Scientists rename genes to stop Microsoft Excel from misreading them as dates (2020)

Original Article ↗ Hacker News Discussion ↗

Scientists in genomics have long struggled with Microsoft Excel automatically converting certain gene names (like SEPT2) into dates, corrupting research data shared via CSV files. Commenters debate whether this is primarily a usability failure in Excel’s aggressive type inference, a “skill issue” for researchers who don’t lock columns to text or use alternative tools, or an inevitable trade-off of convenient implicit conversions that underpin Excel’s popularity. The thread also touches on broader themes: institutional lock-in to Microsoft tools, the limits of CSV as a format without schema, and the tension between strict typing and ease of use in widely deployed software.

Excel auto-conversion & new toggle

  • Excel historically auto-converts “gene-like” strings (e.g., SEPT2, MARCHF1) into dates or numbers, corrupting scientific data, especially via CSV round-trips.
  • In 2023 Microsoft added a per-file toggle to disable automatic data conversion on Windows/macOS; some see this as overdue and still imperfect (edge cases, macros, older versions).
  • Others argue Excel already allowed column types (Format Cells → Text, Import Text Wizard, Power Query), so the problem is user workflow and awareness.

Implicit conversions as a broader problem

  • Several comments frame this as part of a larger “implicit conversion / type safety” problem seen in JavaScript, MySQL, YAML, PHP, etc., sometimes called another “billion-dollar mistake” alongside null pointers.
  • Others counter that permissive conversions also enabled huge productivity gains, and are widely chosen because they’re convenient.
  • Debate over whether the right framing is “type safety” (actionable) vs. vague “expectations vs. reality” mismatches.

Why scientists still use Excel

  • Many argue scientists are constrained by institutional IT, licensing, and standardization; they cannot freely choose alternative tools even if they’d like to.
  • Some say most domain experts have already moved off Excel for this use case, but a minority (plus admins, clinicians, editors) can still corrupt shared data.
  • Pro-renaming view: changing a handful of gene names is a low-friction, field-wide mitigation compared to retraining everyone and replacing tools.

CSV, formats, and alternatives

  • CSV is criticized as inherently ambiguous (no schema, no types); Excel’s double-click CSV behavior silently converts and discards leading zeros, plus signs, and gene-like IDs.
  • Workarounds include: pre-formatting columns as text, using import wizards, small macros, or encoding text as Excel formulas in CSV (="Data").
  • Alternatives mentioned: LibreOffice (similar auto-conversion but slower and less compatible), JSON with a hypothetical .xljson extension, databases, Jupyter/R/Python, Emacs org-mode.

Responsibility and expectations

  • Some blame scientists for not mastering their tools; others call this unreasonable given Excel’s hidden defaults and moving target behavior.
  • Several comments emphasize this is as much a social/organizational problem (IT gatekeepers, vendor lock-in, user training) as a technical one.