Total Complexity | 144 |
Total Lines | 1190 |
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 | # |
||
49 | class DataPortal(object): |
||
50 | def __init__(self, |
||
51 | env, |
||
52 | equity_daily_reader=None, |
||
53 | equity_minute_reader=None, |
||
54 | future_daily_reader=None, |
||
55 | future_minute_reader=None, |
||
56 | adjustment_reader=None): |
||
57 | self.env = env |
||
58 | |||
59 | # This is a bit ugly, but is here for performance reasons. In minute |
||
60 | # simulations, we need to very quickly go from dt -> (# of minutes |
||
61 | # since Jan 1 2002 9:30 Eastern). |
||
62 | # |
||
63 | # The clock that heartbeats the simulation has all the necessary |
||
64 | # information to do this calculation very quickly. This value is |
||
65 | # calculated there, and then set here |
||
66 | self.cur_data_offset = 0 |
||
67 | |||
68 | self.views = {} |
||
69 | |||
70 | self._asset_finder = env.asset_finder |
||
71 | |||
72 | self._carrays = { |
||
73 | 'open': {}, |
||
74 | 'high': {}, |
||
75 | 'low': {}, |
||
76 | 'close': {}, |
||
77 | 'volume': {}, |
||
78 | 'sid': {}, |
||
79 | 'dt': {}, |
||
80 | } |
||
81 | |||
82 | self._adjustment_reader = adjustment_reader |
||
83 | |||
84 | # caches of sid -> adjustment list |
||
85 | self._splits_dict = {} |
||
86 | self._mergers_dict = {} |
||
87 | self._dividends_dict = {} |
||
88 | |||
89 | # Cache of sid -> the first trading day of an asset, even if that day |
||
90 | # is before 1/2/2002. |
||
91 | self._asset_start_dates = {} |
||
92 | self._asset_end_dates = {} |
||
93 | |||
94 | # Handle extra sources, like Fetcher. |
||
95 | self._augmented_sources_map = {} |
||
96 | self._extra_source_df = None |
||
97 | |||
98 | self.MINUTE_PRICE_ADJUSTMENT_FACTOR = 0.001 |
||
99 | |||
100 | self._equity_daily_reader = equity_daily_reader |
||
101 | self._equity_minute_reader = equity_minute_reader |
||
102 | self._future_daily_reader = future_daily_reader |
||
103 | self._future_minute_reader = future_minute_reader |
||
104 | |||
105 | # The following values are used by _minute_offset to calculate the |
||
106 | # index into the minute bcolz date. |
||
107 | |||
108 | # A lookup of table every minute to the corresponding day, to avoid |
||
109 | # calling `.date()` on every lookup. |
||
110 | self._minutes_to_day = {} |
||
111 | # A map of days (keyed by midnight) to a DatetimeIndex of market |
||
112 | # minutes for that day. |
||
113 | self._minutes_by_day = {} |
||
114 | # A dict of day to the offset into the minute bcolz on which that |
||
115 | # days data starts. |
||
116 | self._day_offsets = None |
||
117 | |||
118 | def handle_extra_source(self, source_df, sim_params): |
||
119 | """ |
||
120 | Extra sources always have a sid column. |
||
121 | |||
122 | We expand the given data (by forward filling) to the full range of |
||
123 | the simulation dates, so that lookup is fast during simulation. |
||
124 | """ |
||
125 | if source_df is None: |
||
126 | return |
||
127 | |||
128 | self._extra_source_df = source_df |
||
129 | |||
130 | # source_df's sid column can either consist of assets we know about |
||
131 | # (such as sid(24)) or of assets we don't know about (such as |
||
132 | # palladium). |
||
133 | # |
||
134 | # In both cases, we break up the dataframe into individual dfs |
||
135 | # that only contain a single asset's information. ie, if source_df |
||
136 | # has data for PALLADIUM and GOLD, we split source_df into two |
||
137 | # dataframes, one for each. (same applies if source_df has data for |
||
138 | # AAPL and IBM). |
||
139 | # |
||
140 | # We then take each child df and reindex it to the simulation's date |
||
141 | # range by forward-filling missing values. this makes reads simpler. |
||
142 | # |
||
143 | # Finally, we store the data. For each column, we store a mapping in |
||
144 | # self.augmented_sources_map from the column to a dictionary of |
||
145 | # asset -> df. In other words, |
||
146 | # self.augmented_sources_map['days_to_cover']['AAPL'] gives us the df |
||
147 | # holding that data. |
||
148 | |||
149 | if sim_params.emission_rate == "daily": |
||
150 | source_date_index = self.env.days_in_range( |
||
151 | start=sim_params.period_start, |
||
152 | end=sim_params.period_end |
||
153 | ) |
||
154 | else: |
||
155 | source_date_index = self.env.minutes_for_days_in_range( |
||
156 | start=sim_params.period_start, |
||
157 | end=sim_params.period_end |
||
158 | ) |
||
159 | |||
160 | # Break the source_df up into one dataframe per sid. This lets |
||
161 | # us (more easily) calculate accurate start/end dates for each sid, |
||
162 | # de-dup data, and expand the data to fit the backtest start/end date. |
||
163 | grouped_by_sid = source_df.groupby(["sid"]) |
||
164 | group_names = grouped_by_sid.groups.keys() |
||
165 | group_dict = {} |
||
166 | for group_name in group_names: |
||
167 | group_dict[group_name] = grouped_by_sid.get_group(group_name) |
||
168 | |||
169 | for identifier, df in iteritems(group_dict): |
||
170 | # Before reindexing, save the earliest and latest dates |
||
171 | earliest_date = df.index[0] |
||
172 | latest_date = df.index[-1] |
||
173 | |||
174 | # Since we know this df only contains a single sid, we can safely |
||
175 | # de-dupe by the index (dt) |
||
176 | df = df.groupby(level=0).last() |
||
177 | |||
178 | # Reindex the dataframe based on the backtest start/end date. |
||
