Introducing MultiNodeObservation class

[jpalm3r]

Introduces MultiNodeObservation class to streamline creation of multiple node observations from a single data source, significantly improving the user experience for network model workflows.

Changes

New Features

• MultiNodeObservation class: Creates multiple NodeObservation objects from a single DataFrame or Dataset
• Auto-assignment: Automatically maps data columns to nodes when counts match
• Flexible naming: Supports both string prefixes ("S" → "S_0", "S_1") and explicit name lists
• Consistent metadata: Applies quantity, aux_items, and attrs uniformly across all observations

API Improvements

• Updated documentation: Added MultiNodeObservation to module docstring alongside existing observation types
• Factory function compatibility: Maintains consistency with existing observation() patterns

Usage Examples

Before (manual creation)

# Tedious: create each observation separately
obs1 = ms.NodeObservation(df, item=0, node=123, quantity=quantity)
obs2 = ms.NodeObservation(df, item=1, node=456, quantity=quantity) 
obs3 = ms.NodeObservation(df, item=2, node=789, quantity=quantity)
observations = [obs1, obs2, obs3]

After (streamlined)

# Efficient: create all observations at once
observations = ms.MultiNodeObservation(df, nodes=[123, 456, 789], quantity=quantity)

Benefits

• 🚀 Reduced boilerplate: Single call replaces multiple manual instantiations
• 🔧 Error prevention: Validates node count against data columns
• 📊 Consistency: Ensures uniform metadata across all node observations
• ⚡ Better UX: Simplified workflow for network modeling scenarios

Technical Details

• Inherits from list for familiar collection behavior
• Validates input data structure against node requirements
• Supports both DataFrame and xarray.Dataset inputs
• Maintains full compatibility with existing NodeObservation functionality

This enhancement makes network data handling more intuitive and reduces the cognitive load for users working with multi-node datasets.

[ryan-kipawa]

General comment: why is there no MultiPointObservation?

I am sort of questioning if we need such an abstraction. Do users need list-like functionality on a collection of observations, or is this more a convenience for constructing many observations from a single source?

One disadvantage I see is that it only works for collections of observations that share the same time index. If it's as a convenience function, then I wonder if it fits better as a static method or something like that on NodeObesrvation?

[jpalm3r]

General comment: why is there no MultiPointObservation?

I am sort of questioning if we need such an abstraction. Do users need list-like functionality on a collection of observations, or is this more a convenience for constructing many observations from a single source?

One disadvantage I see is that it only works for collections of observations that share the same time index. If it's as a convenience function, then I wonder if it fits better as a static method or something like that on NodeObesrvation?

It is just for convenience, after our previous discussions on user-friendliness and a potential use case of one model vs multiple sensors.