Encoding, decimal commas, and silent order-of-magnitude errors

1 · Learn the move · Red-team audit

A European instrument writes 1,5 for one-and-a-half; a parser reading US conventions sees one thousand five hundred, or fifteen, or a string, depending on its mood - and every one of those outcomes can run straight through to a plot without an error message. Encoding is the same class of silent: a cp1252 micro sign or degree symbol in a header breaks a naive utf-8 read, or worse, survives it mangled, and a column called Conc (ug/mL) becomes unrecognizable to the join that expected it. These are not exotic failures; they are what happens every time a file crosses an instrument, a locale, or a decade. The move is a red-team audit of the raw file before any parse: attack it looking for encoding, delimiter, decimal convention, and thousands separators, with evidence quoted for each finding - and a checksum row where a known value proves the parse read what the instrument wrote.

Red-team this raw file before parsing. Attack four fronts, quoting evidence for each finding: (1) ENCODING - non-ASCII bytes, where, and what they should be; (2) DELIMITER - what actually separates fields, including the semicolon-because-comma-is-decimal pattern; (3) DECIMAL CONVENTION - comma or point, with example values that prove it; (4) MAGNITUDE TRAPS - thousands separators, unit prefixes in headers, exponent notation. Then state the exact parse settings, and verify with this control: [a value you know, e.g. sample X concentration is about Y]. Raw file head and tail: [paste]