Automatically detect the column separator in a CSV file

separator.rb

require "csv"

class CSV
  class Separator
    DEFAULT_SEPARATOR = ","

    class Error < StandardError; end

    class DetectionError < Error
      def message
        "unable to detect the separator because there are not enough rows to be certain. At least two rows are required."
      end
    end

    class EmptyFileError < Error
      def message
        "unable to detect the separator because the CSV file is empty. At least two rows are required."
      end
    end

    # Automatically detects the separator in the first 5 lines of any CSV file.
    #
    # A real separator should appear the **same number of times on every line** (since each row has the same number of columns).
    # If a candidate has a consistent, non-zero count across all sampled lines, it's a strong match.
    #
    # @param file_path [Pathname]
    # @param candidates [Array<String>]
    # @param headers_included [Boolean]
    # @return [String] The detected separator
    def self.detect(file_path,  candidates: [ ",", ";", "\t", "|" ], headers_included: true)
      lines = File.foreach(file_path).first(5).select { it.present? }

      # If the CSV file is empty, we're unable to detect the separator.
      raise EmptyFileError if lines.empty?

      # If the CSV file only contains two lines, we're unable to reliably detect the separator.
      raise DetectionError if headers_included && lines.size <= 3 || !headers_included && lines.size <= 2

      scores = candidates.each_with_object({}) do |sep, hash|
        if (counts = lines.map { it.count(sep) }).all? { |c| c > 0 }
          hash[sep] = counts.first
        else
          hash[sep] = 0
        end
      end

      best = scores.max_by { |_, count| count }
      best && best.last > 0 ? best.first : DEFAULT_SEPARATOR
    end
  end
end