As a rule of thumb I'd estimate that cursors are not the simplest, most maintainable or most efficient solution to at least 99% of data manipulation problems in SQL. It's unfortunately very easy for an inexperienced SQL developer to fall into the trap of writing lots of procedural cursor-based solutions before they get to grips with writing set-based SQL statements. That's because many people just starting out with SQL will already be familiar with procedural programming languages and so cursor based solutions just seem more familiar and more convenient. To steer beginners away from that trap is one reason why it's common to find people advising that you should avoid cursors as much as possible.

Of course it's an over-simplification to say that cursors should be avoided altogether. That's not always possible and sometimes cursor-based solutions are more efficient or even simpler than any alternative. My advice for someone starting out in SQL would be to avoid learning about cursors until you are very familiar with the rest of the SQL language and then to seek a second opinion if you still think a cursor might be a good idea.

As already stated, ref cursors in Oracle are something different and nothing to do with what standard SQL and other DBMSs call a cursor. The above comments do not apply to ref cursors.