179 | # This makes reads easier during the backtest. |
||
180 | df = df.reindex(index=source_date_index, method='ffill') |
||
181 | |||
182 | if not isinstance(identifier, Asset): |
||
183 | # for fake assets we need to store a start/end date |
||
184 | self._asset_start_dates[identifier] = earliest_date |
||
185 | self._asset_end_dates[identifier] = latest_date |
||
186 | |||
187 | for col_name in df.columns.difference(['sid']): |
||
188 | if col_name not in self._augmented_sources_map: |
||
189 | self._augmented_sources_map[col_name] = {} |
||
190 | |||
191 | self._augmented_sources_map[col_name][identifier] = df |
||
192 | |||
193 | def _open_minute_file(self, field, asset): |
||
194 | sid_str = str(int(asset)) |
||
195 | |||
196 | try: |
||
197 | carray = self._carrays[field][sid_str] |
||
198 | except KeyError: |
||
199 | carray = self._carrays[field][sid_str] = \ |
||
200 | self._get_ctable(asset)[field] |
||
201 | |||
202 | return carray |
||
203 | |||
204 | def _get_ctable(self, asset): |
||
205 | sid = int(asset) |
||
206 | |||
207 | if isinstance(asset, Future): |
||
208 | if self._future_minute_reader.sid_path_func is not None: |
||
209 | path = self._future_minute_reader.sid_path_func( |
||
210 | self._future_minute_reader.rootdir, sid |
||
211 | ) |
||
212 | else: |
||
213 | path = "{0}/{1}.bcolz".format( |
||
214 | self._future_minute_reader.rootdir, sid) |
||
215 | elif isinstance(asset, Equity): |
||
216 | if self._equity_minute_reader.sid_path_func is not None: |
||
217 | path = self._equity_minute_reader.sid_path_func( |
||
218 | self._equity_minute_reader.rootdir, sid |
||
219 | ) |
||
220 | else: |
||
221 | path = "{0}/{1}.bcolz".format( |
||
222 | self._equity_minute_reader.rootdir, sid) |
||
223 | |||
224 | else: |
||
225 | # TODO: Figure out if assets should be allowed if neither, and |
||
226 | # why this code path is being hit. |
||
227 | if self._equity_minute_reader.sid_path_func is not None: |
||
228 | path = self._equity_minute_reader.sid_path_func( |
||
229 | self._equity_minute_reader.rootdir, sid |
||
230 | ) |
||
231 | else: |
||
232 | path = "{0}/{1}.bcolz".format( |
||
233 | self._equity_minute_reader.rootdir, sid) |
||
234 | |||
235 | return bcolz.open(path, mode='r') |
||
236 | |||
237 | def get_previous_value(self, asset, field, dt, data_frequency): |
||
238 | """ |
||
239 | Given an asset and a column and a dt, returns the previous value for |
||
240 | the same asset/column pair. If this data portal is in minute mode, |
||
241 | it's the previous minute value, otherwise it's the previous day's |
||
242 | value. |
||
243 | |||
244 | Parameters |
||
245 | --------- |
||
246 | asset : Asset |
||
247 | The asset whose data is desired. |
||
248 | |||
249 | field: string |
||
250 | The desired field of the asset. Valid values are "open", |
||
251 | "open_price", "high", "low", "close", "close_price", "volume", and |
||
252 | "price". |
||
253 | |||
254 | dt: pd.Timestamp |
||
255 | The timestamp from which to go back in time one slot. |
||
256 | |||
257 | data_frequency: string |
||
258 | The frequency of the data to query; i.e. whether the data is |
||
259 | 'daily' or 'minute' bars |
||
260 | |||
261 | Returns |
||
262 | ------- |
||
263 | The value of the desired field at the desired time. |
||
264 | """ |
||
265 | if data_frequency == 'daily': |
||
266 | prev_dt = self.env.previous_trading_day(dt) |
||
267 | elif data_frequency == 'minute': |
||
268 | prev_dt = self.env.previous_market_minute(dt) |
||
269 | |||
270 | return self.get_spot_value(asset, field, prev_dt, data_frequency) |
||
271 | |||
272 | def _check_extra_sources(self, asset, column, day): |
||
273 | # If we have an extra source with a column called "price", only look |
||
274 | # at it if it's on something like palladium and not AAPL (since our |
||
275 | # own price data always wins when dealing with assets). |
||
276 | look_in_augmented_sources = column in self._augmented_sources_map and \ |
||
277 | not (column in BASE_FIELDS and isinstance(asset, Asset)) |
||
278 | |||
279 | if look_in_augmented_sources: |
||
280 | # we're being asked for a field in an extra source |
||
281 | try: |
||
282 | return self._augmented_sources_map[column][asset].\ |
||
283 | loc[day, column] |
||
284 | except: |
||
285 | log.error( |
||
286 | "Could not find value for asset={0}, day={1}," |
||
287 | "column={2}".format( |
||
288 | str(asset), |
||
289 | str(day), |
||
290 | str(column))) |
||
291 | |||
292 | raise KeyError |
||
293 | |||
294 | def get_spot_value(self, asset, field, dt, data_frequency): |
||
295 | """ |
||
296 | Public API method that returns a scalar value representing the value |
||
297 | of the desired asset's field at either the given dt. |
||
298 | |||
299 | Parameters |
||
300 | --------- |
||
301 | asset : Asset |
||
302 | The asset whose data is desired.gith |
||
303 | |||
304 | field: string |
||
305 | The desired field of the asset. Valid values are "open", |
||
306 | "open_price", "high", "low", "close", "close_price", "volume", and |
||
307 | "price". |
||
308 | |||
309 | dt: pd.Timestamp |
||
310 | The timestamp for the desired value. |
||
311 | |||
312 | data_frequency: string |
||
313 | The frequency of the data to query; i.e. whether the data is |
||
314 | 'daily' or 'minute' bars |
||
315 | |||
316 | Returns |
||
317 | ------- |
||
318 | The value of the desired field at the desired time. |
||
319 | """ |
