bsearch.rb

# This code has edge cases on it's edge cases. If you use it in production, or even at all, it may eat your cat/dog/hamster

require 'time'

def time_from_string(line)
  m = /\A\[(?<time>\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2} \+\d{4})\]/.match(line)
  if m
    return Time.parse(m[:time])
  end
end

def read_until_time(io)
  loop do
    c = io.getc
    if c == "\n"
      return time_from_string(io.readline)
    end
    raise EOFError if io.size == io.pos
    io.seek(io.pos - c.bytesize - 1)
  end
end

def within_window(time, window)
  time >= window[0] && time <= window[1]
end

# Expects io to be seeked to before the start of the window
def read_data_from_pos(io, window)
  begin
    io.each_line {|ln|
      t = time_from_string(ln)
      if within_window(t, window)
        puts ln
      elsif t > window[1]
        return
      end
    }
  rescue EOFError
  end
end

def search_for_date(filename, window_of_interest)
  io = File.open filename
  window_end = io.size
  window_start = 0

  begin
    loop do
      new_seek = ((window_end - window_start) / 2) + window_start
      io.seek(new_seek)

      # Search back for date if it's not found at current position
      time = read_until_time(io)

      if within_window(time, window_of_interest)
        before_window = nil
        seek_to = io.pos - 1024
        loop do
          io.seek(seek_to)
          time = read_until_time(io)
          if within_window(time, window_of_interest)
            seek_to = seek_to - 1024
          else
            read_data_from_pos(io, window_of_interest)
            return
          end
        end
      elsif time < window_of_interest[0]
        # seek forwards
        window_start = io.pos
      else
        # seek backwards
        window_end = io.pos
      end
    end
  rescue EOFError
    puts "Read to the end of the file but no matching entries"
  end
end

window_of_interest = [Time.parse("2043-01-25 12:35:00"), Time.parse("2043-01-25 12:40:00")]
search_for_date('log', window_of_interest)