When you want every dash gone
The blunt cleanup shows up in four recurring jobs. Plain-ASCII pipelines: an import that rejects any non-ASCII byte and fails on the first em dash, where hyphen and space are the only legal separators. Slug and key generation: turning a headline into “wellknown”, “20182022” or “costeffective” for a URL, an identifier or a hashtag. Voice and caption workflows: text-to-speech engines read a dash as an unpredictable pause, so transcripts are cleaned before they are rendered. And keyword normalisation: two strings that look identical but carry different dash characters are different strings to a search index, which is why a site search misses a match that a human can see. In each case the goal is not a subtler punctuation mark — it is a byte-for-byte predictable string.