AS (SELECT SUM(val * val) AS l FROM w)
SELECT i, val, grad(f, val) AS df_dval FROM w     -- 2·val, row by row
""")

In a SELECT over a table, grad(f, val) is, at each row, the derivative of the number the CTE f computes with respect to that row's val (w.val when the SELECT reads several tables). The query is ordinary SQL; nothing in it is labelled for ddx, and the one function ddx gives a meaning to is ddx_stop_gradient(x), JAX's lax.stop_gradient.

A gradient is what an optimizer steps along, so a query scoring anything (a model's loss, a curve's misfit, an energy) can be optimized in SQL: a step, params - lr * grad(f)(params), is one expression.

jvp and vjp in SQL

Forward mode is jvp(e, table.column, tangent), a column of the SELECT that computes e: its tangent at each row along tangent, a CTE or table shaped like table (its dims, then the column's tangent), or a number, the same at every row, as jax.jvp returns it beside the value. Along 1, it is the derivative of an expression row by row: SELECT x, jvp(sin(x), t.x, 1) AS dx FROM t is cos(x). Along a table:

ctx.sql("""
WITH h AS (
  SELECT x.sample, w.out, tanh(SUM(x.val * w.val)) AS val
  FROM x JOIN w ON x.inp = w.inp GROUP BY x.sample, w.out),
dw AS (SELECT inp, out, 0.01 AS val FROM w)
SELECT sample, out, val, jvp(val, w.val, dw) AS val_tangent FROM h
""")

One SELECT's jvp columns share a tangent; tangents along two tables are two SELECTs, and add. vjp(f, column, cotangent) is the reverse: in a SELECT over a table, the pull-back to each row of a cotangent shaped like the CTE f's output. Calls nest: f, e's SELECT, or a tangent can read another call, so a jvp of a CTE of gradients is a Hessian-vector product, in SQL.

This is the Python distribution of ddx, a thin wrapper over the ddx-ad engine.

Install

pip install ddxdb                  # the programs, for any engine
pip install "ddxdb[datafusion]"    # + Context and ddxdb.ad, on DataFusion
pip install "ddxdb[duckdb]"        # + ddxdb.duckdb, on DuckDB

Context, for DataFusion

A real SessionContext subclass whose .sql() runs grad first: every inherited method, property and constructor argument works unchanged. It lives in ddxdb.datafusion (a subclass needs its base class at import time) and is re-exported as ddxdb.Context, imported on first use, so import ddxdb needs no engine.

On a plain SessionContext, ddxdb.ad.sql(ctx, statement) does the same, and ddxdb.ad.sql_all runs several statements that take grad of one loss for the price of one backward pass.

A training loop runs the same statements on every batch. ddxdb.ad.prepare prepares them once, as jax.jit traces a function once:

step = ad.prepare(ctx, [step_w, step_b])
for batch in batches:
    ctx.deregister_table("batch")
    ctx.register_record_batches("batch", [batch])
    w, b = step()                       # what sql_all would return

Its first call computes the calls as sql_all does. Later calls run the same plans again, asking DataFusion for no plan of a query, so they skip what a call of sql_all repeats: about a quarter of a small GPT's training step. A table may change its rows, or be registered again, between calls. A call after a table's schema, a view or a setting changes prepares the statements again.

On DuckDB

ddxdb.duckdb.sql(con, statement) is the same on a DuckDB connection, and ddxdb.ad.sql accepts one too:

import duckdb
from ddxdb import duckdb as ad

con = duckdb.connect()
con.register("w", arrow_table)          # or a table, a view, a temp table
ad.sql(con, """
WITH loss AS (SELECT SUM(val * val) AS l FROM w)
SELECT i, val - 0.1 * grad(loss, val) AS val FROM w
""")

Programs, and how they compose

ddxdb.ad.grad, vjp and jvp give the programs underneath, and each takes SQL or a program, and the columns to differentiate with respect to (wrt): a column as "table.column", or a list of them, which one backward pass differentiates together. They compose: ad.jvp(ctx, ad.grad(ctx, loss, wrt), wrt) gives Hessian-vector products, checked against jax.jvp(jax.grad(f)). Both kinds of program (BackwardProgram, ForwardProgram) list the tables you register before they run (inputs, as InputTables) and the tables they leave (value and gradients, as OutputTables). ad.run(ctx, program, inputs) runs one on its inputs (a cotangent or tangents, as DataFrames or Arrow tables, in the order of program.inputs) and returns what it computed, value and gradients, as DataFrames, tangents named {column}_tangent; nothing stays on the context:

hvp = ad.jvp(ctx, ad.grad(ctx, "SELECT SUM(val * val * val) AS l FROM w", "w.val"), "w.val")
out = ad.run(ctx, hvp, [ctx.sql("SELECT i, 1.0 AS val FROM w")])
out.gradients[0]   # i, val (the gradient), val_tangent (H·v)

Other engines

ddxdb.grad_plan(plan_bytes, wrt) differentiates the serialized Substrait plan an engine's producer writes, and ddxdb.run(backend, program) runs the result on any object with four methods, select_all, returns_rows, materialize and drop_table (the ddxdb.Backend protocol); vjp_plan and jvp_plan are the same for the other two. They are ddx-ad's own Rust, the same code the Rust adapter runs. Pass namespace="__ddx_mine_" to get the same program from the same plan every time.

