Patch 1217

[mephistobooks]

Hello, rivella50-san.

I found one more. It seems that the index of results are unshifted. So there is a difference from C/C++. https://gist.github.com/mephistobooks/7992785

Could you please check the above gist text?

The suggestion of mine for this is correcting a for-statement in ta_func_call() in talib.c.

thank you, mephistobooks

[mephistobooks]

rivella50-san, I'm sorry but patch-1217 still has bugs. Please forget this. (I did not get the meaning of begind and nbelem well..., but it seems there is a problem.)

[rivella50]

Hi there, for MACD you could try with this real world example code:

require 'rubygems'
require 'talib_ruby'
require 'date'
require 'time'
require 'uri'
require 'net/http'

  class Macd

    # ascending ordered data
    def load_stock_history(share_id)
      url_share = "http://ichart.finance.yahoo.com/table.csv?s=SHARE_ID&TODAY&g=d&a=0&b=2&c=2000&ignore=.csv"
      response = Net::HTTP.get_response(URI.parse(url_share.sub('SHARE_ID', share_id)))
      response.body
    end

    def build_daily_chart_items(data, horizon)
      items = []
      lines = data.split("\n")
      (1..(lines.length-1)).each do |i|
        line_items = lines[i].split(',')
        item_date = DateTime.strptime(line_items[0], '%Y-%m-%d')
        break if item_date < horizon
        item = { :date => DateTime.new(item_date.year, item_date.month, item_date.day, 23, 59, 59),
                        :open => line_items[1].to_f, :high => line_items[2].to_f,
                        :low => line_items[3].to_f, :close => line_items[4].to_f }
        items << item
      end
      items.reverse
    end

    # items are asc sorted
    def build_closes(items)
      closes = []

      items.each do |item|
        closes << item[:close]
      end

      return closes
    end

  end

slower_period = 12
longer_period = 26
macd_period   = 9
l = TaLib::Function.new("MACD")
stoch = Macd.new
data =  stoch.load_stock_history("ABBN.VX")
items = stoch.build_daily_chart_items(data, (Date.today.<< 12))

values = stoch.build_closes(items)
values_length = values.length

out_k    = Array.new(values_length)
out_d    = Array.new(values_length)
out_hist = Array.new(values_length)

l.in_real(0, values)
l.out_real(0, out_k)
l.out_real(1, out_d)
l.out_real(2, out_hist)
l.opt_int(0, slower_period)
l.opt_int(1, longer_period)
l.opt_int(2, macd_period)
p "lookback is #{l.lookback}"
l.call(0, values_length-1)

p "k: #{out_k.join(',')}"
p "d: #{out_d.join(',')}"
p "hist: #{out_hist.join(',')}"

For the printed outputs you can see that talib leaves more than one empty slot at the beginning. Since this is filled by talib i wouldn't change that in the wrapper as well. It would be interesting which outputs you get when you use the same input data as in the example above. Thanks for your efforts.

[mephistobooks]

Thank you for your confirming, rivella50-san.

But what I would like to say is:

• TA_MACD and TA_MACDEXT (with EMA) is functionally equivalent by the definition. But they give us different results under the same condition. This is happened only when signal period is 1.
• Output elements are populated into specified output array from the index 0 in C/C++. On the other hand, talib_ruby doesn't.
  • Suppose we have 3000 input data ([0..2999]) and do some analysis for last 1000 data (input data[2000...2999]). In this time, BegIdx become 2000, NBElem is 1000. And the result data are populated into out_SOMEARRAY[0...999] in C/C++.
  • On the other hand, talib_ruby outputs result data from BegIdx(in source out_start at ta_func_call()). So, result is out_SOMEARRAY[2000..2999]. out_SOMEARRAY has 2000 empty slots.
  • That is, TA-Lib don't make empty slots but talib_ruby is making.
  • I would think that this is not good if talib_ruby's purpose is the wrapper of TA-Lib.