TL;DR: pandas needs pyarrow installed to read parquet files and it is missing. Run `pip install pyarrow`, restart your interpreter or kernel, and re-run read_parquet.

## The error
```text
ImportError: Missing optional dependency 'pyarrow'. Use pip or conda to install pyarrow.
```

## Fix it
1. Install pyarrow:
   ```
   pip install pyarrow
   ```
   Expected: install completes.
2. Restart your Python interpreter or notebook kernel so the new package is picked up. Expected: fresh session.
3. Retry the read:
   ```python
   import pandas as pd
   df = pd.read_parquet("data.parquet")
   print(df.head())
   ```
   Expected: prints the first rows.

## When this applies
Use this when read_parquet, to_parquet, or read_feather raises the missing-pyarrow message.

## When it does not apply
If the error names fastparquet instead, either engine works; pyarrow is the recommended one. If pyarrow is installed but the read still fails, check the file is valid parquet.

## Compatibility
pandas 1.x and 2.x. pyarrow wheels exist for all major platforms; very old 32-bit systems may need fastparquet instead.

## Root cause
Parquet support in pandas is delegated to an external engine, pyarrow by default. pandas does not hard-depend on it, so a fresh environment hits this on first parquet use.

## Edge cases
In notebooks, installing without restarting the kernel leaves the old import state; the restart is required. If you pin pandas, also pin a compatible pyarrow; major version skew can break the engine interface.