The SQL calls run on another engine too: ddxdb.rewrite(backend, statements) runs the programs the statements' grad, vjp and jvp calls need on a backend that also has plan_sql (SQL to a serialized Substrait plan), and returns the statements rewritten to read what they computed. Run them on the engine inside the with block; the tables they read are dropped when it ends:

with ddxdb.rewrite(backend, [step_w, step_b]) as rewritten:
    for sql in rewritten.statements:
        engine.sql(sql)

In a training loop, pass the same ddxdb.Programs() each step (ddxdb.rewrite(backend, statements, programs=programs)), and each program is built once and run again on later steps. A program is kept under the backend's plan_key(sql), a text that changes whenever the engine's plan of the query could (its tables' schemas, its views' definitions), so a backend without that method gets nothing kept. ddxdb.ad.sql and Context.sql keep programs for each context this way on their own. ddxdb.prepare(backend, statements) prepares them instead, as ddxdb.ad.prepare does: its run() returns a Rewritten like ddxdb.rewrite's.

What you can write

Inside the loss: + - * /; the chain rule for the trig / inverse-trig / exp / log / hyperbolic set plus abs and erf; power with a constant base or exponent; joins, GROUP BY, SUM, AVG, MAX, MIN, CASE, greatest and least.

ddx_stop_gradient(x) is x, held constant by differentiation, as JAX's lax.stop_gradient is. Context registers it, and so does ddxdb.duckdb, on the connection it's given, unless that already has one. On a plain SessionContext, call ddxdb.ad.register_stop_gradient(ctx). On DataFusion it keeps its argument's type. On DuckDB it takes DOUBLE, so a forward value computed through it on a REAL column is DOUBLE, though the gradient stays REAL.

A loss may call a function you register, such as erf, which ddx differentiates and neither engine defines. On DataFusion, register it with ctx.register_udf. On DuckDB, use ddxdb.duckdb.create_function(con, name, function, parameters, return_type, **kwargs), which takes con.create_function's arguments. ddx plans on a connection of its own, so it can't see a function registered with con.create_function alone. For a cursor to plan with those functions too, make it with ddxdb.duckdb.cursor(con).

import math
from ddxdb import duckdb as ad

ad.create_function(con, "erf", math.erf, ["DOUBLE"], "DOUBLE", side_effects=False)
ddxdb.supported_functions()               # ['abs', 'acos', 'asin', ...]

The unary functions ddx has a rule for, read from the engine rather than restated.

Errors are typed

An unsupported construct is always an error, never a silently wrong number — this is a numerical-correctness library, and a plausible-looking wrong derivative is the worst thing it could produce. The kind of failure is a class, so you can catch the one you can act on. All of them derive from ddxdb.DdxError:

  • UnsupportedExpression: no rule for something in the query.
  • NotScalar: grad of a query that is not a loss (one row, one column).
  • UnknownColumn: a column the loss does not read.
  • InvalidColumn: a column that cannot be differentiated (not a float, or in a table whose rows do not have unique dims).

Development

pip install maturin pytest
maturin develop --uv
python -m pytest tests/

Building compiles protoc from source for the substrait crate (it needs cmake and a C++ compiler), which takes a few minutes the first time.

License

Licensed under Apache-2.0, the same as the rest of ddx.

Metadata

Release files for ddxdb 0.4.10

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for ddxdb 0.4.10
File Size Uploaded Provenance Actions
ddxdb-0.4.10.tar.gz 429.1 kB With provenance Details for ddxdb-0.4.10.tar.gz

Built distributions (wheels)

