This seems more useful if re-framed as "Why pulling data you won't use is bad for performance."
SELECT * seems like a strawman to me. I don't often see it in the wild anymore. But pulling more columns than you need is extremely common; it happens in every codebase I've ever seen. I habitually do it myself. More-or-less every time I choose to re-use a single function for retrieving data in several places. Because then the function needs to get the union of all the columns that all the callers need.
Which may be a reasonable trade-off. There's almost always a need to strike a balance between maintainability and performance. But it's also nice to have occasional reminders to re-assess what you've been doing. This particular practice tends to have a particularly high cost, and one that, depending on how you configure your environments, may be much larger in production than it is in development or CI.
SELECT * seems like a strawman to me. I don't often see it in the wild anymore. But pulling more columns than you need is extremely common; it happens in every codebase I've ever seen. I habitually do it myself. More-or-less every time I choose to re-use a single function for retrieving data in several places. Because then the function needs to get the union of all the columns that all the callers need.
Which may be a reasonable trade-off. There's almost always a need to strike a balance between maintainability and performance. But it's also nice to have occasional reminders to re-assess what you've been doing. This particular practice tends to have a particularly high cost, and one that, depending on how you configure your environments, may be much larger in production than it is in development or CI.