Summary

Fixes agenttype_reporters not collecting data when using batch_run(), making it impossible to use selective agent data collection with parameter sweeps and parallel execution.

Bug / Issue

Closes #2804

Users reported that agenttype_reporters worked perfectly in regular model runs but failed to collect data when using batch_run().

What was expected: Data from agenttype_reporters should appear in batch_run results, just like agent_reporters data does.

What actually happened: The agenttype data was silently dropped - no errors just missing columns in the results DataFrame.

Implementation

The issue was in the _collect_data() helper function in mesa/batchrunner.py. This function extracts data from the DataCollector after model execution, but it only looked at _agent_records and completely ignored _agenttype_records.

Changes made:

1. Added iteration through dc._agenttype_records in _collect_data()
2. Process agenttype data using the same structure as agent data
3. Combine both into the all_agents_data list for consistent output

It handles all edge cases:

• Models with only agent_reporters
• Models with only agenttype_reporters
• Models with both reporter types
• Models with neither (model-only data)

Testing

Added test coverage in tests/test_batch_run.py:

1. test_batch_run_agenttype_reporters(): Verifies agenttype_reporters data is collected correctly

   • Tests with different parameter values (n_agents: [3, 5])
   • Validates correct number of rows (8 agents = 8 rows)
   • Confirms all expected columns are present

2. test_batch_run_agenttype_and_agent_reporters(): Tests mixed usage of both reporter types

   • Ensures both data types coexist in results
   • Verifies no data loss from either reporter type
   • Checks proper handling of duplicate agent entries

Test Results:

• All 20 batch_run tests pass (18 existing + 2 new)
• All datacollection tests pass (no regressions)