Fix false "incomplete entity data" warning on time-varying populations

[chraibi]

Loading a normal pedestrian dataset raises a user-facing error popup claiming data is missing, when nothing is missing:

LogHdf5SimulationReader: Converted Juelich to Mobius: 493 entities, 1442486 samples, duration=1198.80s, fps=5.00
LogMobiusUserFeedback: Error: Incomplete entity data: HDF5: peak entity count 386 < MaxAgents 493.
  Some entities may be missing. (AgentDataSubsystem - RunHdf5SimDataGatheringLoop)

Cause

The check compares two quantities that measure different things:

• MaxAgents comes from ConvertJuelichToMobiusFormat as SortedIds.Num() — distinct entity IDs across the whole run.
• PeakEntityCount is the maximum entries in NumOfAgentsPerTimeStep — agents in the busiest single timestep.

These are equal only if every agent is present for the entire simulation. In any dataset where agents enter and leave, the peak concurrent count is legitimately lower than the total, and the warning fires.

Verified against the file above by reading the HDF5 directly:

unique ids (= MaxAgents):        493
peak concurrent agents:          386
agents present for the whole run:  0 of 493
median lifespan:                3193 of 5993 frames

All 493 entities are present in the data. Not one of them spans the full run, so the condition is guaranteed to trigger. It would stay quiet only on a synthetic file where the population never changes.

The spawn path is unaffected — SpawnMaxPedestrians uses GetMaxAgents(), so 493 MASS entities are created and the ones without samples at a given timestep simply do not render. This is a spurious warning, not a sizing bug.

Change

Measure the property the check was actually after: how many declared entities have no samples at all.

TBitArray<> ObservedEntities(false, MaxAgents);
// ... mark each Sample.EntityId ...
if (ObservedEntityCount < MaxAgents)
    → "HDF5: %d of %d entities have no samples. Data may be incomplete."

That still catches genuinely truncated data — an entity declared in metadata but absent from every timestep — while staying quiet on normal pedestrian flow. PeakEntityCount and its loop are removed as they now have no other use.

One pass over Hdf5Data.Samples with a TBitArray sized to MaxAgents; negligible against the 1.4M samples already being iterated in this function.

The neighbouring timestep-truncation check is left untouched — it compares like with like and is correct as written.

Verification

macOS, UE 5.5.4. Compiles clean; the warning no longer fires for the dataset above, whose 493 entities all have samples.

[Copilot]

Pull request overview

This PR fixes a false user-facing “Incomplete entity data” warning when loading HDF5 simulations with time-varying populations (agents entering/leaving over time). It replaces an incorrect comparison of peak concurrent agents vs. total distinct entities with a coverage check that verifies each declared entity appears in at least one sample.

Changes:

• Removed the PeakEntityCount computation based on per-timestep sample counts.
• Added an “entity coverage” check that tracks which entity IDs appear across all samples using a TBitArray.
• Updated the warning message to report how many declared entities have zero samples.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[chraibi]

Superseded by #18 — same false 'incomplete entity data' warning, cleaner fix (tracks which entity IDs ever appear instead of comparing peak concurrent count).