| Total Complexity | 146 |
| Total Lines | 1201 |
| Duplicated Lines | 0 % |
Complex classes like zipline.data.DataPortal often do a lot of different things. To break such a class down, we need to identify a cohesive component within that class. A common approach to find such a component is to look for fields/methods that share the same prefixes, or suffixes.
Once you have determined the fields that belong together, you can apply the Extract Class refactoring. If the component makes sense as a sub-class, Extract Subclass is also a candidate, and is often faster.
| 1 | # |
||
| 65 | class DataPortal(object): |
||
| 66 | def __init__(self, |
||
| 67 | env, |
||
| 68 | sim_params=None, |
||
| 69 | equity_minute_reader=None, |
||
| 70 | minutes_futures_path=None, |
||
| 71 | daily_equities_path=None, |
||
| 72 | adjustment_reader=None, |
||
| 73 | futures_sid_path_func=None): |
||
| 74 | self.env = env |
||
| 75 | |||
| 76 | # This is a bit ugly, but is here for performance reasons. In minute |
||
| 77 | # simulations, we need to very quickly go from dt -> (# of minutes |
||
| 78 | # since Jan 1 2002 9:30 Eastern). |
||
| 79 | # |
||
| 80 | # The clock that heartbeats the simulation has all the necessary |
||
| 81 | # information to do this calculation very quickly. This value is |
||
| 82 | # calculated there, and then set here |
||
| 83 | self.cur_data_offset = 0 |
||
| 84 | |||
| 85 | self.views = {} |
||
| 86 | |||
| 87 | self._daily_equities_path = daily_equities_path |
||
| 88 | self._minutes_futures_path = minutes_futures_path |
||
| 89 | |||
| 90 | self._asset_finder = env.asset_finder |
||
| 91 | |||
| 92 | self._carrays = { |
||
| 93 | 'open': {}, |
||
| 94 | 'high': {}, |
||
| 95 | 'low': {}, |
||
| 96 | 'close': {}, |
||
| 97 | 'volume': {}, |
||
| 98 | 'sid': {}, |
||
| 99 | 'dt': {}, |
||
| 100 | } |
||
| 101 | |||
| 102 | self._adjustment_reader = adjustment_reader |
||
| 103 | |||
| 104 | # caches of sid -> adjustment list |
||
| 105 | self._splits_dict = {} |
||
| 106 | self._mergers_dict = {} |
||
| 107 | self._dividends_dict = {} |
||
| 108 | |||
| 109 | # Cache of sid -> the first trading day of an asset, even if that day |
||
| 110 | # is before 1/2/2002. |
||
| 111 | self._asset_start_dates = {} |
||
| 112 | self._asset_end_dates = {} |
||
| 113 | |||
| 114 | # Handle extra sources, like Fetcher. |
||
| 115 | self._augmented_sources_map = {} |
||
| 116 | self._extra_source_df = None |
||
| 117 | |||
| 118 | self._sim_params = sim_params |
||
| 119 | |||
| 120 | self._futures_sid_path_func = futures_sid_path_func |
||
| 121 | |||
| 122 | self.MINUTE_PRICE_ADJUSTMENT_FACTOR = 0.001 |
||
| 123 | |||
| 124 | if daily_equities_path is not None: |
||
| 125 | self._daily_bar_reader = BcolzDailyBarReader(daily_equities_path) |
||
| 126 | else: |
||
| 127 | self._daily_bar_reader = None |
||
| 128 | |||
| 129 | self._equity_minute_reader = equity_minute_reader |
||
| 130 | |||
| 131 | # The following values are used by _minute_offset to calculate the |
||
| 132 | # index into the minute bcolz date. |
||
| 133 | |||
| 134 | # A lookup of table every minute to the corresponding day, to avoid |
||
| 135 | # calling `.date()` on every lookup. |
||
| 136 | self._minutes_to_day = {} |
||
| 137 | # A map of days (keyed by midnight) to a DatetimeIndex of market |
||
| 138 | # minutes for that day. |
||
| 139 | self._minutes_by_day = {} |
||
| 140 | # A dict of day to the offset into the minute bcolz on which that |
||
| 141 | # days data starts. |
||
| 142 | self._day_offsets = None |
||
| 143 | |||
| 144 | def handle_extra_source(self, source_df): |
||
| 145 | """ |
||
| 146 | Extra sources always have a sid column. |
||
| 147 | |||
| 148 | We expand the given data (by forward filling) to the full range of |
||
| 149 | the simulation dates, so that lookup is fast during simulation. |
||
| 150 | """ |
||
| 151 | if source_df is None: |
||
| 152 | return |
||
| 153 | |||
| 154 | self._extra_source_df = source_df |
||
| 155 | |||
| 156 | # source_df's sid column can either consist of assets we know about |
||
| 157 | # (such as sid(24)) or of assets we don't know about (such as |
||
| 158 | # palladium). |
||
| 159 | # |
||
| 160 | # In both cases, we break up the dataframe into individual dfs |
||
| 161 | # that only contain a single asset's information. ie, if source_df |
||
| 162 | # has data for PALLADIUM and GOLD, we split source_df into two |
||
| 163 | # dataframes, one for each. (same applies if source_df has data for |
||
| 164 | # AAPL and IBM). |
||
| 165 | # |
||
| 166 | # We then take each child df and reindex it to the simulation's date |
||
| 167 | # range by forward-filling missing values. this makes reads simpler. |
||
| 168 | # |
||
| 169 | # Finally, we store the data. For each column, we store a mapping in |
||
| 170 | # self.augmented_sources_map from the column to a dictionary of |
||
| 171 | # asset -> df. In other words, |
||
| 172 | # self.augmented_sources_map['days_to_cover']['AAPL'] gives us the df |
||
| 173 | # holding that data. |
||
| 174 | |||
| 175 | if self._sim_params.emission_rate == "daily": |
||
| 176 | source_date_index = self.env.days_in_range( |
||
| 177 | start=self._sim_params.period_start, |
||
| 178 | end=self._sim_params.period_end |
||
| 179 | ) |
||
| 180 | else: |
||
| 181 | source_date_index = self.env.minutes_for_days_in_range( |
||
| 182 | start=self._sim_params.period_start, |
||
| 183 | end=self._sim_params.period_end |
||
| 184 | ) |
||
| 185 | |||
| 186 | # Break the source_df up into one dataframe per sid. This lets |
||
| 187 | # us (more easily) calculate accurate start/end dates for each sid, |
||
| 188 | # de-dup data, and expand the data to fit the backtest start/end date. |
||
| 189 | grouped_by_sid = source_df.groupby(["sid"]) |
||
| 190 | group_names = grouped_by_sid.groups.keys() |
||
| 191 | group_dict = {} |
||
| 192 | for group_name in group_names: |
||
