Let me know if I should add comments about justification to the commit message.

You can write it here. You can also report an issue first, and then reference it in both the commit message and the PR, which usually makes it easier to track things.

Updated commit message to explain justification and made test more robust.

@tristil Thank you! But CSV is not an enumerable. An instance of CSV is more like a cursor that always points to a row and you can fetch any of the columns by index, header name or regex. That's why each doesn't return the row (your question in Google Groups). In fact a CSV instance has just one instance variable for storing all rows, but it just stores one at a time, for efficiency (to avoid allocating new arrays for each row). You can invoke to_a on a CSV to get a copy of that row.

For an Enumerable (well, Iterator), you can use CSV.each_row:

CSV.each_row(string_or_io).with_index.each { |row, index| ... }

@asterite Okay, I understand now, and I see that this is well documented on the class. What confused me, besides not reading the documentation well, is that there is an each method which just calls next for you. My thinking was that anything that implements each should be an enumerable, but I guess the idea is that this could create problems because e.g. if you did CSV.new(string_or_io).each(&.itself) you would get back the same instance over and over again? I'm still vague on the principle that governs when something that implements each should include Enumerable, but I am satisfied that I can use the other interface, which cleans up my code. Thanks!

While I'm here, would it make sense to implement the headers functionality in CSV Parser? If so, I can take a shot at it.

@tristil The headers functionality is already there. What do you need to do?

@asterite So my new code is

require "csv"

def get_rows
  text = File.read("vendors.csv")

  CSV.each_row(text).map_with_index do |row, index|
    next if index == 0

    row
  end.compact
end

rows = get_rows

The only reason I need the index is to skip the first row, which I would get for free with headers: true if I were using the standard interface. Right now it seems like you have to choose between using the iterator/parsing interface but you don't get the headers switch or using the standard interface. Seems like it would be good to have CSV.each_row(text, headers: true) for consistency?

@asterite I see now that wouldn't work unless it was returning hashes instead of Array(String), which is a bigger change than I was talking about. Looking at the Ruby interface they switch to returning Row objects when you set headers: true. OTOH it's a bit confusing that you have a CSV::Row class but each_row yields an Array(String).

Just for note, you can also use Iterator#skip to skip header, like:

CSV.each_row(text).skip(1).to_a

And I think CSV.each_row is a lightweight convenient(utility) method for those who don't want to use the full CSV implementation. In that case, returning an array might be reasonable.

@chaniks That skip is nice, thanks.