||
320 | extra_source_val = self._check_extra_sources( |
||
321 | asset, |
||
322 | field, |
||
323 | dt, |
||
324 | ) |
||
325 | |||
326 | if extra_source_val is not None: |
||
327 | return extra_source_val |
||
328 | |||
329 | if field not in BASE_FIELDS: |
||
330 | raise KeyError("Invalid column: " + str(field)) |
||
331 | |||
332 | column_to_use = BASE_FIELDS[field] |
||
333 | |||
334 | if isinstance(asset, int): |
||
335 | asset = self._asset_finder.retrieve_asset(asset) |
||
336 | |||
337 | self._check_is_currently_alive(asset, dt) |
||
338 | |||
339 | if data_frequency == "daily": |
||
340 | day_to_use = dt |
||
341 | day_to_use = normalize_date(day_to_use) |
||
342 | return self._get_daily_data(asset, column_to_use, day_to_use) |
||
343 | else: |
||
344 | if isinstance(asset, Future): |
||
345 | return self._get_minute_spot_value_future( |
||
346 | asset, column_to_use, dt) |
||
347 | else: |
||
348 | return self._get_minute_spot_value( |
||
349 | asset, column_to_use, dt) |
||
350 | |||
351 | def _get_minute_spot_value_future(self, asset, column, dt): |
||
352 | # Futures bcolz files have 1440 bars per day (24 hours), 7 days a week. |
||
353 | # The file attributes contain the "start_dt" and "last_dt" fields, |
||
354 | # which represent the time period for this bcolz file. |
||
355 | |||
356 | # The start_dt is midnight of the first day that this future started |
||
357 | # trading. |
||
358 | |||
359 | # figure out the # of minutes between dt and this asset's start_dt |
||
360 | start_date = self._get_asset_start_date(asset) |
||
361 | minute_offset = int((dt - start_date).total_seconds() / 60) |
||
362 | |||
363 | if minute_offset < 0: |
||
364 | # asking for a date that is before the asset's start date, no dice |
||
365 | return 0.0 |
||
366 | |||
367 | # then just index into the bcolz carray at that offset |
||
368 | carray = self._open_minute_file(column, asset) |
||
369 | result = carray[minute_offset] |
||
370 | |||
371 | # if there's missing data, go backwards until we run out of file |
||
372 | while result == 0 and minute_offset > 0: |
||
373 | minute_offset -= 1 |
||
374 | result = carray[minute_offset] |
||
375 | |||
376 | if column != 'volume': |
||
377 | return result * self.MINUTE_PRICE_ADJUSTMENT_FACTOR |
||
378 | else: |
||
379 | return result |
||
380 | |||
381 | def setup_offset_cache(self, minutes_by_day, minutes_to_day, trading_days): |
||
382 | # TODO: This case should not be hit, but is when tests are setup |
||
383 | # with data_frequency of daily, but run with minutely. |
||
384 | if self._equity_minute_reader is None: |
||
385 | return |
||
386 | |||
387 | self._minutes_to_day = minutes_to_day |
||
388 | self._minutes_by_day = minutes_by_day |
||
389 | start = trading_days[0] |
||
390 | first_trading_day_idx = self._equity_minute_reader.trading_days.\ |
||
391 | searchsorted(start) |
||
392 | self._day_offsets = { |
||
393 | day: (i + first_trading_day_idx) * 390 |
||
394 | for i, day in enumerate(trading_days)} |
||
395 | |||
396 | def _minute_offset(self, dt): |
||
397 | if self._day_offsets is not None: |
||
398 | try: |
||
399 | day = self._minutes_to_day[dt] |
||
400 | minutes = self._minutes_by_day[day] |
||
401 | return self._day_offsets[day] + minutes.get_loc(dt) |
||
402 | except KeyError: |
||
403 | return None |
||
404 | |||
405 | def _get_minute_spot_value(self, asset, column, dt): |
||
406 | # if dt is before the first market minute, minute_index |
||
407 | # will be 0. if it's after the last market minute, it'll |
||
408 | # be len(minutes_for_day) |
||
409 | minute_offset_to_use = self._minute_offset(dt) |
||
410 | |||
411 | if minute_offset_to_use is None: |
||
412 | given_day = pd.Timestamp(dt.date(), tz='utc') |
||
413 | day_index = self._equity_minute_reader.trading_days.searchsorted( |
||
414 | given_day) |
||
415 | |||
416 | # if dt is before the first market minute, minute_index |
||
417 | # will be 0. if it's after the last market minute, it'll |
||
418 | # be len(minutes_for_day) |
||
419 | minute_index = self.env.market_minutes_for_day(given_day).\ |
||
420 | searchsorted(dt) |
||
421 | |||
422 | minute_offset_to_use = (day_index * 390) + minute_index |
||
423 | |||
424 | carray = self._equity_minute_reader._open_minute_file(column, asset) |
||
425 | result = carray[minute_offset_to_use] |
||
426 | |||
427 | if result == 0: |
||
428 | # if the given minute doesn't have data, we need to seek |
||
429 | # backwards until we find data. This makes the data |
||
430 | # forward-filled. |
||
431 | |||
432 | # get this asset's start date, so that we don't look before it. |
||
433 | start_date = self._get_asset_start_date(asset) |
||
434 | start_date_idx = self._equity_minute_reader.trading_days.\ |
||
435 | searchsorted(start_date) |
||
436 | start_day_offset = start_date_idx * 390 |
||
437 | |||
438 | original_start = minute_offset_to_use |
||
439 | |||
440 | while result == 0 and minute_offset_to_use > start_day_offset: |
||
441 | minute_offset_to_use -= 1 |
||
442 | result = carray[minute_offset_to_use] |
||
443 | |||
444 | # once we've found data, we need to check whether it needs |
||
445 | # to be adjusted. |
||
446 | if result != 0: |
||
447 | minutes = self.env.market_minute_window( |
||
448 | start=dt, |
||
449 | count=(original_start - minute_offset_to_use + 1), |
||
450 | step=-1 |
||
451 | ).order() |
||
452 | |||