| 193 | group_dict[group_name] = grouped_by_sid.get_group(group_name) |
||
| 194 | |||
| 195 | for identifier, df in iteritems(group_dict): |
||
| 196 | # Before reindexing, save the earliest and latest dates |
||
| 197 | earliest_date = df.index[0] |
||
| 198 | latest_date = df.index[-1] |
||
| 199 | |||
| 200 | # Since we know this df only contains a single sid, we can safely |
||
| 201 | # de-dupe by the index (dt) |
||
| 202 | df = df.groupby(level=0).last() |
||
| 203 | |||
| 204 | # Reindex the dataframe based on the backtest start/end date. |
||
| 205 | # This makes reads easier during the backtest. |
||
| 206 | df = df.reindex(index=source_date_index, method='ffill') |
||
| 207 | |||
| 208 | if not isinstance(identifier, Asset): |
||
| 209 | # for fake assets we need to store a start/end date |
||
| 210 | self._asset_start_dates[identifier] = earliest_date |
||
| 211 | self._asset_end_dates[identifier] = latest_date |
||
| 212 | |||
| 213 | for col_name in df.columns.difference(['sid']): |
||
| 214 | if col_name not in self._augmented_sources_map: |
||
| 215 | self._augmented_sources_map[col_name] = {} |
||
| 216 | |||
| 217 | self._augmented_sources_map[col_name][identifier] = df |
||
| 218 | |||
| 219 | def _open_minute_file(self, field, asset): |
||
| 220 | sid_str = str(int(asset)) |
||
| 221 | |||
| 222 | try: |
||
| 223 | carray = self._carrays[field][sid_str] |
||
| 224 | except KeyError: |
||
| 225 | carray = self._carrays[field][sid_str] = \ |
||
| 226 | self._get_ctable(asset)[field] |
||
| 227 | |||
| 228 | return carray |
||
| 229 | |||
| 230 | def _get_ctable(self, asset): |
||
| 231 | sid = int(asset) |
||
| 232 | |||
| 233 | if isinstance(asset, Future): |
||
| 234 | if self._futures_sid_path_func is not None: |
||
| 235 | path = self._futures_sid_path_func( |
||
| 236 | self._minutes_futures_path, sid |
||
| 237 | ) |
||
| 238 | else: |
||
| 239 | path = "{0}/{1}.bcolz".format(self._minutes_futures_path, sid) |
||
| 240 | elif isinstance(asset, Equity): |
||
| 241 | if self._equity_minute_reader.sid_path_func is not None: |
||
| 242 | path = self._equity_minute_reader.sid_path_func( |
||
| 243 | self._equity_minute_reader.rootdir, sid |
||
| 244 | ) |
||
| 245 | else: |
||
| 246 | path = "{0}/{1}.bcolz".format( |
||
| 247 | self._equity_minute_reader.rootdir, sid) |
||
| 248 | |||
| 249 | else: |
||
| 250 | # TODO: Figure out if assets should be allowed if neither, and |
||
| 251 | # why this code path is being hit. |
||
| 252 | if self._equity_minute_reader.sid_path_func is not None: |
||
| 253 | path = self._equity_minute_reader.sid_path_func( |
||
| 254 | self._equity_minute_reader.rootdir, sid |
||
| 255 | ) |
||
| 256 | else: |
||
| 257 | path = "{0}/{1}.bcolz".format( |
||
| 258 | self._equity_minute_reader.rootdir, sid) |
||
| 259 | |||
| 260 | return bcolz.open(path, mode='r') |
||
| 261 | |||
| 262 | def get_previous_value(self, asset, field, dt, data_frequency): |
||
| 263 | """ |
||
| 264 | Given an asset and a column and a dt, returns the previous value for |
||
| 265 | the same asset/column pair. If this data portal is in minute mode, |
||
| 266 | it's the previous minute value, otherwise it's the previous day's |
||
| 267 | value. |
||
| 268 | |||
| 269 | Parameters |
||
| 270 | --------- |
||
| 271 | asset : Asset |
||
| 272 | The asset whose data is desired. |
||
| 273 | |||
| 274 | field: string |
||
| 275 | The desired field of the asset. Valid values are "open", |
||
| 276 | "open_price", "high", "low", "close", "close_price", "volume", and |
||
| 277 | "price". |
||
| 278 | |||
| 279 | dt: pd.Timestamp |
||
| 280 | The timestamp from which to go back in time one slot. |
||
| 281 | |||
| 282 | data_frequency: string |
||
| 283 | The frequency of the data to query; i.e. whether the data is |
||
| 284 | 'daily' or 'minute' bars |
||
| 285 | |||
| 286 | Returns |
||
| 287 | ------- |
||
| 288 | The value of the desired field at the desired time. |
||
| 289 | """ |
||
| 290 | if data_frequency == 'daily': |
||
| 291 | prev_dt = self.env.previous_trading_day(dt) |
||
| 292 | elif data_frequency == 'minute': |
||
| 293 | prev_dt = self.env.previous_market_minute(dt) |
||
| 294 | |||
| 295 | return self.get_spot_value(asset, field, prev_dt, data_frequency) |
||
| 296 | |||
| 297 | def _check_extra_sources(self, asset, column, day): |
||
| 298 | # If we have an extra source with a column called "price", only look |
||
| 299 | # at it if it's on something like palladium and not AAPL (since our |
||
| 300 | # own price data always wins when dealing with assets). |
||
| 301 | look_in_augmented_sources = column in self._augmented_sources_map and \ |
||
| 302 | not (column in BASE_FIELDS and isinstance(asset, Asset)) |
||
| 303 | |||
| 304 | if look_in_augmented_sources: |
||
| 305 | # we're being asked for a field in an extra source |
||
| 306 | try: |
||
| 307 | return self._augmented_sources_map[column][asset].\ |
||
| 308 | loc[day, column] |
||
| 309 | except: |
||
| 310 | log.error( |
||
| 311 | "Could not find value for asset={0}, day={1}," |
||
| 312 | "column={2}".format( |
||
| 313 | str(asset), |
||
| 314 | str(day), |
||
| 315 | str(column))) |
||
| 316 | |||
| 317 | raise KeyError |
||
| 318 | |||
| 319 | def get_spot_value(self, asset, field, dt, data_frequency): |
||
| 320 | """ |
||
| 321 | Public API method that returns a scalar value representing the value |
||
| 322 | of the desired asset's field at either the given dt. |
||
| 323 | |||
| 324 | Parameters |
||
| 325 | --------- |
||
| 326 | asset : Asset |
||
| 327 | The asset whose data is desired.gith |
||
| 328 | |||
| 329 | field: string |
||
| 330 | The desired field of the asset. Valid values are "open", |
||
| 331 | "open_price", "high", "low", "close", "close_price", "volume", and |
||
| 332 | "price". |
||
| 333 | |||
| 334 | dt: pd.Timestamp |
||
| 335 | The timestamp for the desired value. |
||
| 336 | |||
| 337 | data_frequency: string |
||
| 338 | The frequency of the data to query; i.e. whether the data is |
