Bug: value/uncertainty AutoDigits does not normalize decimal input before finding bottom digit but value AutoDigits does

[jagerber48]

See

from decimal import Decimal

from sciform import AutoDigits, Formatter

formatter = Formatter(ndigits=AutoDigits)

print(formatter(Decimal("1.00")))
# 1
print(formatter(Decimal("1.000")))
# 1
print(formatter(Decimal("1.00"), Decimal("0.10")))
# 1.00 ± 0.10
print(formatter(Decimal("1.000"), Decimal("0.100")))
# 1.000 ± 0.100

These two behaviors shouldn't differ. It's not actually obvious which should be preferred. Due to current implementation details, the value decimal is always normalized before determining the least significant decimal place, while in the value/uncertainty mode the normalization does not necessarily happen.

My inclination is that normalization should happen at the first input step and at the formatting step to guarantee equal inputs get equal outputs. The argument against this is that one of the purposes of the Decimal encoding is to preserve significance of digits which are not technically significant digits.

[jagerber48]

Closed by #149, All imports are normalized early