examples/async_btgym_workers.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Asynchronous BTgym environment setup.\n",
    "****\n",
    "- This example shows base setup for asynchronious execution of multiply BTgym environment instances.\n",
    "\n",
    "\n",
    "- The main idea is to utilize separate BTgymDataFeedServer process to draw random episode data samples from one main dataset and pass it to every running environment.\n",
    "\n",
    "\n",
    "- This is quite a memory saver since every env. instance holds only small piece of data (single episode). Besides, sampling and dataset management is performed in it's own process."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import sys\n",
    "sys.path.insert(0,'..')\n",
    "\n",
    "import time\n",
    "import multiprocessing\n",
    "\n",
    "from btgym import BTgymEnv, BTgymDataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class Worker(multiprocessing.Process):\n",
    "    env = None\n",
    "    \n",
    "    def __init__(self, env_class, worker_id, num_episodes, env_config):\n",
    "        super(Worker, self).__init__()\n",
    "        self.env_class = env_class\n",
    "        self.num_episodes = num_episodes\n",
    "        self.env_config = env_config\n",
    "        self.worker_id = worker_id\n",
    "        \n",
    "    def run(self):\n",
    "        print('worker_{}: making environment...'.format(self.worker_id))\n",
    "        self.env = self.env_class(**self.env_config)\n",
    "        \n",
    "        for episode in range(self.num_episodes):\n",
    "            print('worker_{}: episode {} started.'.format(self.worker_id, episode))\n",
    "            obs = self.env.reset()\n",
    "            done = False\n",
    "            while not done:\n",
    "                # Just repeat `hold` action:\n",
    "                obs, reward, done, info = self.env.step(0)\n",
    "            # Get and report statistic:\n",
    "            stat = self.env.get_stat() \n",
    "            print(\n",
    "                'worker_{}: episode {} finished, {} steps made within {} seconds.'.\n",
    "                  format(self.worker_id, stat['episode'], stat['length'], stat['runtime'])\n",
    "            )   \n",
    "        self.env.close()\n",
    "        print('worker_{}: environmnt closed.'.format(self.worker_id))\n",
    "        \n",
    "        \n",
    "        \n",
    "     "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[2017-12-18 19:25:28,687] Data_master `reset()` called prior to `reset_data()` with [possibly inconsistent] defaults.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "worker_0: making environment...\n",
      "worker_1: making environment...\n",
      "worker_2: making environment...\n",
      "worker_3: making environment...\n",
      "worker_4: making environment...\n",
      "worker_5: making environment...\n",
      "worker_6: making environment...\n",
      "worker_7: making environment...\n",
      "worker_0: episode 0 started.\n",
      "worker_1: episode 0 started.\n",
      "worker_2: episode 0 started.\n",
      "worker_3: episode 0 started.\n",
      "worker_4: episode 0 started.\n",
      "worker_5: episode 0 started.\n",
      "worker_6: episode 0 started.\n",
      "worker_7: episode 0 started.\n",
      "worker_0: episode 0 finished, 2874 steps made within 0:00:04.970312 seconds.\n",
      "worker_0: episode 1 started.\n",
      "worker_1: episode 0 finished, 2874 steps made within 0:00:04.996122 seconds.\n",
      "worker_1: episode 1 started.\n",
      "worker_2: episode 0 finished, 2874 steps made within 0:00:05.133555 seconds.\n",