453 | # only need to check for adjustments if we've gone back |
||
454 | # far enough to cross the day boundary. |
||
455 | if minutes[0].date() != minutes[-1].date(): |
||
456 | # create a np array of size minutes, fill it all with |
||
457 | # the same value. and adjust the array. |
||
458 | arr = np.array([result] * len(minutes), |
||
459 | dtype=np.float64) |
||
460 | self._apply_all_adjustments( |
||
461 | data=arr, |
||
462 | asset=asset, |
||
463 | dts=minutes, |
||
464 | field=column |
||
465 | ) |
||
466 | |||
467 | # The first value of the adjusted array is the value |
||
468 | # we want. |
||
469 | result = arr[0] |
||
470 | |||
471 | if column != 'volume': |
||
472 | return result * self.MINUTE_PRICE_ADJUSTMENT_FACTOR |
||
473 | else: |
||
474 | return result |
||
475 | |||
476 | def _get_daily_data(self, asset, column, dt): |
||
477 | while True: |
||
478 | try: |
||
479 | value = self._equity_daily_reader.spot_price(asset, dt, column) |
||
480 | if value != -1: |
||
481 | return value |
||
482 | else: |
||
483 | dt -= tradingcalendar.trading_day |
||
484 | except NoDataOnDate: |
||
485 | return 0 |
||
486 | |||
487 | def _get_history_daily_window(self, assets, end_dt, bar_count, |
||
488 | field_to_use): |
||
489 | """ |
||
490 | Internal method that returns a dataframe containing history bars |
||
491 | of daily frequency for the given sids. |
||
492 | """ |
||
493 | day_idx = tradingcalendar.trading_days.searchsorted(end_dt.date()) |
||
494 | days_for_window = tradingcalendar.trading_days[ |
||
495 | (day_idx - bar_count + 1):(day_idx + 1)] |
||
496 | |||
497 | if len(assets) == 0: |
||
498 | return pd.DataFrame(None, |
||
499 | index=days_for_window, |
||
500 | columns=None) |
||
501 | |||
502 | data = [] |
||
503 | |||
504 | for asset in assets: |
||
505 | if isinstance(asset, Future): |
||
506 | data.append(self._get_history_daily_window_future( |
||
507 | asset, days_for_window, end_dt, field_to_use |
||
508 | )) |
||
509 | else: |
||
510 | data.append(self._get_history_daily_window_equity( |
||
511 | asset, days_for_window, end_dt, field_to_use |
||
512 | )) |
||
513 | |||
514 | return pd.DataFrame( |
||
515 | np.array(data).T, |
||
516 | index=days_for_window, |
||
517 | columns=assets |
||
518 | ) |
||
519 | |||
520 | def _get_history_daily_window_future(self, asset, days_for_window, |
||
521 | end_dt, column): |
||
522 | # Since we don't have daily bcolz files for futures (yet), use minute |
||
523 | # bars to calculate the daily values. |
||
524 | data = [] |
||
525 | data_groups = [] |
||
526 | |||
527 | # get all the minutes for the days NOT including today |
||
528 | for day in days_for_window[:-1]: |
||
529 | minutes = self.env.market_minutes_for_day(day) |
||
530 | |||
531 | values_for_day = np.zeros(len(minutes), dtype=np.float64) |
||
532 | |||
533 | for idx, minute in enumerate(minutes): |
||
534 | minute_val = self._get_minute_spot_value_future( |
||
535 | asset, column, minute |
||
536 | ) |
||
537 | |||
538 | values_for_day[idx] = minute_val |
||
539 | |||
540 | data_groups.append(values_for_day) |
||
541 | |||
542 | # get the minutes for today |
||
543 | last_day_minutes = pd.date_range( |
||
544 | start=self.env.get_open_and_close(end_dt)[0], |
||
545 | end=end_dt, |
||
546 | freq="T" |
||
547 | ) |
||
548 | |||
549 | values_for_last_day = np.zeros(len(last_day_minutes), dtype=np.float64) |
||
550 | |||
551 | for idx, minute in enumerate(last_day_minutes): |
||
552 | minute_val = self._get_minute_spot_value_future( |
||
553 | asset, column, minute |
||
554 | ) |
||
555 | |||
556 | values_for_last_day[idx] = minute_val |
||
557 | |||
558 | data_groups.append(values_for_last_day) |
||
559 | |||
560 | for group in data_groups: |
||
561 | if len(group) == 0: |
||
562 | continue |
||
563 | |||
564 | if column == 'volume': |
||
565 | data.append(np.sum(group)) |
||
566 | elif column == 'open': |
||
567 | data.append(group[0]) |
||
568 | elif column == 'close': |
||
569 | data.append(group[-1]) |
||
570 | elif column == 'high': |
||
571 | data.append(np.amax(group)) |
||
572 | elif column == 'low': |
||
573 | data.append(np.amin(group)) |
||
574 | |||
575 | return data |
||
576 | |||
577 | def _get_history_daily_window_equity(self, asset, days_for_window, |
||
578 | end_dt, field_to_use): |
||
579 | sid = int(asset) |
||
580 | ends_at_midnight = end_dt.hour == 0 and end_dt.minute == 0 |
||
581 | |||
582 | # get the start and end dates for this sid |
||
583 | end_date = self._get_asset_end_date(asset) |
||
584 | |||
585 | if ends_at_midnight or (days_for_window[-1] > end_date): |
||
586 | # two cases where we use daily data for the whole range: |
||
587 | # 1) the history window ends at midnight utc. |
||
588 | # 2) the last desired day of the window is after the |
||
589 | # last trading day, use daily data for the whole range. |
||
590 | return self._get_daily_window_for_sid( |
||
591 | asset, |
||
592 | field_to_use, |
||
593 | days_for_window, |
||
594 | extra_slot=False |
||
595 | ) |
||
596 | else: |
||
597 | # for the last day of the desired window, use minute |
||
598 | # data and aggregate it. |
||
599 | all_minutes_for_day = self.env.market_minutes_for_day( |
||
600 | pd.Timestamp(end_dt.date())) |
||
601 | |||
602 | last_minute_idx = all_minutes_for_day.searchsorted(end_dt) |
||
603 | |||
604 | # these are the minutes for the partial day |
||
605 | minutes_for_partial_day =\ |
||
