The default behavior when using data connectors is to retrieve data into memory for visual analysis to then occur, where the data is aggregated and filtered in memory. This retrieval may be the consumption of a whole dataset, or through the use of parameters for the retrieval of a dynamically selected subset of the data. This approach is however limited by the memory of the machine, and the overhead of retrieving and processing large volumes of data on the desktop.
On-demand queries provide ROLAP functionality to the Altair Visual Data Discovery products, where the aggregation and filtering tasks are largely offloaded to the underlying data repository.
The SQL dialect is necessary, given that the software will dynamically generate SQL or q query for:
q Filter domains (Categorical Listing and Min/Max for Numeric Fields)
q Aggregated and Filtered Data returned in the visualizations
Each filter and visualization are driven by a separately generated SQL or q query, ensuing that each query is optimized for index utilization, and returns the minimum amount of data.
This on-demand capability dramatically reduces the amount of data to be transferred across the network and onto the desktop and ensures that the heavy data intensive tasks occur at a server. However, when using this mode, the following functionality is disabled:
q Percentile Filtering
q Copy Raw Data
q Pivot and Unpivot Data Transforms
q Non-Additive Data support.
q Shared selection across visualizations
q Numeric Bucketing
q Time Part Specific Options (Decade, Quarter, Weekday, Millisecond, Nanosecond)
q R Transform
q Python Transform
In Panopticon 2020.0, data preview for on-demand data sources is not supported yet. However, the on-demand data can be displayed on the workbook dashboards.