      "worker_2: episode 1 started.\n",
      "worker_3: episode 0 finished, 2874 steps made within 0:00:05.132922 seconds.\n",
      "worker_3: episode 1 started.\n",
      "worker_4: episode 0 finished, 2874 steps made within 0:00:05.430693 seconds.\n",
      "worker_4: episode 1 started.\n",
      "worker_6: episode 0 finished, 2874 steps made within 0:00:05.225778 seconds.\n",
      "worker_6: episode 1 started.\n",
      "worker_5: episode 0 finished, 2874 steps made within 0:00:05.371958 seconds.\n",
      "worker_5: episode 1 started.\n",
      "worker_7: episode 0 finished, 2874 steps made within 0:00:05.310113 seconds.\n",
      "worker_7: episode 1 started.\n",
      "worker_0: episode 1 finished, 2874 steps made within 0:00:04.669963 seconds.\n",
      "worker_0: environmnt closed.\n",
      "worker_1: episode 1 finished, 2874 steps made within 0:00:04.599960 seconds.\n",
      "worker_1: environmnt closed.\n",
      "worker_2: episode 1 finished, 2874 steps made within 0:00:04.532229 seconds.\n",
      "...worker_0 has joined.\n",
      "...worker_1 has joined.\n",
      "worker_2: environmnt closed.\n",
      "worker_3: episode 1 finished, 2874 steps made within 0:00:04.537144 seconds.\n",
      "worker_3: environmnt closed.\n",
      "...worker_2 has joined.\n",
      "...worker_3 has joined.\n",
      "worker_4: episode 1 finished, 2874 steps made within 0:00:04.346328 seconds.\n",
      "worker_4: environmnt closed.\n",
      "worker_5: episode 1 finished, 2874 steps made within 0:00:04.337170 seconds.\n",
      "worker_5: environmnt closed.\n",
      "worker_6: episode 1 finished, 2874 steps made within 0:00:04.462971 seconds.\n",
      "worker_6: environmnt closed.\n",
      "worker_7: episode 1 finished, 2874 steps made within 0:00:04.302940 seconds.\n",
      "worker_7: environmnt closed.\n",
      "...worker_4 has joined.\n",
      "...worker_5 has joined.\n",
      "...worker_6 has joined.\n",
      "...worker_7 has joined.\n",
      "data_master: environment closed.\n"
     ]
    }
   ],
   "source": [
    "# Provide data:\n",
    "MyDataset = BTgymDataset(\n",
    "    filename='../examples/data/DAT_ASCII_EURUSD_M1_2016.csv',\n",
    "    start_weekdays=[0, 1, 2, 3, 4],\n",
    "    episode_duration={'days': 1, 'hours': 23, 'minutes': 55},\n",
    "    start_00=False,\n",
    "    time_gap={'hours': 12},\n",
    ")\n",
    "\n",
    "# Setup:\n",
    "num_workers = 8   # set it according to available CPU cores.\n",
    "num_episodes = 2  # individual workplan.\n",
    "base_port = 5000  # worker environment <--> BTgymServer communictaion, individual for every worker.\n",
    "data_port = 4999  # BTgymServer <--> BTgymDataFeedServer communication, same for all workers.\n",
    "\n",
    "workers = []\n",
    "\n",
    "# Make data-master environment. In this example it will only serves as coordinator\n",
    "# to start/stop data server, but can be exploited as full-time worker as well;\n",
    "# just keep it runnung until others exit.\n",
    "data_master = BTgymEnv(\n",
    "    dataset=MyDataset,  # It is the only environment here for which dataset is required:\n",
    "    port=5050,\n",
    "    data_port=data_port,\n",
    "    data_master=True,\n",
    "    connect_timeout=10,  # set server connection timeout to 10 second (default is 60).\n",
    "    verbose=0,\n",
    ")\n",
    "\n",
    "o = data_master.reset() # fake reset() to start data_server_process\n",
    "\n",
    "# Make and launch workers in separate processes:\n",
    "for i in range(num_workers):\n",
