Slugify Article Titles — Solution & ExplanationProblems / Slugify Article Titles / EditorialApproach

Normalize case first. lower(title) shrinks the problem: the junk-detection class is just [^a-z0-9].

Squash junk runs. regexp_replace(..., '[^a-z0-9]+', '-', 'g') — the + makes each *run* of junk one match (so " & " becomes one hyphen, not three), and 'g' applies it across the whole string.

Trim the edges. Leading/trailing junk became leading/trailing hyphens; '^-+|-+$' with 'g' removes both anchors in a single pass.

Key InsightText normalization pipelines are best built as ordered, idempotent passes — lowercase, canonicalize separators, trim — where each pass makes the next one simpler, and running the pipeline twice changes nothing. The two classic slugify bugs are both single-flag mistakes: forgetting 'g' (only the first junk run collapses) and forgetting the + (every junk character becomes its own hyphen, yielding sql---duckdb). Idempotence is the property to test for: a correct slug, slugified again, must come back unchanged.Hints

Lowercase FIRST — then the character class only needs [^a-z0-9], with no uppercase letters to worry about.

The + quantifier in [^a-z0-9]+ is what collapses 'Data--Engineering' to a single hyphen: the whole junk run is one match, replaced once.

Titles with leading/trailing junk leave a stray hyphen at the edges after pass one. A second regexp_replace with the anchored pattern ^-+|-+$ cleans both ends in one go.

DuckDB's regexp_replace replaces only the FIRST match unless you pass the 'g' flag — forgetting it slugifies only the first word boundary.

Solve Slugify Article Titles yourself →