In March, we reviewed the tables' release and identified numerous opportunities to improve performance and the general experience in Quadratic.
AI improvements
AI will now wait for code execution before executing its next step. This allows AI to immediately identify errors if they exist after code runs and attempt a fix immediately afterward.
Version history
You can now view the version history for all of your files. Prior versions can be created as new files.
Performance improvements
We drastically improved Python reference performance, enabling Python to reference more data faster.
q.cells() calls for Python are up to 500% faster, and JavaScript calls are up to 900% faster. Our benchmark was a 1M cell table that was brought down from roughly 8 seconds to 2 seconds in Python and 33 seconds to 1 second in JavaScript.
(We tested this on top-tier hardware. We're actively working on improvements so that average consumer hardware can have a more powerful experience in Quadratic with medium to large spreadsheets.)
Bug fixes
We're currently laser-focused on stability and performance in Quadratic. Crucial to stability is ensuring the sheet behaves as expected. We pushed dozens of bug fixes in March. Some of those are littered in the PR links below.
PR links
This is a non-exhaustive list from March.
Python performance
AI waits for code to execute
Version history
Fix chart refs from Formulas
Remove copy as PNG from HTML chart menus
Don't highlight table after copy as PNG
Table column wrap fix
- should go before $
Fix missing data when switching data types
Single finger mobile drag
Formula color change
Billing settings for non-owners
Connections flow improvement
GCP support
Preserve data table name after re-run
Optimized copy from contiguous-2D
Viewport improvements
Follow our roadmap on GitHub.