    "    # Worker environment configuration:\n",
    "    env_config=dict(\n",
    "        port=base_port + i,\n",
    "        data_port=data_port,\n",
    "        data_master=False,  # This option forces environmnet to seek for datafeed provider server\n",
    "                            # rather to use own dataset. Thus, no dataset is specified for this env.\n",
    "        connect_timeout=10,\n",
    "        verbose=0,\n",
    "    )\n",
    "    \n",
    "    # Make:\n",
    "    worker = Worker(\n",
    "        env_class=BTgymEnv,\n",
    "        worker_id=i,\n",
    "        num_episodes=num_episodes,\n",
    "        env_config=env_config,\n",
    "    )\n",
    "    \n",
    "    # Launch:\n",
    "    worker.daemon = False\n",
    "    worker.start()\n",
    "    workers.append(worker)\n",
    "    # Artificial async, no need for this in real training:\n",
    "    time.sleep(0.1)\n",
    "    \n",
    "# Wait everyone to finish:\n",
    "for worker in workers:\n",
    "    worker.join()\n",
    "    print('...worker_{} has joined.'.format(worker.worker_id))\n",
    "    \n",
    "# shutdown data server :\n",
    "data_master.close()\n",
    "\n",
    "print('data_master: environment closed.')\n",
    "     "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Under the hood:\n",
    "*****\n",
    "...for those who interested in what internal communications are going on: \n",
    "\n",
    "- just set **`verbose=2*`**:\n",
    "\n",
    "\n",
    "*-DEBUG level - do not use that in real life due to data overflood."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[2017-12-18 19:25:53,607] Custom Dataset class used.\n",
      "[2017-12-18 19:25:53,608] Connecting data_server...\n",
      "[2017-12-18 19:25:53,648] DataServer PID: 32495\n",
      "[2017-12-18 19:25:54,352] Loaded 372678 records from <../examples/data/DAT_ASCII_EURUSD_M1_2016.csv>.\n",
      "[2017-12-18 19:25:54,438] Data summary:\n",
      "                open           high            low          close    volume\n",
      "count  372678.000000  372678.000000  372678.000000  372678.000000  372678.0\n",
      "mean        1.107109       1.107198       1.107019       1.107108       0.0\n",
      "std         0.024843       0.024840       0.024847       0.024844       0.0\n",
      "min         1.035250       1.035470       1.035220       1.035220       0.0\n",
      "25%         1.092140       1.092230       1.092040       1.092140       0.0\n",
      "50%         1.113530       1.113610       1.113450       1.113530       0.0\n",
      "75%         1.124710       1.124780       1.124630       1.124710       0.0\n",
      "max         1.161440       1.161600       1.160770       1.161450       0.0\n",
      "[2017-12-18 19:25:54,441] If_sample: data_ready: False, fresh_sample: False\n",
      "[2017-12-18 19:25:54,651] Pinging data_server at: tcp://127.0.0.1:4999 ...\n",
      "[2017-12-18 19:25:54,653] DataServer received <{'ctrl': 'ping!'}>\n",
      "[2017-12-18 19:25:54,655] DataServer sent: {'ctrl': 'waiting for control keys:  <_reset_data>, <_get_data>, <_get_info>, <_stop>.'}\n",
      "[2017-12-18 19:25:54,656] Data_server seems ready with response: <{'ctrl': 'waiting for control keys:  <_reset_data>, <_get_data>, <_get_info>, <_stop>.'}>\n",
      "[2017-12-18 19:25:54,656] If_sample: data_ready: False, fresh_sample: False\n",
      "[2017-12-18 19:25:54,658] DataServer received <{'ctrl': '_get_info'}>\n",
      "[2017-12-18 19:25:54,659] Sending info for #0.\n",
      "[2017-12-18 19:25:54,661] If_sample: data_ready: False, fresh_sample: False\n",
