IRx

IRx is the semantic-analysis, LLVM-lowering, and native-runtime layer of the ArxLang ecosystem. The PyPI distribution is pyirx; the Python import is irx.

Status: functional experimental backend. It supports substantial compiler and runtime flows but does not lower every ASTx node.

Responsibilities

analyze ASTx modules and attach resolved semantic sidecars
validate types, calls, imports, templates, classes, and FFI boundaries
lower supported nodes to LLVM IR with llvmlite
emit objects, link executables with Clang, and run built artifacts
report structured semantic, lowering, native-compile, and link diagnostics
activate native runtime features only when a compilation unit needs them

Native Apache Arrow runtime

IRx integrates Apache Arrow through native C++ runtime code and opaque C ABI handles. Current layers include:

primitive Arrow arrays and Arrow C Data interoperability
homogeneous fixed-width arrow::Tensor values
DataFrames backed by arrow::Table
Series backed by arrow::ChunkedArray
RecordBatch schemas, builders, nullable data, and Arrow IPC streams

The RecordBatch API interoperates with PyArrow and supports signed/unsigned integers, floats, booleans, UTF-8 and large UTF-8 strings, dates, timestamps, and times. pyarrow and arx-arrowcpp-sources provide the Arrow C++ build and link metadata.

Install

pip install pyirx

Native builds require an LLVM/Clang-compatible toolchain. Arrow runtime features require a C++ compiler.

Translate ASTx to LLVM IR

import astx

from irx.builder import Builder

builder = Builder()
module = builder.module()

body = astx.Block()
body.append(astx.FunctionReturn(astx.LiteralInt32(0)))
prototype = astx.FunctionPrototype(
    name="main",
    args=astx.Arguments(),
    return_type=astx.Int32(),
)
module.block.append(astx.FunctionDef(prototype=prototype, body=body))

llvm_ir = builder.translate(module)
print(llvm_ir)

translate() returns LLVM IR text without linking. Use build() to emit and link a native artifact, then run() to execute the built program.

RecordBatch API

The direct Python API lives in irx.record_batch. In a source checkout, build its standalone shared library before first use:

python -c "from irx.builder.runtime.record_batch import build_record_batch_shared_library; build_record_batch_shared_library()"

from irx.record_batch import (
    IrxColumnType,
    RecordBatchBuilder,
    RecordBatchSchema,
)

schema = RecordBatchSchema()
schema.add_field("id", IrxColumnType.INT32, nullable=False)

builder = RecordBatchBuilder(schema)
builder.append_int32(0, 42)
batch = builder.finish()

assert batch.get_int32(0, 0) == 42

batch.release()
builder.release()
schema.release()

Runtime features

Registered features include libc, libm, assertions, buffer, list, array, tensor, dataframe, and record_batch. Features may contribute external symbols, C/C++ sources, object files, libraries, and linker flags.

Boundaries

IRx does not parse Arx/AIX source and does not define high-level query or DataFrame language semantics. Arrow C++ stays behind the native runtime; LLVM does not encode Arrow object layouts directly.

Documentation: https://arxlang.org/irx/

License: Apache-2.0.

Metadata

Release files for pyirx 1.24.1

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

Source distribution (sdist)

Source distribution for pyirx 1.24.1
File	Size	Uploaded
pyirx-1.24.1.tar.gz	211.4 kB	Aug 5, 2026	Details

Built distribution (wheel)

Table of built distributions (wheels) for pyirx 1.24.1
File	Interpreter	ABI	Platform	Reset
pyirx-1.24.1-py3-none-any.whl 277.5 kB Aug 5, 2026	Python 3	none	any	Details

Total release size: 488.9 kB

Release files / pyirx-1.24.1.tar.gz

Download URL	pyirx-1.24.1.tar.gz
Size	211.4 kB
Tags	Source
SHA-256 checksum How to use checksums	`17b8613cc8027974559b1bdd6c1eaaca7efcf75e7b7e8c6a39a38772104b995d`
BLAKE2b-256 checksum How to use checksums	`16db99adcac45f0592b425e16f4a678f512a30c7050458c02511ee63131c82cf`
Upload date	Aug 5, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`poetry/2.4.1 CPython/3.10.20 Linux/6.17.0-1020-azure`

Release files / pyirx-1.24.1-py3-none-any.whl

Download URL	pyirx-1.24.1-py3-none-any.whl
Size	277.5 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`add96105982a3a6170a29c256e62097242fa18df1217d91bd99e1fbf27ff0bfa`
BLAKE2b-256 checksum How to use checksums	`9f0381079216a16e575e1c53025eeb57fe5816f2743fb9ea1f5cba976c439d0e`
Upload date	Aug 5, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`poetry/2.4.1 CPython/3.10.20 Linux/6.17.0-1020-azure`