CSV.parse - Support Headers

[stellarpower]

Feature Request

Whilst it's good that CSV is implemented as an efficient parser that iterates over data as it's needed, sometimes it's convenient to read the lot in as one big array of arrays.

The constructor for CSV has an optional headers argument for handling a header row automatically, whilst the parse static method returns an array of arrays of strings for grabbing the thing as a table. Often I'd like to get a table in one line, but also use a header row.

So I'd like to propose adding this to the parse method, and I think the most suitable option would be, if headers is true in parse, to return an array of hashes, so that I can index by row and then extract the cell through its column key from the hash.

Thanks

[Blacksmoke16]

Would probably have to be a new overload, as adding it to the existing .parse method would change its return type and be a breaking change.

[AgileIndustrialComplex]

@stellarpower you should be able to achieve this with code like

require "csv"

csv = CSV.new("Name, Age\nJohn, 20\nPeter, 30", headers: true)
rows = [] of Hash(String, String)
while csv.next
    rows << csv.row.to_h
end

[stellarpower]

I can thanks, and in my project I extended the base class to do it, it's just nice to have a one-liner in the standard library. I have to jump through multiple languages in a day, and I'd say 80+% of the Crystal I write involves CSV, JSON/YAML, or less often XML, and it's nice to be able to read in things like that with minimal overhead. If I've just been working in Python say, either I need to flip my brain and remember it or it's a bit of extra typing that adds noise (File.read is vastly nicer than with open(...) as ..., and don't even get me started on simply flattening a list!).

I guess similarly a file keyword argument would be nice, so I don't have to bother opening it myself either. And it'd also be convenient to take an array of array of (things that are safe to serialise) or an array of hashes to (things that are safe to serialise) and just output that to file without needing to use the builder and iterate. Perhaps there's a more general discussion here on CSV. I know there have been efforts to keep the standard library slim, but also these one-line solutions to things that mean there's less you need to hold in your brain are part of what always made Crystal and Ruby nice to work in.