tart_cargo
Curated cargo of TART radio astronomy packages for stimela 2.0. This allows the use of TART telescope tools as part of the stimela 2.0 radio astronomy pipeline.
Author: Tim Molteno
Based on the tart_tools package which does most of the heavy lifting.
Installation and Use
GitHub Codespaces (Easiest)
The easiest way to get started is using GitHub Codespaces, which provides a pre-configured development environment in your browser:
- Click the green "Code" button on the GitHub repository page
- Select the "Codespaces" tab
- Click "Create codespace on main" (or your desired branch)
The codespace will automatically set up Ubuntu with Python 3.12, Apptainer, and all required dependencies. Once ready, you can run tart-cargo directly in the terminal.
Local Installation
You will need a Linux or Windows computer with apptainer, and python installed.
Linux
Most linux has python preinstalled.. Packages are available for download from apptainer.org. On recent debian
apt install apptainer squashfuse fuse2fs gocryptfs
Windows
install the Windows Subsystem for Linux (WSL), open PowerShell as an administrator and run the command wsl --install. This will enable the necessary features and install the default Linux distribution, which is usually Ubuntu.
To install pip on WSL, first ensure that Python is installed by running sudo apt install python3. Then, install pip using the command ```sudo apt install python3-pip``.
Use of tart_cargo
then create a python virtual environment
python3 -m venv myvenv
source myvenv/bin/activate
pip install tart_cargo
This should install all the necessary dependencies (including stimela).
Then create a stimela recipe. An example is shown below...
#!/usr/bin/env -S stimela run -l
# This basic TART Stimela 2.0 processing pipeline
opts:
log:
dir: logs/log-{config.run.datetime}
name: log-{info.fqname}
nest: 2
symlink: log
_include:
- (tartcargo):
- tart2ms.yml
- tart-download-data.yml
- disko.yml
- disko-draw.yml
tart-image:
info: TART imaging using stimela
inputs:
tart:
dtype: str
required: true
info: "TART name (e.g. mu-udm)"
outputs:
ms:
dtype: MS
default: '{current.tart}.ms'
hdf:
dtype: File
default: =STRIPEXT(current.ms) + '-vis.hdf'
svg:
dtype: File
default: =STRIPEXT(current.ms) + '.svg'
assign:
api: 'https://api.elec.ac.nz/tart/{recipe.tart}'
steps:
download-hdf:
cab: tart-download-data
params:
api: =recipe.api
vis: true
file: =recipe.hdf
create-ms:
cab: tart2ms
params:
hdf: =steps.download-hdf.file
ms: =recipe.ms
clobber: true
single-field: true
rephase: obs-midpoint
disko-image:
cab: disko
params:
ms: =recipe.ms
HDF: '{recipe.tart}.h5'
healpix: true
fov: 170deg
res: 0.5deg
lasso: true
alpha: 0.006
l1-ratio: 0.02
project-image:
cab: disko-draw
params:
hdf: '{recipe.tart}.h5'
show-sources: true
SVG: =recipe.svg
You can run this recipe as follows:
stimela run example_recipe.yml tart=mu-udm
This will download data, and form an image from the specified TART telescope.
Metadata
Release files for tart-cargo 0.4.12
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| tart_cargo-0.4.12.tar.gz | 17.1 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| tart_cargo-0.4.12-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 30.5 kB
Release files / tart_cargo-0.4.12.tar.gz
| Download URL | tart_cargo-0.4.12.tar.gz |
|---|---|
| Size | 17.1 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
ba82b4ecac670d2e6f0ec9d3b80b811a8fa20d96bbac6ba886ca1a6fba193f1e
|
|
BLAKE2b-256 checksum How to use checksums |
69bb390e9afdb11df92f296fd3499c7a6bfe7092e005c21b9089aa9de6161963
|
| 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 Aug 30, 2026.
Transparency logRelease files / tart_cargo-0.4.12-py3-none-any.whl
| Download URL | tart_cargo-0.4.12-py3-none-any.whl |
|---|---|
| Size | 13.4 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
c5ce2af395ca749f17d7a0d2579e9f73c40e071e3d69cbd281a9b6bc793d43f6
|
|
BLAKE2b-256 checksum How to use checksums |
65fd4b87a983681f339052c3d5911c20842586ff226281764aa918db99a9fd4b
|
| 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 Aug 30, 2026.
Transparency log