Table of built distributions (wheels) for ddxdb 0.4.10
File Actions
ddxdb-0.4.10-cp310-abi3-win_amd64.whl CPython 3.10 abi3 Windows x86-64 Details for ddxdb-0.4.10-cp310-abi3-win_amd64.whl
ddxdb-0.4.10-cp310-abi3-manylinux_2_28_x86_64.whl CPython 3.10 abi3 Linux glibc 2.28+ x86-64 Details for ddxdb-0.4.10-cp310-abi3-manylinux_2_28_x86_64.whl
ddxdb-0.4.10-cp310-abi3-manylinux_2_28_aarch64.whl CPython 3.10 abi3 Linux glibc 2.28+ ARM64 Details for ddxdb-0.4.10-cp310-abi3-manylinux_2_28_aarch64.whl
ddxdb-0.4.10-cp310-abi3-macosx_11_0_arm64.whl CPython 3.10 abi3 macOS 11.0+ ARM64 Details for ddxdb-0.4.10-cp310-abi3-macosx_11_0_arm64.whl
ddxdb-0.4.10-cp310-abi3-macosx_10_12_x86_64.whl CPython 3.10 abi3 macOS 10.12+ x86-64 Details for ddxdb-0.4.10-cp310-abi3-macosx_10_12_x86_64.whl

Total release size: 46.5 MB

Release files / ddxdb-0.4.10.tar.gz

Download URL ddxdb-0.4.10.tar.gz
Size 429.1 kB
Tags Source
SHA-256 checksum
How to use checksums
3cfd78727d2f5a32e7ecc9432c5bdd2ec976309080e047a62d1169cf233380c5
BLAKE2b-256 checksum
How to use checksums
a80fd092c8df9ff41ff9e7e502a6b37681424fcbe4524d92dc2b7cae8b63c837
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 16, 2026.

Transparency log

Release files / ddxdb-0.4.10-cp310-abi3-win_amd64.whl

Download URL ddxdb-0.4.10-cp310-abi3-win_amd64.whl
Size 9.3 MB
Tags CPython 3.10 Windows x86-64 abi3
SHA-256 checksum
How to use checksums
a9c03165c09ca209f852721e5027ac311ce11737c733fc48f7e2035378bfa172
BLAKE2b-256 checksum
How to use checksums
215bb010701e40525e8102ca23aedb1e433f0cb8522bb9c86e2031751b674b1b
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 16, 2026.

Transparency log

Release files / ddxdb-0.4.10-cp310-abi3-manylinux_2_28_x86_64.whl

Download URL ddxdb-0.4.10-cp310-abi3-manylinux_2_28_x86_64.whl
Size 10.6 MB
Tags CPython 3.10 Linux glibc 2.28+ x86-64 abi3
SHA-256 checksum
How to use checksums
32d79afc4d2f577d0f4b5c7fbc07a51a62c8897cc4668c02aa8e13739dfed23a
BLAKE2b-256 checksum
How to use checksums
04c98a0d05230be841f8e370a3dd787af60cd5adcc918c51f53f25b263e68505
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 16, 2026.

Transparency log

Release files / ddxdb-0.4.10-cp310-abi3-manylinux_2_28_aarch64.whl

Download URL ddxdb-0.4.10-cp310-abi3-manylinux_2_28_aarch64.whl
Size 8.8 MB
Tags CPython 3.10 Linux glibc 2.28+ ARM64 abi3
SHA-256 checksum
How to use checksums
4d0d8fcde20dbf0214ec702fa24cc6dd13b18c263f8070f423c15aff4fb5cd23
BLAKE2b-256 checksum
How to use checksums
149a1da4b9b7da2c57b61d8c6e9a343dc2e8c43ff8488dca12357b42e66dba15
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 16, 2026.

Transparency log

Release files / ddxdb-0.4.10-cp310-abi3-macosx_11_0_arm64.whl

Download URL ddxdb-0.4.10-cp310-abi3-macosx_11_0_arm64.whl
Size 8.5 MB
Tags CPython 3.10 abi3 macOS 11.0+ ARM64
SHA-256 checksum
How to use checksums
80927158160b0caf0da6c8cfdaa77ad54cf5d817772b77a32283428d60a4e8f4
BLAKE2b-256 checksum
How to use checksums
143c5bb6d26dd9f38bdf9d4d606df9faa89bddb3af51676d9bbc2efcfc0d5f4b
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 16, 2026.

Transparency log

Release files / ddxdb-0.4.10-cp310-abi3-macosx_10_12_x86_64.whl

Download URL ddxdb-0.4.10-cp310-abi3-macosx_10_12_x86_64.whl
Size 8.7 MB
Tags CPython 3.10 abi3 macOS 10.12+ x86-64
SHA-256 checksum
How to use checksums
c4c1a6389bb6c1668fa2a2a6be837f264c8a817173a2064907dffc897838b817
BLAKE2b-256 checksum
How to use checksums
1099f6415afaca4dccb13346d7a421dbbaf8ded4aab5e8985d3e26284925f616
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 16, 2026.

Transparency log

Security for ddxdb 0.4.10

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

All 6 files in this release have been uploaded with provenance metadata.

Select from 6 files

Attestation for ddxdb-0.4.10.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 16, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page