Improve Decimal Precision in XBRL Mapping

[PapediPoo]

Improve Decimal Precision in XBRL Mapping

After reviewing the feedback on the newly released Brel Report, it was pointed out that the mapping table currently uses floats for the decimal type, which may lead to precision errors and does not fully conform to XBRL’s 18-digit precision requirement.

Issue

• 32-bit floats have ~7 digits of precision.
• 64-bit floats have ~16 digits of precision.
• XBRL-compliant processors require at least 18 digits of precision.
• Using floating-point numbers could introduce rounding errors and compliance issues.

Proposed Consideration

We should assess whether using decimal libraries would be a better approach for handling high-precision decimal values, such as:

• Python’s decimal library (decimal.Decimal)
• pandas-decimal

Criteria of Acceptance

• Identify potential refactoring needed for compatibility
• Do refactoring
• Verify compliance with XBRL precision requirements.
• Add test cases

[ghislainfourny]

That's a great point, @PapediPoo
Do you know if the pandas decimal library is better than the Python library? Otherwise we could use one for storage and support getting both kinds in the API.