||
| 339 | 'daily' or 'minute' bars |
||
| 340 | |||
| 341 | Returns |
||
| 342 | ------- |
||
| 343 | The value of the desired field at the desired time. |
||
| 344 | """ |
||
| 345 | extra_source_val = self._check_extra_sources( |
||
| 346 | asset, |
||
| 347 | field, |
||
| 348 | dt, |
||
| 349 | ) |
||
| 350 | |||
| 351 | if extra_source_val is not None: |
||
| 352 | return extra_source_val |
||
| 353 | |||
| 354 | if field not in BASE_FIELDS: |
||
| 355 | raise KeyError("Invalid column: " + str(field)) |
||
| 356 | |||
| 357 | column_to_use = BASE_FIELDS[field] |
||
| 358 | |||
| 359 | if isinstance(asset, int): |
||
| 360 | asset = self._asset_finder.retrieve_asset(asset) |
||
| 361 | |||
| 362 | self._check_is_currently_alive(asset, dt) |
||
| 363 | |||
| 364 | if data_frequency == "daily": |
||
| 365 | day_to_use = dt |
||
| 366 | day_to_use = normalize_date(day_to_use) |
||
| 367 | return self._get_daily_data(asset, column_to_use, day_to_use) |
||
| 368 | else: |
||
| 369 | if isinstance(asset, Future): |
||
| 370 | return self._get_minute_spot_value_future( |
||
| 371 | asset, column_to_use, dt) |
||
| 372 | else: |
||
| 373 | return self._get_minute_spot_value( |
||
| 374 | asset, column_to_use, dt) |
||
| 375 | |||
| 376 | def _get_minute_spot_value_future(self, asset, column, dt): |
||
| 377 | # Futures bcolz files have 1440 bars per day (24 hours), 7 days a week. |
||
| 378 | # The file attributes contain the "start_dt" and "last_dt" fields, |
||
| 379 | # which represent the time period for this bcolz file. |
||
| 380 | |||
| 381 | # The start_dt is midnight of the first day that this future started |
||
| 382 | # trading. |
||
| 383 | |||
| 384 | # figure out the # of minutes between dt and this asset's start_dt |
||
| 385 | start_date = self._get_asset_start_date(asset) |
||
| 386 | minute_offset = int((dt - start_date).total_seconds() / 60) |
||
| 387 | |||
| 388 | if minute_offset < 0: |
||
| 389 | # asking for a date that is before the asset's start date, no dice |
||
| 390 | return 0.0 |
||
| 391 | |||
| 392 | # then just index into the bcolz carray at that offset |
||
| 393 | carray = self._open_minute_file(column, asset) |
||
| 394 | result = carray[minute_offset] |
||
| 395 | |||
| 396 | # if there's missing data, go backwards until we run out of file |
||
| 397 | while result == 0 and minute_offset > 0: |
||
| 398 | minute_offset -= 1 |
||
| 399 | result = carray[minute_offset] |
||
| 400 | |||
| 401 | if column != 'volume': |
||
| 402 | return result * self.MINUTE_PRICE_ADJUSTMENT_FACTOR |
||
| 403 | else: |
||
| 404 | return result |
||
| 405 | |||
| 406 | def setup_offset_cache(self, minutes_by_day, minutes_to_day): |
||
| 407 | # TODO: This case should not be hit, but is when tests are setup |
||
| 408 | # with data_frequency of daily, but run with minutely. |
||
| 409 | if self._equity_minute_reader is None: |
||
| 410 | return |
||
| 411 | |||
| 412 | self._minutes_to_day = minutes_to_day |
||
| 413 | self._minutes_by_day = minutes_by_day |
||
| 414 | if self._sim_params is not None: |
||
| 415 | start = self._sim_params.trading_days[0] |
||
| 416 | first_trading_day_idx = self._equity_minute_reader.trading_days.\ |
||
| 417 | searchsorted(start) |
||
| 418 | self._day_offsets = { |
||
| 419 | day: (i + first_trading_day_idx) * 390 |
||
| 420 | for i, day in enumerate( |
||
| 421 | self._sim_params.trading_days)} |
||
| 422 | |||
| 423 | def _minute_offset(self, dt): |
||
| 424 | if self._day_offsets is not None: |
||
| 425 | try: |
||
| 426 | day = self._minutes_to_day[dt] |
||
| 427 | minutes = self._minutes_by_day[day] |
||
| 428 | return self._day_offsets[day] + minutes.get_loc(dt) |
||
| 429 | except KeyError: |
||
| 430 | return None |
||
| 431 | |||
| 432 | def _get_minute_spot_value(self, asset, column, dt): |
||
| 433 | # if dt is before the first market minute, minute_index |
||
| 434 | # will be 0. if it's after the last market minute, it'll |
||
| 435 | # be len(minutes_for_day) |
||
| 436 | minute_offset_to_use = self._minute_offset(dt) |
||
| 437 | |||
| 438 | if minute_offset_to_use is None: |
||
| 439 | given_day = pd.Timestamp(dt.date(), tz='utc') |
||
| 440 | day_index = self._equity_minute_reader.trading_days.searchsorted( |
||
| 441 | given_day) |
||
| 442 | |||
| 443 | # if dt is before the first market minute, minute_index |
||
| 444 | # will be 0. if it's after the last market minute, it'll |
||
| 445 | # be len(minutes_for_day) |
||
| 446 | minute_index = self.env.market_minutes_for_day(given_day).\ |
||
| 447 | searchsorted(dt) |
||
| 448 | |||
| 449 | minute_offset_to_use = (day_index * 390) + minute_index |
||
| 450 | |||
| 451 | carray = self._equity_minute_reader._open_minute_file(column, asset) |
||
| 452 | result = carray[minute_offset_to_use] |
||
| 453 | |||
| 454 | if result == 0: |
||
| 455 | # if the given minute doesn't have data, we need to seek |
||
| 456 | # backwards until we find data. This makes the data |
||
| 457 | # forward-filled. |
||
| 458 | |||
| 459 | # get this asset's start date, so that we don't look before it. |
||
| 460 | start_date = self._get_asset_start_date(asset) |
||
| 461 | start_date_idx = self._equity_minute_reader.trading_days.\ |
||
| 462 | searchsorted(start_date) |
||
| 463 | start_day_offset = start_date_idx * 390 |
||
| 464 | |||
| 465 | original_start = minute_offset_to_use |
||
| 466 | |||
| 467 | while result == 0 and minute_offset_to_use > start_day_offset: |
||
| 468 | minute_offset_to_use -= 1 |
||
| 469 | result = carray[minute_offset_to_use] |
||
| 470 | |||
| 471 | # once we've found data, we need to check whether it needs |
||
| 472 | # to be adjusted. |
||
| 473 | if result != 0: |
||
| 474 | minutes = self.env.market_minute_window( |
||
| 475 | start=dt, |
||
| 476 | count=(original_start - minute_offset_to_use + 1), |
||
| 477 | step=-1 |
||
| 478 | ).order() |
||
| 479 | |||
| 480 | # only need to check for adjustments if we've gone back |