      "[2017-12-18 19:25:54,663] ...done.\n",
      "[2017-12-18 19:25:54,664] Base Cerebro class used. Base Strategy class used.\n",
      "[2017-12-18 19:25:54,673] Inferring `state_raw` high/low values form dataset: 1.035220 / 1.161600.\n",
      "[2017-12-18 19:25:54,674] Obs. shape: OrderedDict([('raw_state', Box(4, 4))])\n",
      "[2017-12-18 19:25:54,715] BTgymServer PID: 32500\n",
      "[2017-12-18 19:25:54,719] BtgymServer: pinging data_server at: tcp://127.0.0.1:4999 ...\n",
      "[2017-12-18 19:25:54,721] DataServer received <{'ctrl': 'ping!'}>\n",
      "[2017-12-18 19:25:54,723] DataServer sent: {'ctrl': 'waiting for control keys:  <_reset_data>, <_get_data>, <_get_info>, <_stop>.'}\n",
      "[2017-12-18 19:25:54,725] If_sample: data_ready: False, fresh_sample: False\n",
      "[2017-12-18 19:25:54,725] BTgymServer: Data_server seems ready with response: <{'ctrl': 'waiting for control keys:  <_reset_data>, <_get_data>, <_get_info>, <_stop>.'}>\n",
      "[2017-12-18 19:25:55,715] Server started, pinging tcp://127.0.0.1:5050 ...\n",
      "[2017-12-18 19:25:55,717] Server Control mode: received <{'ctrl': 'ping!'}>\n",
      "[2017-12-18 19:25:55,719] Server sent: {'ctrl': 'send control keys: <_reset>, <_getstat>, <_render>, <_stop>.'}\n",
      "[2017-12-18 19:25:55,720] Server seems ready with response: <{'ctrl': 'send control keys: <_reset>, <_getstat>, <_render>, <_stop>.'}>\n",
      "[2017-12-18 19:25:55,721] Environment is ready.\n",
      "[2017-12-18 19:25:55,726] Connecting data_server...\n",
      "[2017-12-18 19:25:55,727] Pinging data_server at: tcp://127.0.0.1:4999 ...\n",
      "[2017-12-18 19:25:55,728] DataServer received <{'ctrl': 'ping!'}>\n",
      "[2017-12-18 19:25:55,729] DataServer sent: {'ctrl': 'waiting for control keys:  <_reset_data>, <_get_data>, <_get_info>, <_stop>.'}\n",
      "[2017-12-18 19:25:55,731] Data_server seems ready with response: <{'ctrl': 'waiting for control keys:  <_reset_data>, <_get_data>, <_get_info>, <_stop>.'}>\n",
      "[2017-12-18 19:25:55,731] If_sample: data_ready: False, fresh_sample: False\n",
      "[2017-12-18 19:25:55,733] DataServer received <{'ctrl': '_get_info'}>\n",
      "[2017-12-18 19:25:55,734] Sending info for #0.\n",
      "[2017-12-18 19:25:55,736] If_sample: data_ready: False, fresh_sample: False\n",
      "[2017-12-18 19:25:55,737] ...done.\n",
      "[2017-12-18 19:25:55,739] Base Cerebro class used. Base Strategy class used.\n",
      "[2017-12-18 19:25:55,746] Inferring `state_raw` high/low values form dataset: 1.035220 / 1.161600.\n",
      "[2017-12-18 19:25:55,747] Obs. shape: OrderedDict([('raw_state', Box(4, 4))])\n",
      "[2017-12-18 19:25:55,791] BTgymServer PID: 32505\n",
      "[2017-12-18 19:25:55,795] BtgymServer: pinging data_server at: tcp://127.0.0.1:4999 ...\n",
      "[2017-12-18 19:25:55,797] DataServer received <{'ctrl': 'ping!'}>\n",
      "[2017-12-18 19:25:55,799] DataServer sent: {'ctrl': 'waiting for control keys:  <_reset_data>, <_get_data>, <_get_info>, <_stop>.'}\n",
      "[2017-12-18 19:25:55,801] If_sample: data_ready: False, fresh_sample: False\n",
      "[2017-12-18 19:25:55,801] BTgymServer: Data_server seems ready with response: <{'ctrl': 'waiting for control keys:  <_reset_data>, <_get_data>, <_get_info>, <_stop>.'}>\n",
      "[2017-12-18 19:25:56,790] Server started, pinging tcp://127.0.0.1:5052 ...\n",
      "[2017-12-18 19:25:56,792] Server Control mode: received <{'ctrl': 'ping!'}>\n",
      "[2017-12-18 19:25:56,794] Server sent: {'ctrl': 'send control keys: <_reset>, <_getstat>, <_render>, <_stop>.'}\n",