606 | all_minutes_for_day[0:(last_minute_idx + 1)] |
||
607 | |||
608 | daily_data = self._get_daily_window_for_sid( |
||
609 | sid, |
||
610 | field_to_use, |
||
611 | days_for_window[0:-1] |
||
612 | ) |
||
613 | |||
614 | minute_data = self._get_minute_window_for_equity( |
||
615 | sid, |
||
616 | field_to_use, |
||
617 | minutes_for_partial_day |
||
618 | ) |
||
619 | |||
620 | if field_to_use == 'volume': |
||
621 | minute_value = np.sum(minute_data) |
||
622 | elif field_to_use == 'open': |
||
623 | minute_value = minute_data[0] |
||
624 | elif field_to_use == 'close': |
||
625 | minute_value = minute_data[-1] |
||
626 | elif field_to_use == 'high': |
||
627 | minute_value = np.amax(minute_data) |
||
628 | elif field_to_use == 'low': |
||
629 | minute_value = np.amin(minute_data) |
||
630 | |||
631 | # append the partial day. |
||
632 | daily_data[-1] = minute_value |
||
633 | |||
634 | return daily_data |
||
635 | |||
636 | def _get_history_minute_window(self, assets, end_dt, bar_count, |
||
637 | field_to_use): |
||
638 | """ |
||
639 | Internal method that returns a dataframe containing history bars |
||
640 | of minute frequency for the given sids. |
||
641 | """ |
||
642 | # get all the minutes for this window |
||
643 | minutes_for_window = self.env.market_minute_window( |
||
644 | end_dt, bar_count, step=-1)[::-1] |
||
645 | |||
646 | first_trading_day = self._equity_minute_reader.first_trading_day |
||
647 | |||
648 | # but then cut it down to only the minutes after |
||
649 | # the first trading day. |
||
650 | modified_minutes_for_window = minutes_for_window[ |
||
651 | minutes_for_window.slice_indexer(first_trading_day)] |
||
652 | |||
653 | modified_minutes_length = len(modified_minutes_for_window) |
||
654 | |||
655 | if modified_minutes_length == 0: |
||
656 | raise ValueError("Cannot calculate history window that ends" |
||
657 | "before 2002-01-02 14:31 UTC!") |
||
658 | |||
659 | data = [] |
||
660 | bars_to_prepend = 0 |
||
661 | nans_to_prepend = None |
||
662 | |||
663 | if modified_minutes_length < bar_count: |
||
664 | first_trading_date = first_trading_day.date() |
||
665 | if modified_minutes_for_window[0].date() == first_trading_date: |
||
666 | # the beginning of the window goes before our global trading |
||
667 | # start date |
||
668 | bars_to_prepend = bar_count - modified_minutes_length |
||
669 | nans_to_prepend = np.repeat(np.nan, bars_to_prepend) |
||
670 | |||
671 | if len(assets) == 0: |
||
672 | return pd.DataFrame( |
||
673 | None, |
||
674 | index=modified_minutes_for_window, |
||
675 | columns=None |
||
676 | ) |
||
677 | |||
678 | for asset in assets: |
||
679 | asset_minute_data = self._get_minute_window_for_asset( |
||
680 | asset, |
||
681 | field_to_use, |
||
682 | modified_minutes_for_window |
||
683 | ) |
||
684 | |||
685 | if bars_to_prepend != 0: |
||
686 | asset_minute_data = np.insert(asset_minute_data, 0, |
||
687 | nans_to_prepend) |
||
688 | |||
689 | data.append(asset_minute_data) |
||
690 | |||
691 | return pd.DataFrame( |
||
692 | np.array(data).T, |
||
693 | index=minutes_for_window, |
||
694 | columns=assets |
||
695 | ) |
||
696 | |||
697 | def get_history_window(self, assets, end_dt, bar_count, frequency, field, |
||
698 | ffill=True): |
||
699 | """ |
||
700 | Public API method that returns a dataframe containing the requested |
||
701 | history window. Data is fully adjusted. |
||
702 | |||
703 | Parameters |
||
704 | --------- |
||
705 | assets : list of zipline.data.Asset objects |
||
706 | The assets whose data is desired. |
||
707 | |||
708 | bar_count: int |
||
709 | The number of bars desired. |
||
710 | |||
711 | frequency: string |
||
712 | "1d" or "1m" |
||
713 | |||
714 | field: string |
||
715 | The desired field of the asset. |
||
716 | |||
717 | ffill: boolean |
||
718 | Forward-fill missing values. Only has effect if field |
||
719 | is 'price'. |
||
720 | |||
721 | Returns |
||
722 | ------- |
||
723 | A dataframe containing the requested data. |
||
724 | """ |
||
725 | try: |
||
726 | field_to_use = BASE_FIELDS[field] |
||
727 | except KeyError: |
||
728 | raise ValueError("Invalid history field: " + str(field)) |
||
729 | |||
730 | # sanity check in case sids were passed in |
||
731 | assets = [(self.env.asset_finder.retrieve_asset(asset) if |
||
732 | isinstance(asset, int) else asset) for asset in assets] |
||
733 | |||
734 | if frequency == "1d": |
||
735 | df = self._get_history_daily_window(assets, end_dt, bar_count, |
||
736 | field_to_use) |
||
737 | elif frequency == "1m": |
||
738 | df = self._get_history_minute_window(assets, end_dt, bar_count, |
||
739 | field_to_use) |
||
740 | else: |
||
741 | raise ValueError("Invalid frequency: {0}".format(frequency)) |
||
742 | |||
743 | # forward-fill if needed |
||
744 | if field == "price" and ffill: |
||
745 | df.fillna(method='ffill', inplace=True) |
||
746 | |||
747 | return df |
||
748 | |||
749 | def _get_minute_window_for_asset(self, asset, field, minutes_for_window): |
||
750 | """ |
||
751 | Internal method that gets a window of adjusted minute data for an asset |
||
752 | and specified date range. Used to support the history API method for |
||
753 | minute bars. |
||
754 | |||
755 | Missing bars are filled with NaN. |
||
756 | |||
757 | Parameters |
||
758 | ---------- |
||
759 | asset : Asset |
||
760 | The asset whose data is desired. |
||
761 | |||
762 | field: string |
||
