In Print Layout Atlas Mode, Date string arrays are interpreted as non-numeric fields for Scatter Plot

[jo-chemla]

Describe the bug
Using the data-plotly plugin is incredibly useful, both during standard interaction with the data as well as within the Print Composer, so thanks a lot for that!
I've been trying using a data-plotly plot in Atlas mode in the Print Composer, and encountered the following issue. Dates which are stored as string lists (inside a vector layer where each single feature has one attribute which is a list of (0 to N) strings) are not interpreted as numeric values by data-plotly. The X Field can be set to attribute(@atlas_feature , 'dates'), but on the screenshot, we can see that the dates are plotted each in its own column rather than at the correct position on the X axis.

To Reproduce
Steps to reproduce the behavior:

1. Create a new vector scratch layer that contains an attribute with type list of strings, and enter some dates formatted as YYYY-MM-DD.
2. Enable atlas mode, use the vector layer as the atlas coverage layer, Add a Data-Plotly Item
3. Set the plot-type to scalar plot, X field to attribute(@atlas_feature , 'dates')
4. See error

Screenshots

Desktop (please complete the following information):

• OS: Windows 11
• QGIS release QGIS 3.30
• DataPlotly 4.0.3 updated at mar. juin 27

Additional context
When producing a scatter plot (in or outside of the print composer) with a layer which has a property that is a QDate or ISO-string (each feature has a single date as attribute, not a string list as attribute), then the dates are correctly interpreted as numeric values and correctly spaced.

[jo-chemla]

Hi there, any pointer would be very much appreciated in order to try to solve this bug. Seems that the core/plot_types/scatter create_trace method creates a graph_objs.Scatter( x=settings.x, ...) where settings is built from core/plot_factory.py, eg the settings.x=xx list is filled with x = x_expression.evaluate(context).

Might indicate that there is an expression that could be written to parse dates correctly on dataplotly depending on qgis cvector layer properties/column types.