      "[2017-12-18 19:25:56,796] Server seems ready with response: <{'ctrl': 'send control keys: <_reset>, <_getstat>, <_render>, <_stop>.'}>\n",
      "[2017-12-18 19:25:56,797] Environment is ready.\n",
      "[2017-12-18 19:25:56,799] DataServer received <{'ctrl': '_get_info'}>\n",
      "[2017-12-18 19:25:56,800] Sending info for #0.\n",
      "[2017-12-18 19:25:56,802] If_sample: data_ready: False, fresh_sample: False\n",
      "[2017-12-18 19:25:56,804] Data_master `reset()` called prior to `reset_data()` with [possibly inconsistent] defaults.\n",
      "[2017-12-18 19:25:56,805] Server Control mode: received <{'ctrl': '_done'}>\n",
      "[2017-12-18 19:25:56,807] Server sent: {'ctrl': 'send control keys: <_reset>, <_getstat>, <_render>, <_stop>.'}\n",
      "[2017-12-18 19:25:56,809] FORCE CONTROL MODE attempt: 1.\n",
      "Response: {'ctrl': 'send control keys: <_reset>, <_getstat>, <_render>, <_stop>.'}\n",
      "[2017-12-18 19:25:56,811] DataServer received <{'ctrl': '_reset_data', 'kwargs': {}}>\n",
      "[2017-12-18 19:25:57,610] Loaded 372678 records from <../examples/data/DAT_ASCII_EURUSD_M1_2016.csv>.\n",
      "[2017-12-18 19:25:57,622] DataServer sent: {'ctrl': 'Dataset has been reset with kwargs: {}'}\n",
      "[2017-12-18 19:25:57,624] [_reset_data]: data_is_ready: True\n",
      "[2017-12-18 19:25:57,626] Dataset seems ready with response: <{'ctrl': 'Dataset has been reset with kwargs: {}'}>\n",
      "[2017-12-18 19:25:57,626] If_sample: data_ready: True, fresh_sample: False\n",
      "[2017-12-18 19:25:57,627] Server Control mode: received <{'ctrl': '_done'}>\n",
      "[2017-12-18 19:25:57,628] Maximum sample time duration set to: 1 day, 23:55:00.\n",
      "[2017-12-18 19:25:57,630] Server sent: {'ctrl': 'send control keys: <_reset>, <_getstat>, <_render>, <_stop>.'}\n",
      "[2017-12-18 19:25:57,630] Respective number of steps: 2875.\n",
      "[2017-12-18 19:25:57,632] FORCE CONTROL MODE attempt: 1.\n",
      "Response: {'ctrl': 'send control keys: <_reset>, <_getstat>, <_render>, <_stop>.'}\n",
      "[2017-12-18 19:25:57,632] Maximum allowed data time gap set to: 12:00:00.\n",
      "[2017-12-18 19:25:57,633] Server Control mode: received <{'ctrl': '_reset'}>\n",
      "\n",
      "[2017-12-18 19:25:57,635] Starting episode.\n",
      "[2017-12-18 19:25:57,636] Episode start: 2016-07-19 23:39:00, weekday: 1.\n",
      "[2017-12-18 19:25:57,659] Episode duration: 1 day, 23:58:00.\n",
      "[2017-12-18 19:25:57,661] Total episode time gap: 0:03:00.\n",
      "[2017-12-18 19:25:57,663] Sample accepted.\n",
      "[2017-12-18 19:25:57,665] Episode id: <_btgym_random_sample_2016-07-19 23:39:00>.\n",
      "[2017-12-18 19:25:57,689] Got fresh: episode #0 metadata:\n",
      "{'type': False, 'first_row': 204586, 'trial_num': False, 'sample_num': False}\n",
      "[2017-12-18 19:25:57,693] DataServer received <{'ctrl': '_get_data'}>\n",
      "[2017-12-18 19:25:57,707] Sending episode #0 data {'metadata': {'type': False, 'first_row': 204586, 'trial_num': False, 'sample_num': False}, 'datafeed': <backtrader.feeds.pandafeed.PandasDirectData object at 0x10ba48080>, 'episode_stat':               open         high          low        close  volume\n",
      "count  2875.000000  2875.000000  2875.000000  2875.000000  2875.0\n",
      "mean      1.101559     1.101645     1.101472     1.101559     0.0\n",
      "std       0.001155     0.001159     0.001159     0.001155     0.0\n",
      "min       1.098210     1.098430     1.097950     1.098150     0.0\n",
      "25%       1.100840     1.100930     1.100750     1.100840     0.0\n",