763 | The specific field to return. "open", "high", "close_price", etc. |
||
764 | |||
765 | minutes_for_window: pd.DateTimeIndex |
||
766 | The list of minutes representing the desired window. Each minute |
||
767 | is a pd.Timestamp. |
||
768 | |||
769 | Returns |
||
770 | ------- |
||
771 | A numpy array with requested values. |
||
772 | """ |
||
773 | if isinstance(asset, int): |
||
774 | asset = self.env.asset_finder.retrieve_asset(asset) |
||
775 | |||
776 | if isinstance(asset, Future): |
||
777 | return self._get_minute_window_for_future(asset, field, |
||
778 | minutes_for_window) |
||
779 | else: |
||
780 | return self._get_minute_window_for_equity(asset, field, |
||
781 | minutes_for_window) |
||
782 | |||
783 | def _get_minute_window_for_future(self, asset, field, minutes_for_window): |
||
784 | # THIS IS TEMPORARY. For now, we are only exposing futures within |
||
785 | # equity trading hours (9:30 am to 4pm, Eastern). The easiest way to |
||
786 | # do this is to simply do a spot lookup for each desired minute. |
||
787 | return_data = np.zeros(len(minutes_for_window), dtype=np.float64) |
||
788 | for idx, minute in enumerate(minutes_for_window): |
||
789 | return_data[idx] = \ |
||
790 | self._get_minute_spot_value_future(asset, field, minute) |
||
791 | |||
792 | # Note: an improvement could be to find the consecutive runs within |
||
793 | # minutes_for_window, and use them to read the underlying ctable |
||
794 | # more efficiently. |
||
795 | |||
796 | # Once futures are on 24-hour clock, then we can just grab all the |
||
797 | # requested minutes in one shot from the ctable. |
||
798 | |||
799 | # no adjustments for futures, yay. |
||
800 | return return_data |
||
801 | |||
802 | def _get_minute_window_for_equity(self, asset, field, minutes_for_window): |
||
803 | # each sid's minutes are stored in a bcolz file |
||
804 | # the bcolz file has 390 bars per day, starting at 1/2/2002, regardless |
||
805 | # of when the asset started trading and regardless of half days. |
||
806 | # for a half day, the second half is filled with zeroes. |
||
807 | |||
808 | # find the position of start_dt in the entire timeline, go back |
||
809 | # bar_count bars, and that's the unadjusted data |
||
810 | raw_data = self._equity_minute_reader._open_minute_file(field, asset) |
||
811 | |||
812 | start_idx = max( |
||
813 | self._equity_minute_reader._find_position_of_minute( |
||
814 | minutes_for_window[0]), |
||
815 | 0) |
||
816 | end_idx = self._equity_minute_reader._find_position_of_minute( |
||
817 | minutes_for_window[-1]) + 1 |
||
818 | |||
819 | if end_idx == 0: |
||
820 | # No data to return for minute window. |
||
821 | return np.full(len(minutes_for_window), np.nan) |
||
822 | |||
823 | return_data = np.zeros(len(minutes_for_window), dtype=np.float64) |
||
824 | |||
825 | data_to_copy = raw_data[start_idx:end_idx] |
||
826 | |||
827 | num_minutes = len(minutes_for_window) |
||
828 | |||
829 | # data_to_copy contains all the zeros (from 1pm to 4pm of an early |
||
830 | # close). num_minutes is the number of actual trading minutes. if |
||
831 | # these two have different lengths, that means that we need to trim |
||
832 | # away data due to early closes. |
||
833 | if len(data_to_copy) != num_minutes: |
||
834 | # get a copy of the minutes in Eastern time, since we depend on |
||
835 | # an early close being at 1pm Eastern. |
||
836 | eastern_minutes = minutes_for_window.tz_convert("US/Eastern") |
||
837 | |||
838 | # accumulate a list of indices of the last minute of an early |
||
839 | # close day. For example, if data_to_copy starts at 12:55 pm, and |
||
840 | # there are five minutes of real data before 180 zeroes, we would |
||
841 | # put 5 into last_minute_idx_of_early_close_day, because the fifth |
||
842 | # minute is the last "real" minute of the day. |
||
843 | last_minute_idx_of_early_close_day = [] |
||
844 | for minute_idx, minute_dt in enumerate(eastern_minutes): |
||
845 | if minute_idx == (num_minutes - 1): |
||
846 | break |
||
847 | |||
848 | if minute_dt.hour == 13 and minute_dt.minute == 0: |
||
849 | next_minute = eastern_minutes[minute_idx + 1] |
||
850 | if next_minute.hour != 13: |
||
851 | # minute_dt is the last minute of an early close day |
||
852 | last_minute_idx_of_early_close_day.append(minute_idx) |
||
853 | |||
854 | # spin through the list of early close markers, and use them to |
||
855 | # chop off 180 minutes at a time from data_to_copy. |
||
856 | for idx, early_close_minute_idx in \ |
||
857 | enumerate(last_minute_idx_of_early_close_day): |
||
858 | early_close_minute_idx -= (180 * idx) |
||
859 | data_to_copy = np.delete( |
||
860 | data_to_copy, |
||
861 | range( |
||
862 | early_close_minute_idx + 1, |
||
863 | early_close_minute_idx + 181 |
||
864 | ) |
||
865 | ) |
||
866 | |||
867 | return_data[0:len(data_to_copy)] = data_to_copy |
||
868 | |||
869 | self._apply_all_adjustments( |
||
870 | return_data, |
||
871 | asset, |
||
872 | minutes_for_window, |
||
873 | field, |
||
874 | self.MINUTE_PRICE_ADJUSTMENT_FACTOR |
||
875 | ) |
||
876 | |||
877 | return return_data |
||
878 | |||
879 | def _apply_all_adjustments(self, data, asset, dts, field, |
||
880 | price_adj_factor=1.0): |
||
881 | """ |
||
882 | Internal method that applies all the necessary adjustments on the |
||
883 | given data array. |
||
884 | |||
885 | The adjustments are: |
||
886 | - splits |
||
