Improve SQL Cell table view

[domoritz]

Is your feature request related to a problem? Please describe.

The table view requires manual scrolling and looks overall not as polished (e.g. numbers are not right aligned).

Describe the solution you'd like

Inputs.table https://observablehq.com/@observablehq/input-table automatically loads more data and looks overall more polished. However, there are some things that are better in the SQL table: row numbers, better displays of nulls, and column sizing.

[domoritz]

The accuracy of numbers is also a bit excessive in many examples and it looks like the default font is not optimized for viewing numbers.

[mbostock]

👍 There’s definitely room for improvement in the current SQL cell table design. We’re working on these improvements and it should be out in the near future. I would add the display of dates to the list. Any other specific suggestions you want to make would be welcome!

[mootari]

Another problem with the current implementation is, that the result table is not available in embedded notebooks. Instead, only the default Inspector output is shown.

Example: https://observablehq.com/embed/@observablehq/sql-cells?cell=*

In Observable:

Embed:

[asg017]

Some other improvements I'd like to see on the SQL cell table view (I can file a new issue if needed):

• Ability to add warning messages to table. For example, in Datasette, only the first 1,000 rows are returned, and I'd like to add a warning that says "results possibly truncated" so that authors know if the data is incomplete.
• Hooks to add metadata about a query. For example, adding a message for how long a query took (1.24 seconds), which service executed the query, when the query started, etc.
• Show loading message when querying. If I have a query I'm iterating on that takes >1 second, I can't tell if the results are stale from a previous run, or if my query is just taking longer than usual. I wish the table would show a loading spinner when a query is taking place, so I know when the query is done loading
• Ability to hide the results table. A lot of the time I have some javascript that adds dates + transforms the data a bit, to which I pipe into Inputs.table to view. But then, I have two tables: 1 from the SQL cell that's "raw", and 1 for readers to view/interact with. I know I can have two separate cells like dataRaw and data (and drag the SQL table at the bottom so readers don't see), but doing it over and over again is repetitive.

[kimmolinna]

Just a couple of thing. 1) It would be nice to have Line numbers in SQL Cell queries because then it's easier to find line where the error is. 2) Inputs.tableis handling integers (big int) nicely but I have to cast all big ints to ints for SQL Cell tables

[chriszs]

Enjoying the visual refresh:

Though the visual inconsistency between the two table styles is even more glaring and I totally second everything @asg017 said, especially the loading indicator, which might actually be harder to see with the new gray background?

[chriszs]

To be honest, I'm not sure why one takes a pagination approach and the other an infinite scroll approach.

[CobusT]

Many of the issues mentioned here have now been fixed:

• [x] Pagination through scrolling
• [x] Numbers right aligned
• [x] Better handling of number of digits
• [x] Better font for numbers (and in general for tabled results)
• [x] Indicator of data loading (spinner at the right of the cell editor)
• [x] Ability to hide the table (through the 👁️ icon in the cell editor menu)
• [x] BigInts

Not done yet:

• [ ] You cannot view the table in the embedded view (also reported in #468)
• [ ] Line numbers in the SQL cell editor (not really related to this issue)

[mbostock]

BigInt is now supported, although I notice that we aren’t doing localized formatting for them yet. We’ll fix that.

[mbostock]

Number formatting is also improved in https://github.com/observablehq/feedback/issues/308#issuecomment-986288618:

[domoritz]

That's excellent. Thank you for all the great improvements!

There seems to be an issue with floats from DuckDB that are not recognized (shown as type "other").

This is from https://observablehq.com/@cmudig/duckdb.

[mbostock]

@domoritz It appears “other” is what the DuckDBClient is returning for the schema:

Specifically, it’s because the DuckDBClient doesn’t recognize the Float32 type. I sent you a suggestion to fix this here:

https://observablehq.com/compare/357f1a71f976f173...447275d083e033c7

[domoritz]

Ahh, I see. I really need to make the observable client a proper module with tests and move it out of a notebook. @enjalot and I talked about it last week (https://github.com/duckdb/duckdb-wasm/issues/1056).