      "50%       1.101510     1.101590     1.101440     1.101520     0.0\n",
      "75%       1.102380     1.102430     1.102320     1.102370     0.0\n",
      "max       1.105650     1.105950     1.104860     1.105660     0.0, 'dataset_stat':                 open           high            low          close    volume\n",
      "count  372678.000000  372678.000000  372678.000000  372678.000000  372678.0\n",
      "mean        1.107109       1.107198       1.107019       1.107108       0.0\n",
      "std         0.024843       0.024840       0.024847       0.024844       0.0\n",
      "min         1.035250       1.035470       1.035220       1.035220       0.0\n",
      "25%         1.092140       1.092230       1.092040       1.092140       0.0\n",
      "50%         1.113530       1.113610       1.113450       1.113530       0.0\n",
      "75%         1.124710       1.124780       1.124630       1.124710       0.0\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "max         1.161440       1.161600       1.160770       1.161450       0.0, 'local_step': 0}.\n",
      "[2017-12-18 19:25:57,713] If_sample: data_ready: True, fresh_sample: False\n",
      "[2017-12-18 19:25:57,717] Maximum sample time duration set to: 1 day, 23:55:00.\n",
      "[2017-12-18 19:25:57,719] Data_server responded with datafeed in about 0.0793910026550293 seconds.\n",
      "[2017-12-18 19:25:57,720] Respective number of steps: 2875.\n",
      "[2017-12-18 19:25:57,722] Maximum allowed data time gap set to: 12:00:00.\n",
      "\n",
      "[2017-12-18 19:25:57,729] Kwargs:\n",
      "{'metadata': {'type': False, 'first_row': 204586, 'trial_num': False, 'sample_num': False}}\n",
      "\n",
      "[2017-12-18 19:25:57,731] Episode start: 2016-02-28 20:29:00, weekday: 6.\n",
      "[2017-12-18 19:25:57,735] Not a good day to start, resampling...\n",
      "[2017-12-18 19:25:57,739] Episode start: 2016-10-20 19:25:00, weekday: 3.\n",
      "[2017-12-18 19:25:57,747] Inner time embedding: 4\n",
      "[2017-12-18 19:25:57,748] Episode duration: 4 days, 0:06:00.\n",
      "[2017-12-18 19:25:57,751] COMM recieved: {'action': 'hold'}\n",
      "[2017-12-18 19:25:57,751] Total episode time gap: 2 days, 0:11:00.\n",
      "[2017-12-18 19:25:57,753] Duration too big, resampling...\n",
      "[2017-12-18 19:25:57,755] Env response checker received:\n",
      "({'raw_state': array([[ 1.10113,  1.10122,  1.10113,  1.10122],\n",
      "       [ 1.10121,  1.10125,  1.10121,  1.10125],\n",
      "       [ 1.10123,  1.10123,  1.10111,  1.10112],\n",
      "       [ 1.10113,  1.10113,  1.1011 ,  1.10111]])}, 0.0, False, [{'step': 1, 'time': datetime.datetime(2016, 7, 19, 23, 42), 'action': 'hold', 'broker_message': '-', 'broker_cash': 10.0, 'broker_value': 10.0, 'drawdown': 0.0, 'max_drawdown': 0.0}])\n",
      "as type: <class 'tuple'>\n",
      "\n",
      "[2017-12-18 19:25:57,757] Episode start: 2016-09-18 18:16:00, weekday: 6.\n",
      "[2017-12-18 19:25:57,757] Server Control mode: received <{'ctrl': '_done'}>\n",
      "[2017-12-18 19:25:57,759] Not a good day to start, resampling...\n",
      "[2017-12-18 19:25:57,759] Server sent: {'ctrl': 'send control keys: <_reset>, <_getstat>, <_render>, <_stop>.'}\n",
      "[2017-12-18 19:25:57,761] FORCE CONTROL MODE attempt: 1.\n",
      "Response: {'ctrl': 'send control keys: <_reset>, <_getstat>, <_render>, <_stop>.'}\n",
      "[2017-12-18 19:25:57,761] Episode start: 2016-07-27 06:20:00, weekday: 2.\n",
      "[2017-12-18 19:25:57,762] Server Control mode: received <{'ctrl': '_reset'}>\n",
      "[2017-12-18 19:25:57,764] Starting episode.\n",