887 | - if field != "volume": |
||
888 | - mergers |
||
889 | - dividends |
||
890 | - * 0.001 |
||
891 | - any zero fields replaced with NaN |
||
892 | - all values rounded to 3 digits after the decimal point. |
||
893 | |||
894 | Parameters |
||
895 | ---------- |
||
896 | data : np.array |
||
897 | The data to be adjusted. |
||
898 | |||
899 | asset: Asset |
||
900 | The asset whose data is being adjusted. |
||
901 | |||
902 | dts: pd.DateTimeIndex |
||
903 | The list of minutes or days representing the desired window. |
||
904 | |||
905 | field: string |
||
906 | The field whose values are in the data array. |
||
907 | |||
908 | price_adj_factor: float |
||
909 | Factor with which to adjust OHLC values. |
||
910 | Returns |
||
911 | ------- |
||
912 | None. The data array is modified in place. |
||
913 | """ |
||
914 | self._apply_adjustments_to_window( |
||
915 | self._get_adjustment_list( |
||
916 | asset, self._splits_dict, "SPLITS" |
||
917 | ), |
||
918 | data, |
||
919 | dts, |
||
920 | field != 'volume' |
||
921 | ) |
||
922 | |||
923 | if field != 'volume': |
||
924 | self._apply_adjustments_to_window( |
||
925 | self._get_adjustment_list( |
||
926 | asset, self._mergers_dict, "MERGERS" |
||
927 | ), |
||
928 | data, |
||
929 | dts, |
||
930 | True |
||
931 | ) |
||
932 | |||
933 | self._apply_adjustments_to_window( |
||
934 | self._get_adjustment_list( |
||
935 | asset, self._dividends_dict, "DIVIDENDS" |
||
936 | ), |
||
937 | data, |
||
938 | dts, |
||
939 | True |
||
940 | ) |
||
941 | |||
942 | data *= price_adj_factor |
||
943 | |||
944 | # if anything is zero, it's a missing bar, so replace it with NaN. |
||
945 | # we only want to do this for non-volume fields, because a missing |
||
946 | # volume should be 0. |
||
947 | data[data == 0] = np.NaN |
||
948 | |||
949 | np.around(data, 3, out=data) |
||
950 | |||
951 | def _get_daily_window_for_sid(self, asset, field, days_in_window, |
||
952 | extra_slot=True): |
||
953 | """ |
||
954 | Internal method that gets a window of adjusted daily data for a sid |
||
955 | and specified date range. Used to support the history API method for |
||
956 | daily bars. |
||
957 | |||
958 | Parameters |
||
959 | ---------- |
||
960 | asset : Asset |
||
961 | The asset whose data is desired. |
||
962 | |||
963 | start_dt: pandas.Timestamp |
||
964 | The start of the desired window of data. |
||
965 | |||
966 | bar_count: int |
||
967 | The number of days of data to return. |
||
968 | |||
969 | field: string |
||
970 | The specific field to return. "open", "high", "close_price", etc. |
||
971 | |||
972 | extra_slot: boolean |
||
973 | Whether to allocate an extra slot in the returned numpy array. |
||
974 | This extra slot will hold the data for the last partial day. It's |
||
975 | much better to create it here than to create a copy of the array |
||
976 | later just to add a slot. |
||
977 | |||
978 | Returns |
||
979 | ------- |
||
980 | A numpy array with requested values. Any missing slots filled with |
||
981 | nan. |
||
982 | |||
983 | """ |
||
984 | bar_count = len(days_in_window) |
||
985 | # create an np.array of size bar_count |
||
986 | if extra_slot: |
||
987 | return_array = np.zeros((bar_count + 1,)) |
||
988 | else: |
||
989 | return_array = np.zeros((bar_count,)) |
||
990 | |||
991 | return_array[:] = np.NAN |
||
992 | |||
993 | start_date = self._get_asset_start_date(asset) |
||
994 | end_date = self._get_asset_end_date(asset) |
||
995 | day_slice = days_in_window.slice_indexer(start_date, end_date) |
||
996 | active_days = days_in_window[day_slice] |
||
997 | |||
998 | if active_days.shape[0]: |
||
999 | data = self._equity_daily_reader.history_window(field, |
||
1000 | active_days[0], |
||
1001 | active_days[-1], |
||
1002 | asset) |
||
1003 | return_array[day_slice] = data |
||
1004 | self._apply_all_adjustments( |
||
1005 | return_array, |
||
1006 | asset, |
||
1007 | active_days, |
||
1008 | field, |
||
1009 | ) |
||
1010 | |||
1011 | return return_array |
||
1012 | |||
1013 | @staticmethod |
||
1014 | def _apply_adjustments_to_window(adjustments_list, window_data, |
||
1015 | dts_in_window, multiply): |
||
1016 | if len(adjustments_list) == 0: |
||
1017 | return |
||
1018 | |||
1019 | # advance idx to the correct spot in the adjustments list, based on |
||
1020 | # when the window starts |
||
1021 | idx = 0 |
||
1022 | |||
1023 | while idx < len(adjustments_list) and dts_in_window[0] >\ |
||
1024 | adjustments_list[idx][0]: |
||
1025 | idx += 1 |
||
1026 | |||
1027 | # if we've advanced through all the adjustments, then there's nothing |
||
1028 | # to do. |
||
1029 | if idx == len(adjustments_list): |
||
1030 | return |
||
1031 | |||
1032 | while idx < len(adjustments_list): |
||
1033 | adjustment_to_apply = adjustments_list[idx] |
||
1034 | |||
1035 | if adjustment_to_apply[0] > dts_in_window[-1]: |
||
1036 | break |
||
1037 | |||
1038 | range_end = dts_in_window.searchsorted(adjustment_to_apply[0]) |
||
1039 | if multiply: |
||
1040 | window_data[0:range_end] *= adjustment_to_apply[1] |
||
1041 | else: |
||
1042 | window_data[0:range_end] /= adjustment_to_apply[1] |
||
1043 | |||
1044 | idx += 1 |
||
1045 | |||
1046 | def _get_adjustment_list(self, asset, adjustments_dict, table_name): |
||
1047 | """ |
||
1048 | Internal method that returns a list of adjustments for the given sid. |
||
1049 | |||
1050 | Parameters |
||
1051 | ---------- |