||
| 481 | # far enough to cross the day boundary. |
||
| 482 | if minutes[0].date() != minutes[-1].date(): |
||
| 483 | # create a np array of size minutes, fill it all with |
||
| 484 | # the same value. and adjust the array. |
||
| 485 | arr = np.array([result] * len(minutes), |
||
| 486 | dtype=np.float64) |
||
| 487 | self._apply_all_adjustments( |
||
| 488 | data=arr, |
||
| 489 | asset=asset, |
||
| 490 | dts=minutes, |
||
| 491 | field=column |
||
| 492 | ) |
||
| 493 | |||
| 494 | # The first value of the adjusted array is the value |
||
| 495 | # we want. |
||
| 496 | result = arr[0] |
||
| 497 | |||
| 498 | if column != 'volume': |
||
| 499 | return result * self.MINUTE_PRICE_ADJUSTMENT_FACTOR |
||
| 500 | else: |
||
| 501 | return result |
||
| 502 | |||
| 503 | def _get_daily_data(self, asset, column, dt): |
||
| 504 | while True: |
||
| 505 | try: |
||
| 506 | value = self._daily_bar_reader.spot_price(asset, dt, column) |
||
| 507 | if value != -1: |
||
| 508 | return value |
||
| 509 | else: |
||
| 510 | dt -= tradingcalendar.trading_day |
||
| 511 | except NoDataOnDate: |
||
| 512 | return 0 |
||
| 513 | |||
| 514 | def _get_history_daily_window(self, assets, end_dt, bar_count, |
||
| 515 | field_to_use): |
||
| 516 | """ |
||
| 517 | Internal method that returns a dataframe containing history bars |
||
| 518 | of daily frequency for the given sids. |
||
| 519 | """ |
||
| 520 | day_idx = tradingcalendar.trading_days.searchsorted(end_dt.date()) |
||
| 521 | days_for_window = tradingcalendar.trading_days[ |
||
| 522 | (day_idx - bar_count + 1):(day_idx + 1)] |
||
| 523 | |||
| 524 | if len(assets) == 0: |
||
| 525 | return pd.DataFrame(None, |
||
| 526 | index=days_for_window, |
||
| 527 | columns=None) |
||
| 528 | |||
| 529 | data = [] |
||
| 530 | |||
| 531 | for asset in assets: |
||
| 532 | if isinstance(asset, Future): |
||
| 533 | data.append(self._get_history_daily_window_future( |
||
| 534 | asset, days_for_window, end_dt, field_to_use |
||
| 535 | )) |
||
| 536 | else: |
||
| 537 | data.append(self._get_history_daily_window_equity( |
||
| 538 | asset, days_for_window, end_dt, field_to_use |
||
| 539 | )) |
||
| 540 | |||
| 541 | return pd.DataFrame( |
||
| 542 | np.array(data).T, |
||
| 543 | index=days_for_window, |
||
| 544 | columns=assets |
||
| 545 | ) |
||
| 546 | |||
| 547 | def _get_history_daily_window_future(self, asset, days_for_window, |
||
| 548 | end_dt, column): |
||
| 549 | # Since we don't have daily bcolz files for futures (yet), use minute |
||
| 550 | # bars to calculate the daily values. |
||
| 551 | data = [] |
||
| 552 | data_groups = [] |
||
| 553 | |||
| 554 | # get all the minutes for the days NOT including today |
||
| 555 | for day in days_for_window[:-1]: |
||
| 556 | minutes = self.env.market_minutes_for_day(day) |
||
| 557 | |||
| 558 | values_for_day = np.zeros(len(minutes), dtype=np.float64) |
||
| 559 | |||
| 560 | for idx, minute in enumerate(minutes): |
||
| 561 | minute_val = self._get_minute_spot_value_future( |
||
| 562 | asset, column, minute |
||
| 563 | ) |
||
| 564 | |||
| 565 | values_for_day[idx] = minute_val |
||
| 566 | |||
| 567 | data_groups.append(values_for_day) |
||
| 568 | |||
| 569 | # get the minutes for today |
||
| 570 | last_day_minutes = pd.date_range( |
||
| 571 | start=self.env.get_open_and_close(end_dt)[0], |
||
| 572 | end=end_dt, |
||
| 573 | freq="T" |
||
| 574 | ) |
||
| 575 | |||
| 576 | values_for_last_day = np.zeros(len(last_day_minutes), dtype=np.float64) |
||
| 577 | |||
| 578 | for idx, minute in enumerate(last_day_minutes): |
||
| 579 | minute_val = self._get_minute_spot_value_future( |
||
| 580 | asset, column, minute |
||
| 581 | ) |
||
| 582 | |||
| 583 | values_for_last_day[idx] = minute_val |
||
| 584 | |||
| 585 | data_groups.append(values_for_last_day) |
||
| 586 | |||
| 587 | for group in data_groups: |
||
| 588 | if len(group) == 0: |
||
| 589 | continue |
||
| 590 | |||
| 591 | if column == 'volume': |
||
| 592 | data.append(np.sum(group)) |
||
| 593 | elif column == 'open': |
||
| 594 | data.append(group[0]) |
||
| 595 | elif column == 'close': |
||
| 596 | data.append(group[-1]) |
||
| 597 | elif column == 'high': |
||
| 598 | data.append(np.amax(group)) |
||
| 599 | elif column == 'low': |
||
| 600 | data.append(np.amin(group)) |
||
| 601 | |||
| 602 | return data |
||
| 603 | |||
| 604 | def _get_history_daily_window_equity(self, asset, days_for_window, |
||
| 605 | end_dt, field_to_use): |
||
| 606 | sid = int(asset) |
||
| 607 | ends_at_midnight = end_dt.hour == 0 and end_dt.minute == 0 |
||
| 608 | |||
| 609 | # get the start and end dates for this sid |
||
| 610 | end_date = self._get_asset_end_date(asset) |
||
| 611 | |||
| 612 | if ends_at_midnight or (days_for_window[-1] > end_date): |
||
| 613 | # two cases where we use daily data for the whole range: |
||
| 614 | # 1) the history window ends at midnight utc. |
||
| 615 | # 2) the last desired day of the window is after the |
||
| 616 | # last trading day, use daily data for the whole range. |
||
| 617 | return self._get_daily_window_for_sid( |
||
| 618 | asset, |
||
| 619 | field_to_use, |
||
| 620 | days_for_window, |
||
| 621 | extra_slot=False |
||
| 622 | ) |
||
| 623 | else: |
||
| 624 | # for the last day of the desired window, use minute |
||
| 625 | # data and aggregate it. |
||
| 626 | all_minutes_for_day = self.env.market_minutes_for_day( |
||
| 627 | pd.Timestamp(end_dt.date())) |
||
| 628 | |||
| 629 | last_minute_idx = all_minutes_for_day.searchsorted(end_dt) |
||
| 630 | |||
| 631 | # these are the minutes for the partial day |
||
| 632 | minutes_for_partial_day =\ |
||
| 633 | all_minutes_for_day[0:(last_minute_idx + 1)] |
||
| 634 | |||
| 635 | daily_data = self._get_daily_window_for_sid( |
||
| 636 | sid, |
||
| 637 | field_to_use, |
||
| 638 | days_for_window[0:-1] |
||
| 639 | ) |
||
| 640 | |||
| 641 | minute_data = self._get_minute_window_for_equity( |