      "[2017-12-18 19:25:57,765] Episode duration: 1 day, 23:56:00.\n",
      "[2017-12-18 19:25:57,767] Total episode time gap: 0:01:00.\n",
      "[2017-12-18 19:25:57,769] Sample accepted.\n",
      "[2017-12-18 19:25:57,771] Episode id: <_btgym_random_sample_2016-07-27 06:20:00>.\n",
      "[2017-12-18 19:25:57,791] Got fresh: episode #1 metadata:\n",
      "{'type': False, 'first_row': 212157, 'trial_num': False, 'sample_num': False}\n",
      "[2017-12-18 19:25:57,795] DataServer received <{'ctrl': '_get_data'}>\n",
      "[2017-12-18 19:25:57,807] Sending episode #1 data {'metadata': {'type': False, 'first_row': 212157, 'trial_num': False, 'sample_num': False}, 'datafeed': <backtrader.feeds.pandafeed.PandasDirectData object at 0x10b9f1ba8>, 'episode_stat':               open         high          low        close  volume\n",
      "count  2875.000000  2875.000000  2875.000000  2875.000000  2875.0\n",
      "mean      1.106485     1.106570     1.106406     1.106488     0.0\n",
      "std       0.003530     0.003522     0.003538     0.003529     0.0\n",
      "min       1.096490     1.097390     1.096110     1.096520     0.0\n",
      "25%       1.105855     1.105940     1.105810     1.105850     0.0\n",
      "50%       1.107620     1.107670     1.107570     1.107620     0.0\n",
      "75%       1.108700     1.108800     1.108610     1.108705     0.0\n",
      "max       1.111660     1.111930     1.111190     1.111680     0.0, 'dataset_stat':                 open           high            low          close    volume\n",
      "count  372678.000000  372678.000000  372678.000000  372678.000000  372678.0\n",
      "mean        1.107109       1.107198       1.107019       1.107108       0.0\n",
      "std         0.024843       0.024840       0.024847       0.024844       0.0\n",
      "min         1.035250       1.035470       1.035220       1.035220       0.0\n",
      "25%         1.092140       1.092230       1.092040       1.092140       0.0\n",
      "50%         1.113530       1.113610       1.113450       1.113530       0.0\n",
      "75%         1.124710       1.124780       1.124630       1.124710       0.0\n",
      "max         1.161440       1.161600       1.160770       1.161450       0.0, 'local_step': 1}.\n",
      "[2017-12-18 19:25:57,812] If_sample: data_ready: True, fresh_sample: False\n",
      "[2017-12-18 19:25:57,817] Data_server responded with datafeed in about 0.04974794387817383 seconds.\n",
      "[2017-12-18 19:25:57,820] Maximum sample time duration set to: 1 day, 23:55:00.\n",
      "[2017-12-18 19:25:57,829] Respective number of steps: 2875.\n",
      "[2017-12-18 19:25:57,831] Kwargs:\n",
      "{'metadata': {'type': False, 'first_row': 212157, 'trial_num': False, 'sample_num': False}}\n",
      "[2017-12-18 19:25:57,832] Maximum allowed data time gap set to: 12:00:00.\n",
      "\n",
      "\n",
      "[2017-12-18 19:25:57,844] Episode start: 2016-09-21 06:39:00, weekday: 2.\n",
      "[2017-12-18 19:25:57,850] Inner time embedding: 4\n",
      "[2017-12-18 19:25:57,858] Episode duration: 1 day, 23:57:00.\n",
      "[2017-12-18 19:25:57,860] COMM recieved: {'action': 'hold'}\n",
      "[2017-12-18 19:25:57,865] Total episode time gap: 0:02:00.\n",
      "[2017-12-18 19:25:57,868] Env response checker received:\n",
      "({'raw_state': array([[ 1.10051,  1.10051,  1.10032,  1.10032],\n",
      "       [ 1.10032,  1.10033,  1.10012,  1.10017],\n",
      "       [ 1.10018,  1.1002 ,  1.10013,  1.10017],\n",
