Haskell's record system (analogue of C's structs) has two major deficiencies. 1. Modifying a record value (that is, making a copy of a record value but with di

Haskell's record system (analogue of C's structs) has two major deficiencies. 1. Modifying a record value (that is, making a copy of a record value but with different values in one or two of its fields) is unnecessarily complicated and uncomposable. This makes modifying a subfield painful. 2. Field names are in the global namespace. Thus you cannot have e.g. a field named `map` (conflicts with the standard `map` function); nor a local variable named `owner` in the same scope that you use a field named `owner`; nor may two different record types both have a `name` field. C had this problem in the 70s (which is why e.g. struct tm has tm_sec and tm_min fields instead of sec and min fields), but they solved it a long time ago. The solution to deficiency 1 is to use lenses. Use the lens package from Hackage, but don't read its documentation at first: it generalises the problem exceedingly well, but this makes it harder to understand at first glance. Instead seek out a basic tutorial. At the cost of a short line of boilerplate for each record type, this works well. There is no satisfactory solution to deficiency 2. Some people define each record type in its own module, and import each module qualified. I don't think this scales well. I prefer to put a record-type prefix on each of my field names (i.e. the same thing C programmers were forced to do in the 70s).