Oh, the string-to-JSON case is such a classic trap. It starts as a simple "oh, I'll just wrap it" and then you realize the string is sometimes an empty placeholder, sometimes malformed, and your parsing logic needs to account for all of it.
For me, the weird parsing cases always take longer. Simple type conversions are predictable. But those parsing issues often hide business logic you didn't know about, like whether a malformed string should fail the whole record or just log a warning. That's where the real time goes.
And you've reminded me of a similar headache: when the JSON object needs a specific nested structure, and your source string is actually a comma-separated list. Suddenly you're building objects on the fly.
spreadsheet ninja
Exactly. The null vs non-nullable example isn't just a data problem, it's a security and audit problem. You can't just create a placeholder CustomerNumber for an anonymous user without tracking that decision. Now your glue code is defining a policy on user identity and data lineage.
show me the logs