      "       [ 1.10019,  1.10019,  1.1    ,  1.10001]])}, 0.0, False, [{'step': 1, 'time': datetime.datetime(2016, 7, 27, 6, 23), 'action': 'hold', 'broker_message': '-', 'broker_cash': 10.0, 'broker_value': 10.0, 'drawdown': 0.0, 'max_drawdown': 0.0}])\n",
      "as type: <class 'tuple'>\n",
      "[2017-12-18 19:25:57,867] Sample accepted.\n",
      "[2017-12-18 19:25:57,870] Episode id: <_btgym_random_sample_2016-09-21 06:39:00>.\n",
      "[2017-12-18 19:25:57,870] COMM recieved: {'ctrl': '_done'}\n",
      "[2017-12-18 19:25:57,873] FORCE CONTROL MODE attempt: 1.\n",
      "Response: _DONE SIGNAL RECEIVED\n",
      "[2017-12-18 19:25:57,872] RunStop() invoked with -\n",
      "[2017-12-18 19:25:57,895] Got fresh: episode #2 metadata:\n",
      "{'type': False, 'first_row': 269586, 'trial_num': False, 'sample_num': False}\n",
      "[2017-12-18 19:25:58,847] Episode elapsed time: 0:00:01.080951.\n",
      "[2017-12-18 19:25:58,924] Server Control mode: received <{'ctrl': '_done'}>\n",
      "[2017-12-18 19:25:58,926] Server sent: {'ctrl': 'send control keys: <_reset>, <_getstat>, <_render>, <_stop>.'}\n",
      "[2017-12-18 19:25:58,928] FORCE CONTROL MODE attempt: 2.\n",
      "Response: {'ctrl': 'send control keys: <_reset>, <_getstat>, <_render>, <_stop>.'}\n",
      "[2017-12-18 19:25:58,929] Server Control mode: received <{'ctrl': '_stop'}>\n",
      "[2017-12-18 19:25:58,930] Server is exiting.\n",
      "[2017-12-18 19:25:58,932] Server is exiting. Exit code: None\n",
      "[2017-12-18 19:25:58,933] Environment closed.\n",
      "[2017-12-18 19:25:58,934] COMM recieved: {'ctrl': '_done'}\n",
      "[2017-12-18 19:25:58,936] FORCE CONTROL MODE attempt: 1.\n",
      "Response: _DONE SIGNAL RECEIVED\n",
      "[2017-12-18 19:25:58,936] RunStop() invoked with -\n",
      "[2017-12-18 19:25:59,719] Episode elapsed time: 0:00:02.081797.\n",
      "[2017-12-18 19:25:59,809] Server Control mode: received <{'ctrl': '_done'}>\n",
      "[2017-12-18 19:25:59,812] Server sent: {'ctrl': 'send control keys: <_reset>, <_getstat>, <_render>, <_stop>.'}\n",
      "[2017-12-18 19:25:59,814] FORCE CONTROL MODE attempt: 2.\n",
      "Response: {'ctrl': 'send control keys: <_reset>, <_getstat>, <_render>, <_stop>.'}\n",
      "[2017-12-18 19:25:59,815] Server Control mode: received <{'ctrl': '_stop'}>\n",
      "[2017-12-18 19:25:59,817] Server is exiting.\n",
      "[2017-12-18 19:25:59,819] Server is exiting. Exit code: None\n",
      "[2017-12-18 19:25:59,820] DataServer received <{'ctrl': '_stop'}>\n",
      "[2017-12-18 19:25:59,822] {'ctrl': 'DataServer is exiting.'}\n",
      "[2017-12-18 19:25:59,824] {'ctrl': 'DataServer is exiting.'} Exit code: None\n",
      "[2017-12-18 19:25:59,825] Environment closed.\n"
     ]
    }
   ],
   "source": [
    "MyDataset = BTgymDataset(\n",
    "    filename='../examples/data/DAT_ASCII_EURUSD_M1_2016.csv',\n",
    "    start_weekdays=[0, 1, 2, 3, 4],\n",
    "    episode_duration={'days': 1, 'hours': 23, 'minutes': 55},\n",
    "    start_00=False,\n",
    "    time_gap={'hours': 12},\n",
    ")\n",
    "\n",
    "env1 = BTgymEnv(\n",
    "    dataset=MyDataset,\n",
    "    port=5050,\n",
    "    data_port=4999,\n",
    "    data_master=True,\n",
    "    connect_timeout=5,\n",
    "    verbose=2,\n",
    ")\n",
    "\n",
    "env2 = BTgymEnv(\n",
    "    port=5052,\n",
    "    data_port=4999,\n",
    "    data_master=False,\n",
    "    connect_timeout=5,\n",
    "    verbose=2,\n",
    ")\n",
    "\n",
    "_1 = env1.reset()\n",
    "_2 = env2.reset()\n",
    "\n",
    "env2.close()\n",
    "env1.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "dir(data_master)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}