Dashboards

The

Dashboard

page offers charts and dashboard visualizations based on data stored within Ganymede.

 

To assemble a dashboard:

• Add a dataset from the environment or create one via SQL query.
• Create a chart from the dataset.
• Save the chart to a specified dashboard.
• Modify dashboards and chart layouts by selecting the corresponding dashboard from the Dashboards tab in the top menu bar.

Adding a Dataset

Before creating charts, you will need to add datasets from the environment. These can be either (Physical Datasets) or Virtual Datasets (created via an ANSI SQL query).

Adding Physical Datasets

Physical datasets are outputted by Flows. To add a physical dataset:

• Click on the Datasets button in the top bar:

• Choose a database, schema, and table. Then click
  CREATE DATASET AND CREATE CHART
  .

Adding Virtual Datasets

Virtual datasets are created by querying existing tables in Ganymede.

• Hover over SQL in the top menu bar and click on SQL Lab. This should bring you to a SQL editor, where you can preview tables and run SQL queries.

• Enter your SQL query, select the number of rows to preview, and click
  RUN
  to execute the query.

• To save the virtual dataset, click the
  next to the
  SAVE
  button, and select
  Save Dataset
  .

For a list of available functions and syntax guidance, please reference the SQL in BigQuery documentation.

Saving queries

The

SAVE

button saves SQL queries for future edits, but does not register a query as a virtual dataset.

Creating a chart

Charts can be created from physical or virtual datasets.

To create a chart:

• Hover over Data in the top menu bar and click on Datasets. Select a dataset from the list, which will open the chart creation pane:

• Use the DATA tab to specify chart type and to pivot or aggregate data. The CUSTOMIZE tab allows you to annotate axis labels, specify axis formats, and set the chart title.

Saving charts to dashboards

Once the created chart is ready, click SAVE in the upper-right hand corner to add it to a dashboard:

Calculated columns

Calculated columns are useful for refining data visualization. Examples include:

• Taking the floor or ceiling of a variable to truncate data outside of a desired range
• Categorizing or subsetting data through string manipulation
• Extracting values for aggregation

To create a calculated column:

• Hover over the three dots to the right of the dataset on the Chart Creation page and click Edit Dataset.

• Select the Calculated Columns tab to add new columns to the dataset. Specify the field definition using a SQL expression.

Calculated Column flexibility

Calculated columns are more flexible than what is found in other BI tools offering similar functionality. Beyond aggregation, these fields can be filtered on, sorted by, and/or used to categorize data.

Adding Jinja Templates to SQL Queries

There are two ways to use Jinja templates in SQL queries:

1. Virtual Queries
2. Metrics

Example Using Virtual Queries

An example virtual query that inherits inputs from a filter can be constructed to use Jinja templates using:

{% set variable_name = filter_values('some_column_name')[0] if filter_values('some_column_name') else 1 %}

SELECT 
  *, 
  {{ variable_name }} AS inherited_filter_value,
  value * {{ variable_name }} AS value_multiplied_by_inherited_filter_value
FROM table_name

In this example, a variable is set by calling the filter_values function, a function native to the dashboard environment. The first value chosen is returned, since filter_values returns a list of selected values by default. The if statement checks to see a filtered value was selected on column some_column_name and returns 1 otherwise. This assigns a default value to the query so that the dataset can be created and not raise any errors if there are no active filtered values present on that column. A dataset will not be savable without the if statement. After saving the query to a virtual dataset, any chart in a dashboard will have the Jinja template automatically be rendered.

Examples Using Metrics

Creating a chart, a line chart for instance, requires creating a metric on that variable. You can use filter values by either referencing the new column created in a virtual query or by making the metric as follows:

AVG(value * {{ filter_values('some_column_name')[0] if filter_values('some_column_name') else 1}})

Where this mimics what was done in the virtual dataset above but instead works for both virtual and physical datasets. The if statement is required here as well in order for the chart to be displayed when a filter is not selected on some_column_name.

Notes

Its important to make sure that you have

1. if statements to check if a filter value is present
2. You are referencing the column name that is being filtered on in the dashboard
   a. For instance you have a filter defined in the dashboard named My Filter created on dataset my_dataset on column some_column_name where some_column_name is referenced in the Jinja template
3. The column name is not contained in the dataset you are using the Jinja template in (unless you actually want your chart to be filtered as well)
4. You may need to wrap your Jinja template in a CAST to convert to INT or NUMERIC if necessary
5. Other useful links for more detail or other use cases of Jinja:
   a. From Preset website
   b. From Apache Superset website

Editing dashboards

To edit a dashboard:

• Select Dashboards from the top menu bar and navigate to the desired dashboard. Click
  EDIT DASHBOARD
  to resize and rearrange the layout of different charts.
• To filter data within graphs, click the icon on the left side of the dashboard, then select
  Add filters and dividers
  and use the SCOPING panel to apply filters.

Permissions for editing datasets

Dashboards can only be edited by their owner(s). One method for multiple users to avoid conflicting dataset modifications would be for users of shared datasets to create virtual datasets for their own use.

Tables referenced by dashboard filter

The SCOPING tab on the filters modal allows users to specify which graphs the filter applies to. Note that if 2 charts are backed by different datasets which share a field name, a single filter can be used to manipulate both tables simultaneously.