Timestamp profiles​

When we first started prototyping Rill Developer, we made an intentional choice to go with the quickest-possible dataviz primitives for our column profiles. They weren’t bad; you could see the general distribution and trends in numeric and timestamp columns, and our top-n values for varchar columns certainly gave us an impression of how powerful it is to suffuse a data transformation tool with basic exploratory data analysis.

But our quick-and-dirty data graphic prototypes were quite limited. Especially with the timestamp columns, which are a fantastic way to figure out what is going on with your dataset.

So in 0.3.0, we’re introducing a new timestamp column profile. It has the following features:

• Rolls up your timestamp column into an appropriate granularity based on heuristics – Our new visualization methodology allows you to build intuition around noise and outliers in your data at the speed of inquiry.

• Just as fast as the old timestamp profile code (which was fast) but 10x more useful – when we prototype new data visualizations, we take a lot of care to ensure that the output is fast. Not “data fast”, but “human fast”. These improvements come at no cost to performance and a snappy application.

And with this timestamp column profile, we also begin building out the data graphic primitives we’ll be using throughout the Rill product experience. Our team does in fact have quite a bit of experience with interactive data visualization, and we know the reality ~ we can certainly use a pre-existing library to generate our plots, but we know we will hit a customization wall at a certain point. So we are beginning with the end in mind.

SQL Editor enhancements​

Also included in this 0.3.0 release are a few improvements to the SQL editor:

• Smarter autocomplete suggestions:

  • Source table names are now included in the recommendations! And you can access a table’s column names with dot notation. E.g. myTable. will suggest myTable.col1. Note, in the future, we’ll add column suggestions without requiring the table prefix.

  • Fewer SQL keywords. Now, only the important and DuckDB-specific keywords are included in the recommendations. Previously, the long-tail of SQL keywords were flooding the option list.

• More intuitive keybindings – Enter no longer adds a pesky indentation to the new line. And you can now use Tab to accept an autocomplete suggestion.

• Improved aesthetics – more readable highlights, a clearer autocomplete tooltip, and a couple small design simplifications.