From now on, Swift-Colab will focus on more general usage of Swift, diverging from the S4TF project. It may not receive many new features anymore, but should stay actively maintained. The kernel was designed to work with future versions of Swift, making it usable even if it does go unmaintained. As a final safeguard, the extensive comments and documentation make patching future problems much easier for people besides myself.
I still find Swift-Colab quite useful for personal use cases, enabling persistently stored cloud notebooks geared toward interactive scripting. Plans to support local Jupyter notebooks are postponed indefinitely, as well as a "v3.0" version of Swift-Colab. However, liuliu/swift-jupyter provides a great experience for people seeking local Swift notebooks.
Changes:
- Numerous bug fixes.
- Shortened first cell of the template notebook from 3 to 2 lines.
- Added color to terminal output and refined stack traces for runtime errors.
- IPython and SwiftPlot images now render inline with text, making them appear before (instead of after) a cell finishes executing.
- You can now use Google Drive directly through PythonKit, without switching to Python mode.
- You can now enter command-line input to
%system
commands, making this feature on par with Python!
bash commands.
Not implemented:
%install-swiftpm-import
and%install-swiftpm-environment
magic commands, which would have sped up Swift for TensorFlow compilation considerably.- Conformance to the
JupyterDisplay
protocol fromliuliu/swift-jupyter
. %install-test
command for testing Swift packages inside a Colab notebook.- Support for specifying Swift v5.0-style package specifications.