||
| 642 | sid, |
||
| 643 | field_to_use, |
||
| 644 | minutes_for_partial_day |
||
| 645 | ) |
||
| 646 | |||
| 647 | if field_to_use == 'volume': |
||
| 648 | minute_value = np.sum(minute_data) |
||
| 649 | elif field_to_use == 'open': |
||
| 650 | minute_value = minute_data[0] |
||
| 651 | elif field_to_use == 'close': |
||
| 652 | minute_value = minute_data[-1] |
||
| 653 | elif field_to_use == 'high': |
||
| 654 | minute_value = np.amax(minute_data) |
||
| 655 | elif field_to_use == 'low': |
||
| 656 | minute_value = np.amin(minute_data) |
||
| 657 | |||
| 658 | # append the partial day. |
||
| 659 | daily_data[-1] = minute_value |
||
| 660 | |||
| 661 | return daily_data |
||
| 662 | |||
| 663 | def _get_history_minute_window(self, assets, end_dt, bar_count, |
||
| 664 | field_to_use): |
||
| 665 | """ |
||
| 666 | Internal method that returns a dataframe containing history bars |
||
| 667 | of minute frequency for the given sids. |
||
| 668 | """ |
||
| 669 | # get all the minutes for this window |
||
| 670 | minutes_for_window = self.env.market_minute_window( |
||
| 671 | end_dt, bar_count, step=-1)[::-1] |
||
| 672 | |||
| 673 | first_trading_day = self._equity_minute_reader.first_trading_day |
||
| 674 | |||
| 675 | # but then cut it down to only the minutes after |
||
| 676 | # the first trading day. |
||
| 677 | modified_minutes_for_window = minutes_for_window[ |
||
| 678 | minutes_for_window.slice_indexer(first_trading_day)] |
||
| 679 | |||
| 680 | modified_minutes_length = len(modified_minutes_for_window) |
||
| 681 | |||
| 682 | if modified_minutes_length == 0: |
||
| 683 | raise ValueError("Cannot calculate history window that ends" |
||
| 684 | "before 2002-01-02 14:31 UTC!") |
||
| 685 | |||
| 686 | data = [] |
||
| 687 | bars_to_prepend = 0 |
||
| 688 | nans_to_prepend = None |
||
| 689 | |||
| 690 | if modified_minutes_length < bar_count: |
||
| 691 | first_trading_date = first_trading_day.date() |
||
| 692 | if modified_minutes_for_window[0].date() == first_trading_date: |
||
| 693 | # the beginning of the window goes before our global trading |
||
| 694 | # start date |
||
| 695 | bars_to_prepend = bar_count - modified_minutes_length |
||
| 696 | nans_to_prepend = np.repeat(np.nan, bars_to_prepend) |
||
| 697 | |||
| 698 | if len(assets) == 0: |
||
| 699 | return pd.DataFrame( |
||
| 700 | None, |
||
| 701 | index=modified_minutes_for_window, |
||
| 702 | columns=None |
||
| 703 | ) |
||
| 704 | |||
| 705 | for asset in assets: |
||
| 706 | asset_minute_data = self._get_minute_window_for_asset( |
||
| 707 | asset, |
||
| 708 | field_to_use, |
||
| 709 | modified_minutes_for_window |
||
| 710 | ) |
||
| 711 | |||
| 712 | if bars_to_prepend != 0: |
||
| 713 | asset_minute_data = np.insert(asset_minute_data, 0, |
||
| 714 | nans_to_prepend) |
||
| 715 | |||
| 716 | data.append(asset_minute_data) |
||
| 717 | |||
| 718 | return pd.DataFrame( |
||
| 719 | np.array(data).T, |
||
| 720 | index=minutes_for_window, |
||
| 721 | columns=map(int, assets) |
||
| 722 | ) |
||
| 723 | |||
| 724 | def get_history_window(self, assets, end_dt, bar_count, frequency, field, |
||
| 725 | ffill=True): |
||
| 726 | """ |
||
| 727 | Public API method that returns a dataframe containing the requested |
||
| 728 | history window. Data is fully adjusted. |
||
| 729 | |||
| 730 | Parameters |
||
| 731 | --------- |
||
| 732 | assets : list of zipline.data.Asset objects |
||
| 733 | The assets whose data is desired. |
||
| 734 | |||
| 735 | bar_count: int |
||
| 736 | The number of bars desired. |
||
| 737 | |||
| 738 | frequency: string |
||
| 739 | "1d" or "1m" |
||
| 740 | |||
| 741 | field: string |
||
| 742 | The desired field of the asset. |
||
| 743 | |||
| 744 | ffill: boolean |
||
| 745 | Forward-fill missing values. Only has effect if field |
||
| 746 | is 'price'. |
||
| 747 | |||
| 748 | Returns |
||
| 749 | ------- |
||
| 750 | A dataframe containing the requested data. |
||
| 751 | """ |
||
| 752 | try: |
||
| 753 | field_to_use = BASE_FIELDS[field] |
||
| 754 | except KeyError: |
||
| 755 | raise ValueError("Invalid history field: " + str(field)) |
||
| 756 | |||
| 757 | # sanity check in case sids were passed in |
||
| 758 | assets = [(self.env.asset_finder.retrieve_asset(asset) if |
||
| 759 | isinstance(asset, int) else asset) for asset in assets] |
||
| 760 | |||
| 761 | if frequency == "1d": |
||
| 762 | df = self._get_history_daily_window(assets, end_dt, bar_count, |
||
| 763 | field_to_use) |
||
| 764 | elif frequency == "1m": |
||
| 765 | df = self._get_history_minute_window(assets, end_dt, bar_count, |
||
| 766 | field_to_use) |
||
| 767 | else: |
||
| 768 | raise ValueError("Invalid frequency: {0}".format(frequency)) |
||
| 769 | |||
| 770 | # forward-fill if needed |
||
| 771 | if field == "price" and ffill: |
||
| 772 | df.fillna(method='ffill', inplace=True) |
||
| 773 | |||
| 774 | return df |
||
| 775 | |||
| 776 | def _get_minute_window_for_asset(self, asset, field, minutes_for_window): |
||
| 777 | """ |
||
| 778 | Internal method that gets a window of adjusted minute data for an asset |
||
| 779 | and specified date range. Used to support the history API method for |
||
| 780 | minute bars. |
||
| 781 | |||
| 782 | Missing bars are filled with NaN. |
||
| 783 | |||
| 784 | Parameters |
||
| 785 | ---------- |
||
| 786 | asset : Asset |
||
| 787 | The asset whose data is desired. |
||
| 788 | |||
| 789 | field: string |
||
| 790 | The specific field to return. "open", "high", "close_price", etc. |
||
| 791 | |||
| 792 | minutes_for_window: pd.DateTimeIndex |
||
| 793 | The list of minutes representing the desired window. Each minute |
||
| 794 | is a pd.Timestamp. |
||
| 795 | |||
| 796 | Returns |
||
| 797 | ------- |
||
| 798 | A numpy array with requested values. |
||
| 799 | """ |
||
| 800 | if isinstance(asset, int): |
||
| 801 | asset = self.env.asset_finder.retrieve_asset(asset) |
||
| 802 | |||
