Same answer everywhere

“Port” here means the same behaviour, not the same idea. The behaviour is written once, in the hub’s spec/, and every port runs the same test files from it in its own CI.

What every port runs

File	Rows	What it checks
month-boundaries.csv	7,176	the first and last day of every BS month from 1901 to 2199, in BS and AD
csharp-golden/dates.tsv	1,018	conversion, weekday, day of year, month length, every format, fiscal year and quarter
csharp-golden/add-days.tsv	9,162	adding days
csharp-golden/add-months.tsv	10,180	adding months, clamped and spilled
csharp-golden/format-pattern.tsv	451	custom patterns
parse.tsv		the strict and lenient parsers, errors included

The golden files are the recorded output of the published C# NepDate 2.0.7 package. Where C#’s documentation and its code disagree, the code is the reference.

Beyond the vectors, every port runs an exhaustive round trip over all 109,212 days in the range.

Deliberate differences from C#

A few rows differ from C# on purpose, for example the short names Asa and Aso that keep Ashad and Ashoj apart, or a lenient parser that refuses ambiguous input instead of guessing. Each difference has an id (D-01 to D-15) in PARITY.md, and every port differs in the same way: a port skips a golden row only through a filter that names the deviation.

There are no port-only differences. If one seems necessary, it is proposed in the hub, decided once, and applied to every port.

Check it yourself

The hub has an executable version of the spec. With Python 3.10 or newer and nothing else:

git clone https://github.com/sanamhub/nepdate
cd nepdate
python spec/reference/verify_vectors.py

It checks the algorithm against the month boundaries and the golden rows, plus the formatting and parsing specs.