Say you have a CSV file with a header as the first row. It contains column names, and you have to parse it. I have often seen code that solves this problem but isn’t as convenient and reliable as I wanted. The solution was unreliable and unmaintainable because it used hard-coded column indexes: 0, 1, 2, and so on. I decided to figure out how to use the provided column names instead of numeric indexes. So welcome aboard.

Simple solution

Let’s say we have this CSV:

Make; Model; Year
Audi; 80; 1994
Audi; A6; 2005
BMW; 740; 2001

The simplest solution, with some disadvantages, will look like this:

CSV.foreach(file_path, :headers => true, :col_sep => ';') do |row|
  Car.create(:make => row[0], :model => row[1], :year => row[2])
end

The disadvantages of this solution are obvious: if the column order changes, the system will behave incorrectly and fail.

Robust solution

See how simple it is to write more reliable code that solves this problem:

CSV.foreach(file_path, :headers => true, :col_sep => ';') do |row|
  Car.create(:make => row['Make'], :model => row['Model'], :year => row['Year'])
end

Note that you cannot get row values using symbols corresponding to the original strings (:Make, :Model or :Year), or even using underscored strings/symbols (‘make’ or :make). To achieve this, you should make a monkey patch that resolves the problem. I won’t provide it here — it’s up to you.

This solution is more reliable — it’s ready for changes in column order, it’s readable for humans, and it’s not sensitive to the number of columns. But there is still a possible failure: if there is a mistake in the CSV header, this solution will fail too. In this case, I can give only one piece of advice: check the header and read the column names before parsing. The CSV.foreach method accepts more parameters; you can find them here.

PS. At first, I reinvented the wheel with code that reads the first row and creates a hash of indexes for each column (an underscored symbol of the column name as the key and the index number as the value), but @avsej suggested this solution to me. Although I didn’t find the solution in examples or official docs, I’ve learned one more lesson here — don’t reinvent the wheel. Thanks to @avsej!