| 803 | if isinstance(asset, Future): |
||
| 804 | return self._get_minute_window_for_future(asset, field, |
||
| 805 | minutes_for_window) |
||
| 806 | else: |
||
| 807 | return self._get_minute_window_for_equity(asset, field, |
||
| 808 | minutes_for_window) |
||
| 809 | |||
| 810 | def _get_minute_window_for_future(self, asset, field, minutes_for_window): |
||
| 811 | # THIS IS TEMPORARY. For now, we are only exposing futures within |
||
| 812 | # equity trading hours (9:30 am to 4pm, Eastern). The easiest way to |
||
| 813 | # do this is to simply do a spot lookup for each desired minute. |
||
| 814 | return_data = np.zeros(len(minutes_for_window), dtype=np.float64) |
||
| 815 | for idx, minute in enumerate(minutes_for_window): |
||
| 816 | return_data[idx] = \ |
||
| 817 | self._get_minute_spot_value_future(asset, field, minute) |
||
| 818 | |||
| 819 | # Note: an improvement could be to find the consecutive runs within |
||
| 820 | # minutes_for_window, and use them to read the underlying ctable |
||
| 821 | # more efficiently. |
||
| 822 | |||
| 823 | # Once futures are on 24-hour clock, then we can just grab all the |
||
| 824 | # requested minutes in one shot from the ctable. |
||
| 825 | |||
| 826 | # no adjustments for futures, yay. |
||
| 827 | return return_data |
||
| 828 | |||
| 829 | def _get_minute_window_for_equity(self, asset, field, minutes_for_window): |
||
| 830 | # each sid's minutes are stored in a bcolz file |
||
| 831 | # the bcolz file has 390 bars per day, starting at 1/2/2002, regardless |
||
| 832 | # of when the asset started trading and regardless of half days. |
||
| 833 | # for a half day, the second half is filled with zeroes. |
||
| 834 | |||
| 835 | # find the position of start_dt in the entire timeline, go back |
||
| 836 | # bar_count bars, and that's the unadjusted data |
||
| 837 | raw_data = self._equity_minute_reader._open_minute_file(field, asset) |
||
| 838 | |||
| 839 | start_idx = max( |
||
| 840 | self._equity_minute_reader._find_position_of_minute( |
||
| 841 | minutes_for_window[0]), |
||
| 842 | 0) |
||
| 843 | end_idx = self._equity_minute_reader._find_position_of_minute( |
||
| 844 | minutes_for_window[-1]) + 1 |
||
| 845 | |||
| 846 | if end_idx == 0: |
||
| 847 | # No data to return for minute window. |
||
| 848 | return np.full(len(minutes_for_window), np.nan) |
||
| 849 | |||
| 850 | return_data = np.zeros(len(minutes_for_window), dtype=np.float64) |
||
| 851 | |||
| 852 | data_to_copy = raw_data[start_idx:end_idx] |
||
| 853 | |||
| 854 | num_minutes = len(minutes_for_window) |
||
| 855 | |||
| 856 | # data_to_copy contains all the zeros (from 1pm to 4pm of an early |
||
| 857 | # close). num_minutes is the number of actual trading minutes. if |
||
| 858 | # these two have different lengths, that means that we need to trim |
||
| 859 | # away data due to early closes. |
||
| 860 | if len(data_to_copy) != num_minutes: |
||
| 861 | # get a copy of the minutes in Eastern time, since we depend on |
||
| 862 | # an early close being at 1pm Eastern. |
||
| 863 | eastern_minutes = minutes_for_window.tz_convert("US/Eastern") |
||
| 864 | |||
| 865 | # accumulate a list of indices of the last minute of an early |
||
| 866 | # close day. For example, if data_to_copy starts at 12:55 pm, and |
||
| 867 | # there are five minutes of real data before 180 zeroes, we would |
||
| 868 | # put 5 into last_minute_idx_of_early_close_day, because the fifth |
||
| 869 | # minute is the last "real" minute of the day. |
||
| 870 | last_minute_idx_of_early_close_day = [] |
||
| 871 | for minute_idx, minute_dt in enumerate(eastern_minutes): |
||
| 872 | if minute_idx == (num_minutes - 1): |
||
| 873 | break |
||
| 874 | |||
| 875 | if minute_dt.hour == 13 and minute_dt.minute == 0: |
||
| 876 | next_minute = eastern_minutes[minute_idx + 1] |
||
| 877 | if next_minute.hour != 13: |
||
| 878 | # minute_dt is the last minute of an early close day |
||
| 879 | last_minute_idx_of_early_close_day.append(minute_idx) |
||
| 880 | |||
| 881 | # spin through the list of early close markers, and use them to |
||
| 882 | # chop off 180 minutes at a time from data_to_copy. |
||
| 883 | for idx, early_close_minute_idx in \ |
||
| 884 | enumerate(last_minute_idx_of_early_close_day): |
||
| 885 | early_close_minute_idx -= (180 * idx) |
||
| 886 | data_to_copy = np.delete( |
||
| 887 | data_to_copy, |
||
| 888 | range( |
||
| 889 | early_close_minute_idx + 1, |
||
| 890 | early_close_minute_idx + 181 |
||
| 891 | ) |
||
| 892 | ) |
||
| 893 | |||
| 894 | return_data[0:len(data_to_copy)] = data_to_copy |
||
| 895 | |||
| 896 | self._apply_all_adjustments( |
||
| 897 | return_data, |
||
| 898 | asset, |
||
| 899 | minutes_for_window, |
||
| 900 | field, |
||
| 901 | self.MINUTE_PRICE_ADJUSTMENT_FACTOR |
||
| 902 | ) |
||
| 903 | |||
| 904 | return return_data |
||
| 905 | |||
| 906 | def _apply_all_adjustments(self, data, asset, dts, field, |
||
| 907 | price_adj_factor=1.0): |
||
| 908 | """ |
||
| 909 | Internal method that applies all the necessary adjustments on the |
||
| 910 | given data array. |
||
| 911 | |||
| 912 | The adjustments are: |
||
| 913 | - splits |
||
| 914 | - if field != "volume": |
||
| 915 | - mergers |
||
| 916 | - dividends |
||
| 917 | - * 0.001 |
||
| 918 | - any zero fields replaced with NaN |
||
| 919 | - all values rounded to 3 digits after the decimal point. |
||
| 920 | |||
| 921 | Parameters |
||
| 922 | ---------- |
||
| 923 | data : np.array |
||
| 924 | The data to be adjusted. |
||
| 925 | |||
| 926 | asset: Asset |
||
| 927 | The asset whose data is being adjusted. |
||
| 928 | |||
| 929 | dts: pd.DateTimeIndex |
||
| 930 | The list of minutes or days representing the desired window. |
||
| 931 | |||
| 932 | field: string |
||
| 933 | The field whose values are in the data array. |
||
| 934 | |||
| 935 | price_adj_factor: float |
||
| 936 | Factor with which to adjust OHLC values. |
||
| 937 | Returns |
||
