Opteryx Studio

Loading…

Opteryx studio

Catalog

Details

Select a dataset to see its row count, columns, and partitions.

Variables

No variables defined. Use + in the panel header to add one.

What did that query cost?

Run a query and this tab will break down where its time went, how much data it read and from which datasets, and what came back.

Reach for it when a query is slower or costlier than you expect: the time breakdown tells you whether it waited, thought or worked, and the datasets tell you where the bytes came from.

Example details

Total time	892ms · 862ms in the engine

Where the time went

Queued	14ms
Planning	12ms
Execution	850ms
Other	16ms

Data

Data scanned	340 MB
Rows read	4,812,330
Result	42 rows · 3 columns

Datasets

public.github.events	4,812,330 rows · 340 MB

Where the time went Queued is waiting for a worker, planning is working out how to run the query, execution is reading and processing the data. A big planning number points at a complex query; a big execution number points at the data itself.
Data and datasets Data scanned is what you're billed for, split by dataset below. Rows read far above the result usually means most of what was read got filtered out.

How did Opteryx answer that?

Run a query and this tab will show the steps the engine took — reads, filters, joins and aggregations — as a tree, with the rows and time each step accounted for.

Each step feeds the one it is nested under, so the deepest steps run first. The step handling the most rows is usually the one worth attacking, and a filter that sits close to the read it belongs to is far cheaper than the same filter applied after a join.

Example step

INNER JOIN 8.42ms · 34%

slowest

128K rows (4.1 MB) 32K rows (980 KB)

Step name What this step is doing — reading a table, joining, filtering, sorting, and so on. The line underneath it gives the specifics: which table, which columns, which condition.
Color How costly this step was — red is the single slowest step in the query, amber still takes a meaningful share of the total time, green is comfortably cheap.
Time & share How long this step took, and what percentage of the whole query's time it accounted for.
Rows in / out Rows (and data volume) that flowed into and out of this step. A big drop means a filter did real work; a big jump means a join fanned rows out.

Click any step in a real plan to expand it and see its full detail.

Title

Chart type

Stacked Log scale

X axis

Y axis

Split by

Sort

Run a query that returns at least one numeric column, then choose what to plot on each axis using the controls to the left. Save (above the results) will download the chart as an image or copy it to the clipboard.

Learn

Language reference

Query Basics

Use SELECT … FROM … to read from a dataset.

Add WHERE to filter, ORDER BY to sort, LIMIT to cap results.

Aggregate with COUNT, SUM, AVG, MIN, MAX alongside GROUP BY.

Useful Tips

Results scroll horizontally when columns exceed their minimum width.

Use the Results, Explain, and Details tabs to inspect output.

Upload prepares files first; commit wiring comes later.

Function reference

ABS +

Absolute value.

APPROX_DISTINCT +

HyperLogLog distinct count.

ARRAY_CONTAINS +

Test if array contains item.

ARRAY_CONTAINS_ALL +

Test if array contains all items from set.

COUNT +

Row count.

DATE_TRUNC +

Bucket timestamp to unit.