[Sad Catapella N3][Data integrity] Parameter channel indices silently truncate fractions and booleans #38

Baseline and classification

Batch 3, f784c95e3049483f87466150ffc5764375fa631d. Confirmed input coercion; runtime integration not executed. Distinct from numerical normalization parameter validation.

Evidence

• Parameters.__post_init__ lines 112–119 applies int(c) to every supplied channel index without checking that each is an integral, non-boolean value.
• Parameters.select lines 144–153 also converts the requested channel identifiers via int(c) before membership validation, allowing 1.9 to be interpreted as channel 1 and True as channel 1.
• Tests cover valid integer selection and unknown integer channels, but not fractional inputs, booleans or numeric strings.

Consequence / example

Parameters('z-score','image',None,True,(0.0,10.0),(1.0,1.0), channels=(0.0,1.9)) becomes channels (0,1) without warning. parameters.select([0.9]) silently selects channel 0 instead of rejecting an invalid channel identifier. This can associate normalization parameters with the wrong fluorescence marker.

Fix and tests

Require actual index-like integers (operator.index or explicit numbers.Integral policy) and reject bools. Validate no negatives and strictly increasing unique indices; preserve meaningful error messages. Test float fractions, float whole numbers, bool, text digits, NumPy integer scalar types, valid noncontiguous subsets, and mapping preservation. Do not silently round/truncate. No Numba role.