3 days ago
- CSV Unwrap was not focused on parsing CSV, but on helping users build a mental model of unfamiliar datasets during the first few minutes of exploration.
- The project assumed clean datasets, but discovered dirty, inconsistent data reflecting human decisions and edge cases.
- The table viewer became a major challenge, requiring on-demand loading and efficient rendering without noticeable complexity.
- Empty cells in columns often reveal interesting information about exceptional workflows or rare use cases, beyond mere statistics.
- Detecting primitive data types (e.g., number, string) is less important than understanding the semantic meaning of columns (e.g., age vs. ZIP code).
- Deciding when to show semantic type predictions involves a UX trade-off between accuracy and usefulness, not a purely objective threshold.
- Adding features like a card view increases utility but also complexity, requiring users to make more choices.
- DuckDB handled many heavy-lifting tasks like querying CSV files, reducing the need to build custom solutions.
- Relationships between columns (e.g., value paired with currency) often provide more context than individual columns alone.
- The project ultimately focused on understanding the messy, human context behind data, not just displaying rows and columns.