| 938 | ------- |
||
| 939 | None. The data array is modified in place. |
||
| 940 | """ |
||
| 941 | self._apply_adjustments_to_window( |
||
| 942 | self._get_adjustment_list( |
||
| 943 | asset, self._splits_dict, "SPLITS" |
||
| 944 | ), |
||
| 945 | data, |
||
| 946 | dts, |
||
| 947 | field != 'volume' |
||
| 948 | ) |
||
| 949 | |||
| 950 | if field != 'volume': |
||
| 951 | self._apply_adjustments_to_window( |
||
| 952 | self._get_adjustment_list( |
||
| 953 | asset, self._mergers_dict, "MERGERS" |
||
| 954 | ), |
||
| 955 | data, |
||
| 956 | dts, |
||
| 957 | True |
||
| 958 | ) |
||
| 959 | |||
| 960 | self._apply_adjustments_to_window( |
||
| 961 | self._get_adjustment_list( |
||
| 962 | asset, self._dividends_dict, "DIVIDENDS" |
||
| 963 | ), |
||
| 964 | data, |
||
| 965 | dts, |
||
| 966 | True |
||
| 967 | ) |
||
| 968 | |||
| 969 | data *= price_adj_factor |
||
| 970 | |||
| 971 | # if anything is zero, it's a missing bar, so replace it with NaN. |
||
| 972 | # we only want to do this for non-volume fields, because a missing |
||
| 973 | # volume should be 0. |
||
| 974 | data[data == 0] = np.NaN |
||
| 975 | |||
| 976 | np.around(data, 3, out=data) |
||
| 977 | |||
| 978 | def _get_daily_window_for_sid(self, asset, field, days_in_window, |
||
| 979 | extra_slot=True): |
||
| 980 | """ |
||
| 981 | Internal method that gets a window of adjusted daily data for a sid |
||
| 982 | and specified date range. Used to support the history API method for |
||
| 983 | daily bars. |
||
| 984 | |||
| 985 | Parameters |
||
| 986 | ---------- |
||
| 987 | asset : Asset |
||
| 988 | The asset whose data is desired. |
||
| 989 | |||
| 990 | start_dt: pandas.Timestamp |
||
| 991 | The start of the desired window of data. |
||
| 992 | |||
| 993 | bar_count: int |
||
| 994 | The number of days of data to return. |
||
| 995 | |||
| 996 | field: string |
||
| 997 | The specific field to return. "open", "high", "close_price", etc. |
||
| 998 | |||
| 999 | extra_slot: boolean |
||
| 1000 | Whether to allocate an extra slot in the returned numpy array. |
||
| 1001 | This extra slot will hold the data for the last partial day. It's |
||
| 1002 | much better to create it here than to create a copy of the array |
||
| 1003 | later just to add a slot. |
||
| 1004 | |||
| 1005 | Returns |
||
| 1006 | ------- |
||
| 1007 | A numpy array with requested values. Any missing slots filled with |
||
| 1008 | nan. |
||
| 1009 | |||
| 1010 | """ |
||
| 1011 | bar_count = len(days_in_window) |
||
| 1012 | # create an np.array of size bar_count |
||
| 1013 | if extra_slot: |
||
| 1014 | return_array = np.zeros((bar_count + 1,)) |
||
| 1015 | else: |
||
| 1016 | return_array = np.zeros((bar_count,)) |
||
| 1017 | |||
| 1018 | return_array[:] = np.NAN |
||
| 1019 | |||
| 1020 | start_date = self._get_asset_start_date(asset) |
||
| 1021 | end_date = self._get_asset_end_date(asset) |
||
| 1022 | day_slice = days_in_window.slice_indexer(start_date, end_date) |
||
| 1023 | active_days = days_in_window[day_slice] |
||
| 1024 | |||
| 1025 | if active_days.shape[0]: |
||
| 1026 | data = self._daily_bar_reader.history_window(field, |
||
| 1027 | active_days[0], |
||
| 1028 | active_days[-1], |
||
| 1029 | asset) |
||
| 1030 | return_array[day_slice] = data |
||
| 1031 | self._apply_all_adjustments( |
||
| 1032 | return_array, |
||
| 1033 | asset, |
||
| 1034 | active_days, |
||
| 1035 | field, |
||
| 1036 | ) |
||
| 1037 | |||
| 1038 | return return_array |
||
| 1039 | |||
| 1040 | @staticmethod |
||
| 1041 | def _apply_adjustments_to_window(adjustments_list, window_data, |
||
| 1042 | dts_in_window, multiply): |
||
| 1043 | if len(adjustments_list) == 0: |
||
| 1044 | return |
||
| 1045 | |||
| 1046 | # advance idx to the correct spot in the adjustments list, based on |
||
| 1047 | # when the window starts |
||
| 1048 | idx = 0 |
||
| 1049 | |||
| 1050 | while idx < len(adjustments_list) and dts_in_window[0] >\ |
||
| 1051 | adjustments_list[idx][0]: |
||
| 1052 | idx += 1 |
||
| 1053 | |||
| 1054 | # if we've advanced through all the adjustments, then there's nothing |
||
| 1055 | # to do. |
||
| 1056 | if idx == len(adjustments_list): |
||
| 1057 | return |
||
| 1058 | |||
| 1059 | while idx < len(adjustments_list): |
||
| 1060 | adjustment_to_apply = adjustments_list[idx] |
||
| 1061 | |||
| 1062 | if adjustment_to_apply[0] > dts_in_window[-1]: |
||
| 1063 | break |
||
| 1064 | |||
| 1065 | range_end = dts_in_window.searchsorted(adjustment_to_apply[0]) |
||
| 1066 | if multiply: |
||
| 1067 | window_data[0:range_end] *= adjustment_to_apply[1] |
||
| 1068 | else: |
||
| 1069 | window_data[0:range_end] /= adjustment_to_apply[1] |
||
| 1070 | |||
| 1071 | idx += 1 |
||
| 1072 | |||
| 1073 | def _get_adjustment_list(self, asset, adjustments_dict, table_name): |
||
| 1074 | """ |
||
| 1075 | Internal method that returns a list of adjustments for the given sid. |
||
| 1076 | |||
| 1077 | Parameters |
||
| 1078 | ---------- |
||
| 1079 | asset : Asset |
||
| 1080 | The asset for which to return adjustments. |
||
| 1081 | |||
| 1082 | adjustments_dict: dict |
||
| 1083 | A dictionary of sid -> list that is used as a cache. |
||
| 1084 | |||
| 1085 | table_name: string |
||
| 1086 | The table that contains this data in the adjustments db. |
||
| 1087 | |||
| 1088 | Returns |
||
| 1089 | ------- |
||
| 1090 | adjustments: list |
||
| 1091 | A list of [multiplier, pd.Timestamp], earliest first |
||
| 1092 | |||
| 1093 | """ |
||
| 1094 | if self._adjustment_reader is None: |
||
| 1095 | return [] |
||
| 1096 | |||
| 1097 | sid = int(asset) |
||
| 1098 | |||
| 1099 | try: |
||
| 1100 | adjustments = adjustments_dict[sid] |
||
| 1101 | except KeyError: |
||
| 1102 | adjustments = adjustments_dict[sid] = self._adjustment_reader.\ |
||
| 1103 | get_adjustments_for_sid(table_name, sid) |
||
| 1104 | |||
| 1105 | return adjustments |
||
| 1106 | |||
| 1107 | def _check_is_currently_alive(self, asset, dt): |
