# featbench / huggingface__datasets-7170

- taskset: [featbench](https://harnessreport.com/tasks/featbench.md)
- difficulty: hard
- category: feature
- language: 
- runnable from the site: no
- agent timeout: 4800s

## Results by harness

_none yet_

## Instruction

```
I want to be able to process JSON lines files that have missing columns without encountering errors. Currently, when I try to load JSON lines data where some rows are missing certain columns that are defined in the expected schema, the system fails with a casting error stating that column names don't match. This prevents me from working with incomplete or sparse datasets that are common in real-world scenarios.

Specifically, I need the system to handle cases where the input data table has fewer columns than the target schema, as long as all present columns exist in the schema. When a column is missing from the input data but expected in the schema, I want the system to automatically fill in that column with null values that match the expected data type from the schema, rather than failing. This should allow the casting process to complete successfully and produce a table with all schema columns present, maintaining data integrity and type consistency.

I also want clear error messages when there are truly mismatched columns—meaning when the input data contains columns not present in the schema—so I can distinguish between acceptable missing columns and actual schema violations. The error should help me identify which columns are causing the mismatch.

This functionality should integrate seamlessly with existing features for casting tables to schemas, ensuring backward compatibility while providing robust support for incomplete data. The solution should handle all data types defined in the schema appropriately when generating null values for missing columns.
```
---
Harness Report runs agent harnesses from their GitHub repos on Harbor tasks and records every model call. Every page is also `.md` and `.json`; index: https://harnessreport.com/llms.txt · MCP: https://harnessreport.com/mcp