work / 2022
CanvaSQL
A spatial SQL workspace. Query, viz, filter and dashboard share one board instead of being separate modes.
Direct-manipulation HoneySQL canvas that became CanvaSQL: generated queries, axis/dim tweaking, viz/query scrubbing, then the actual SQL. Metadata, sniff tests, and viz recos themselves defined in HoneySQL and stored as SQL — the tool browses itself. Folded into RVBBIT's data board.
Timeline
- HoneySQL micro-dashboards Procrastination project: dynamic SQL drag-and-drop on HoneySQL. The CanvaSQL ancestor, named out loud.
- HoneySQL workspace Drag-and-drop canvas, context-specific polymorphism. Nest and filter a parent to drive downstream subqueries.
- No modes Why query, then viz, then filter, then dashboard — when the canvas can do all of it at once?
- Visual recommendation browsing CanvasQL: generated query pipelines, recommendation browsing, smooth as butter.
- No canvas to spare My dashboard is packed — what's the point of sub-query lines if I can't see them? Overlay the lineage anyway.
- I don't want a notebook I just want to connect to my database and start looking at stuff. I got you. 127 likes — the pitch.
- The color and shape of data What is the Levenshtein distance between the system in your head and the one in front of you?
- A slightly different take on SQL drag-and-drop Been busy. Excited to get this out in August. More fun than I remember.
- CanvaSQL browses itself Metadata, sniff tests, viz recos — all HoneySQL, all stored as SQL tables.
- Fluid query visualizations Not a menu in an editor panel. Surface generated views in a way that feels like exploration.
- Ladder of abstraction Generated queries → dim/agg tweaking → viz/query scrubbing → hand-edit the SQL. Axis pickers are the *middle* rung.
- Pivot by another SQL query Visual SQL tools need explicit pivot-by-this. Bonus: pivoting by *another query*.
Sources: timeline.yaml