||
| 1108 | sid = int(asset) |
||
| 1109 | |||
| 1110 | if sid not in self._asset_start_dates: |
||
| 1111 | self._get_asset_start_date(asset) |
||
| 1112 | |||
| 1113 | start_date = self._asset_start_dates[sid] |
||
| 1114 | if self._asset_start_dates[sid] > dt: |
||
| 1115 | raise NoTradeDataAvailableTooEarly( |
||
| 1116 | sid=sid, |
||
| 1117 | dt=normalize_date(dt), |
||
| 1118 | start_dt=start_date |
||
| 1119 | ) |
||
| 1120 | |||
| 1121 | end_date = self._asset_end_dates[sid] |
||
| 1122 | if self._asset_end_dates[sid] < dt: |
||
| 1123 | raise NoTradeDataAvailableTooLate( |
||
| 1124 | sid=sid, |
||
| 1125 | dt=normalize_date(dt), |
||
| 1126 | end_dt=end_date |
||
| 1127 | ) |
||
| 1128 | |||
| 1129 | def _get_asset_start_date(self, asset): |
||
| 1130 | self._ensure_asset_dates(asset) |
||
| 1131 | return self._asset_start_dates[asset] |
||
| 1132 | |||
| 1133 | def _get_asset_end_date(self, asset): |
||
| 1134 | self._ensure_asset_dates(asset) |
||
| 1135 | return self._asset_end_dates[asset] |
||
| 1136 | |||
| 1137 | def _ensure_asset_dates(self, asset): |
||
| 1138 | sid = int(asset) |
||
| 1139 | |||
| 1140 | if sid not in self._asset_start_dates: |
||
| 1141 | self._asset_start_dates[sid] = asset.start_date |
||
| 1142 | self._asset_end_dates[sid] = asset.end_date |
||
| 1143 | |||
| 1144 | def get_splits(self, sids, dt): |
||
| 1145 | """ |
||
| 1146 | Returns any splits for the given sids and the given dt. |
||
| 1147 | |||
| 1148 | Parameters |
||
| 1149 | ---------- |
||
| 1150 | sids : list |
||
| 1151 | Sids for which we want splits. |
||
| 1152 | |||
| 1153 | dt: pd.Timestamp |
||
| 1154 | The date for which we are checking for splits. Note: this is |
||
| 1155 | expected to be midnight UTC. |
||
| 1156 | |||
| 1157 | Returns |
||
| 1158 | ------- |
||
| 1159 | list: List of splits, where each split is a (sid, ratio) tuple. |
||
| 1160 | """ |
||
| 1161 | if self._adjustment_reader is None or len(sids) == 0: |
||
| 1162 | return {} |
||
| 1163 | |||
| 1164 | # convert dt to # of seconds since epoch, because that's what we use |
||
| 1165 | # in the adjustments db |
||
| 1166 | seconds = int(dt.value / 1e9) |
||
| 1167 | |||
| 1168 | splits = self._adjustment_reader.conn.execute( |
||
| 1169 | "SELECT sid, ratio FROM SPLITS WHERE effective_date = ?", |
||
| 1170 | (seconds,)).fetchall() |
||
| 1171 | |||
| 1172 | sids_set = set(sids) |
||
| 1173 | splits = [split for split in splits if split[0] in sids_set] |
||
| 1174 | |||
| 1175 | return splits |
||
| 1176 | |||
| 1177 | def get_stock_dividends(self, sid, trading_days): |
||
| 1178 | """ |
||
| 1179 | Returns all the stock dividends for a specific sid that occur |
||
| 1180 | in the given trading range. |
||
| 1181 | |||
| 1182 | Parameters |
||
| 1183 | ---------- |
||
| 1184 | sid: int |
||
| 1185 | The asset whose stock dividends should be returned. |
||
| 1186 | |||
| 1187 | trading_days: pd.DatetimeIndex |
||
| 1188 | The trading range. |
||
| 1189 | |||
| 1190 | Returns |
||
| 1191 | ------- |
||
| 1192 | list: A list of objects with all relevant attributes populated. |
||
| 1193 | All timestamp fields are converted to pd.Timestamps. |
||
| 1194 | """ |
||
| 1195 | |||
| 1196 | if self._adjustment_reader is None: |
||
| 1197 | return [] |
||
| 1198 | |||
| 1199 | if len(trading_days) == 0: |
||
| 1200 | return [] |
||
| 1201 | |||
| 1202 | start_dt = trading_days[0].value / 1e9 |
||
| 1203 | end_dt = trading_days[-1].value / 1e9 |
||
| 1204 | |||
| 1205 | dividends = self._adjustment_reader.conn.execute( |
||
| 1206 | "SELECT * FROM stock_dividend_payouts WHERE sid = ? AND " |
||
| 1207 | "ex_date > ? AND pay_date < ?", (int(sid), start_dt, end_dt,)).\ |
||
| 1208 | fetchall() |
||
| 1209 | |||
| 1210 | dividend_info = [] |
||
| 1211 | for dividend_tuple in dividends: |
||
| 1212 | dividend_info.append({ |
||
| 1213 | "declared_date": dividend_tuple[1], |
||
| 1214 | "ex_date": pd.Timestamp(dividend_tuple[2], unit="s"), |
||
| 1215 | "pay_date": pd.Timestamp(dividend_tuple[3], unit="s"), |
||
| 1216 | "payment_sid": dividend_tuple[4], |
||
| 1217 | "ratio": dividend_tuple[5], |
||
| 1218 | "record_date": pd.Timestamp(dividend_tuple[6], unit="s"), |
||
| 1219 | "sid": dividend_tuple[7] |
||
| 1220 | }) |
||
| 1221 | |||
| 1222 | return dividend_info |
||
| 1223 | |||
| 1224 | def contains(self, asset, field): |
||
| 1225 | return field in BASE_FIELDS or \ |
||
| 1226 | (field in self._augmented_sources_map and |
||
| 1227 | asset in self._augmented_sources_map[field]) |
||
| 1228 | |||
| 1229 | def get_fetcher_assets(self, day): |
||
| 1230 | """ |
||
| 1231 | Returns a list of assets for the current date, as defined by the |
||
| 1232 | fetcher data. |
||
| 1233 | |||
| 1234 | Notes |
||
| 1235 | ----- |
||
| 1236 | Data is forward-filled. If there is no fetcher data defined for day |
||
| 1237 | N, we use day N-1's data (if available, otherwise we keep going back). |
||
| 1238 | |||
| 1239 | Returns |
||
| 1240 | ------- |
||
| 1241 | list: a list of Asset objects. |
||
| 1242 | """ |
||
| 1243 | # return a list of assets for the current date, as defined by the |
||
| 1244 | # fetcher source |
||
| 1245 | if self._extra_source_df is None: |
||
| 1246 | return [] |
||
| 1247 | |||
| 1248 | if day in self._extra_source_df.index: |
||
| 1249 | date_to_use = day |
||
| 1250 | else: |
||
| 1251 | # current day isn't in the fetcher df, go back the last |
||
| 1252 | # available day |
||
| 1253 | idx = self._extra_source_df.index.searchsorted(day) |
||
| 1254 | if idx == 0: |
||
| 1255 | return [] |
||
| 1256 | |||
| 1257 | date_to_use = self._extra_source_df.index[idx - 1] |
||
| 1258 | |||
| 1259 | asset_list = self._extra_source_df.loc[date_to_use]["sid"] |
||
| 1260 | |||
| 1261 | # make sure they're actually assets |
||
| 1262 | asset_list = [asset for asset in asset_list |
||
| 1263 | if isinstance(asset, Asset)] |
||
| 1264 | |||
| 1265 | return asset_list |
||
| 1266 |