I hate parsing CSV. There are so many different implementations it's a constant cat and mouse.. Literally any symbol can be the separator, then the ordering starts getting changed, then since you have to guess what's where you go by type but strings, for example, are sometimes in quotations, other times not, then you have some decimal split with a comma when the values are also separated with commas so you have to track what's a split and what's a decimal comma.. Then you get some line with only 2 elements when you expect 7 and have no clue what to do because there's no documentation for the output and hence what that line means..

If the CSV is not written by me it's always been an exercise in making things as difficult as possible. It might be a tad smaller as a format but I find the parsing to be so ass you need really good reason to use it.

Edit: Oh yeah, and some have a header, others don't and CSV seems to always come from some machine where the techs can come over to do an update, and just reorder everything because fuck your parsing and then you either get lucky and the parser dies, or since you don't really have much info the types just align and you start saving garbage data to your database until a domain expert notices something isn't quite right so you have to find when was the last time someone touched the machines and rollback/reparse everything..