[MarkusNeusinger]

Closes #12224

Summary

• A currency column of three-digit comma groups ("$1,200", "$950", "$12,500", also 1,200 USD, € 1,200, CHF 12,500) read as 1.2, 950 and 12.5. It now reads as 1200, 950 and 12500 with the warning "commas read as thousands separators, as a currency amount has at most two decimals" (telemetry code thousands_separators). A cell with a one- or two-digit fraction ($1,25), a four-digit group or a zero-leading group keeps the column on decimal commas. The 2026-10-15 boundary stays: a dot number in another column is no evidence, and a bare 1,200 keeps its question.
• A file that speaks for decimal commas behaves exactly as on main. An unambiguous decimal-comma column settles the file first, as before. Owner decision of 2026-10-22: a semicolon delimiter, or a column of dot groups (7.184.003) or 1.234 shapes, also keeps the question, so a German fuel receipt's 1,789 € stays 1.789.
• _retype_thousands_commas is shared by the currency rule and the existing cross-column rule. A whole number past int64 now makes such a column text with BIG_INTEGERS on both paths, as _retype_thousands_dots already did, instead of a rounded decimal-comma float.
• Implemented by Gemini 4 Argon in the owner's interactive agy session (commits 670aad1, ae5a78c, e0c1dc9) and reviewed by Claude. An independent Opus review found that the first version ran the currency rule before the file's own decimal-comma evidence, so Datum;Preis;Liter / 01.10.2026;1,789 €;45,5 read 1789. Commit 7e3671c fixes that and adds the review follow-ups.

Plan

Issue #12224 (revised 2026-10-22 after Gemini's report-back).

Test plan

• uv run pytest tests/unit tests/integration -n auto (12,285 passed); new tests in TestThousandsGroups cover the seven acceptance cases, the files that keep their decimal comma (decimal-comma column, semicolon, dot groups, 1.234 shapes), an id column, and the int64 fallback with and without a currency
• uv run ruff check . && uv run ruff format --check ., uv run mypy api core
• uv run python -m tools.changelog check --base origin/main
• Probe against main: every file that speaks for decimal commas (the review's five inputs) gives identical CSV and warnings; only the currency columns change

Not changed on purpose (review notes): a column read as currency thousands is no evidence for another bare 1,250 column (it keeps its question), and a currency named only in the header (price (€)) or in lower case (usd) is not detected.

🤖 Generated with Claude Code

https://claude.ai/code/session_01GdMSqLR5ww4ji74EUSmk9M

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[Copilot]

🟡 Changes recommended

Two decimal-comma safeguards miss valid evidence, allowing neighboring currency values to be multiplied by 1,000.

2 open issues

• Preserve overflow dot-group evidence for currency parsing · New
• Disable currency rule for zero-leading decimal comma evidence · New

What changed in this PR

Updates the agent dataset parser for #12224 to interpret three-digit comma groups in currency columns as thousands, while preserving decimal-comma handling when the file provides contrary evidence.

Changes:

• Adds currency-based interpretation and shared comma-thousands conversion with an int64 overflow fallback.
• Adds regression tests and telemetry classification for the new warning.
• Documents the parsing rule and its exceptions.

File	Description
tests/​unit/​agents/​data/​test_warning_codes.py	Tests classification of the currency warning.
tests/​unit/​agents/​data/​test_parse.py	Covers currency groups, decimal-comma safeguards, identifiers, and overflow.
docs/​concepts/​agent-network.md	Documents currency interpretation and exceptions.
changelog.d/​currency-thousands-commas.md	Records the parsing correction.
agents/​anyplot/​data/​warning_codes.py	Recognizes the new thousands-separator warning.
agents/​anyplot/​data/​parse.py	Implements currency inference and shared conversion helpers.

_{Review effort: Balanced}