||
1052 | asset : Asset |
||
1053 | The asset for which to return adjustments. |
||
1054 | |||
1055 | adjustments_dict: dict |
||
1056 | A dictionary of sid -> list that is used as a cache. |
||
1057 | |||
1058 | table_name: string |
||
1059 | The table that contains this data in the adjustments db. |
||
1060 | |||
1061 | Returns |
||
1062 | ------- |
||
1063 | adjustments: list |
||
1064 | A list of [multiplier, pd.Timestamp], earliest first |
||
1065 | |||
1066 | """ |
||
1067 | if self._adjustment_reader is None: |
||
1068 | return [] |
||
1069 | |||
1070 | sid = int(asset) |
||
1071 | |||
1072 | try: |
||
1073 | adjustments = adjustments_dict[sid] |
||
1074 | except KeyError: |
||
1075 | adjustments = adjustments_dict[sid] = self._adjustment_reader.\ |
||
1076 | get_adjustments_for_sid(table_name, sid) |
||
1077 | |||
1078 | return adjustments |
||
1079 | |||
1080 | def _check_is_currently_alive(self, asset, dt): |
||
1081 | sid = int(asset) |
||
1082 | |||
1083 | if sid not in self._asset_start_dates: |
||
1084 | self._get_asset_start_date(asset) |
||
1085 | |||
1086 | start_date = self._asset_start_dates[sid] |
||
1087 | if self._asset_start_dates[sid] > dt: |
||
1088 | raise NoTradeDataAvailableTooEarly( |
||
1089 | sid=sid, |
||
1090 | dt=normalize_date(dt), |
||
1091 | start_dt=start_date |
||
1092 | ) |
||
1093 | |||
1094 | end_date = self._asset_end_dates[sid] |
||
1095 | if self._asset_end_dates[sid] < dt: |
||
1096 | raise NoTradeDataAvailableTooLate( |
||
1097 | sid=sid, |
||
1098 | dt=normalize_date(dt), |
||
1099 | end_dt=end_date |
||
1100 | ) |
||
1101 | |||
1102 | def _get_asset_start_date(self, asset): |
||
1103 | self._ensure_asset_dates(asset) |
||
1104 | return self._asset_start_dates[asset] |
||
1105 | |||
1106 | def _get_asset_end_date(self, asset): |
||
1107 | self._ensure_asset_dates(asset) |
||
1108 | return self._asset_end_dates[asset] |
||
1109 | |||
1110 | def _ensure_asset_dates(self, asset): |
||
1111 | sid = int(asset) |
||
1112 | |||
1113 | if sid not in self._asset_start_dates: |
||
1114 | self._asset_start_dates[sid] = asset.start_date |
||
1115 | self._asset_end_dates[sid] = asset.end_date |
||
1116 | |||
1117 | def get_splits(self, sids, dt): |
||
1118 | """ |
||
1119 | Returns any splits for the given sids and the given dt. |
||
1120 | |||
1121 | Parameters |
||
1122 | ---------- |
||
1123 | sids : list |
||
1124 | Sids for which we want splits. |
||
1125 | |||
1126 | dt: pd.Timestamp |
||
1127 | The date for which we are checking for splits. Note: this is |
||
1128 | expected to be midnight UTC. |
||
1129 | |||
1130 | Returns |
||
1131 | ------- |
||
1132 | list: List of splits, where each split is a (sid, ratio) tuple. |
||
1133 | """ |
||
1134 | if self._adjustment_reader is None or len(sids) == 0: |
||
1135 | return {} |
||
1136 | |||
1137 | # convert dt to # of seconds since epoch, because that's what we use |
||
1138 | # in the adjustments db |
||
1139 | seconds = int(dt.value / 1e9) |
||
1140 | |||
1141 | splits = self._adjustment_reader.conn.execute( |
||
1142 | "SELECT sid, ratio FROM SPLITS WHERE effective_date = ?", |
||
1143 | (seconds,)).fetchall() |
||
1144 | |||
1145 | sids_set = set(sids) |
||
1146 | splits = [split for split in splits if split[0] in sids_set] |
||
1147 | |||
1148 | return splits |
||
1149 | |||
1150 | def get_stock_dividends(self, sid, trading_days): |
||
1151 | """ |
||
1152 | Returns all the stock dividends for a specific sid that occur |
||
1153 | in the given trading range. |
||
1154 | |||
1155 | Parameters |
||
1156 | ---------- |
||
1157 | sid: int |
||
1158 | The asset whose stock dividends should be returned. |
||
1159 | |||
1160 | trading_days: pd.DatetimeIndex |
||
1161 | The trading range. |
||
1162 | |||
1163 | Returns |
||
1164 | ------- |
||
1165 | list: A list of objects with all relevant attributes populated. |
||
1166 | All timestamp fields are converted to pd.Timestamps. |
||
1167 | """ |
||
1168 | |||
1169 | if self._adjustment_reader is None: |
||
1170 | return [] |
||
1171 | |||
1172 | if len(trading_days) == 0: |
||
1173 | return [] |
||
1174 | |||
1175 | start_dt = trading_days[0].value / 1e9 |
||
1176 | end_dt = trading_days[-1].value / 1e9 |
||
1177 | |||
1178 | dividends = self._adjustment_reader.conn.execute( |
||
1179 | "SELECT * FROM stock_dividend_payouts WHERE sid = ? AND " |
||
1180 | "ex_date > ? AND pay_date < ?", (int(sid), start_dt, end_dt,)).\ |
||
1181 | fetchall() |
||
1182 | |||
1183 | dividend_info = [] |
||
1184 | for dividend_tuple in dividends: |
||
1185 | dividend_info.append({ |
||
1186 | "declared_date": dividend_tuple[1], |
||
1187 | "ex_date": pd.Timestamp(dividend_tuple[2], unit="s"), |
||
1188 | "pay_date": pd.Timestamp(dividend_tuple[3], unit="s"), |
||
1189 | "payment_sid": dividend_tuple[4], |
||
1190 | "ratio": dividend_tuple[5], |
||
1191 | "record_date": pd.Timestamp(dividend_tuple[6], unit="s"), |
||
1192 | "sid": dividend_tuple[7] |
||
1193 | }) |
||
1194 | |||
1195 | return dividend_info |
||
1196 | |||
1197 | def contains(self, asset, field): |
||
1198 | return field in BASE_FIELDS or \ |
||
1199 | (field in self._augmented_sources_map and |
||
1200 | asset in self._augmented_sources_map[field]) |
||
1201 | |||
1202 | def get_fetcher_assets(self, day): |
||
1203 | """ |
||
1204 | Returns a list of assets for the current date, as defined by the |
||
1205 | fetcher data. |
||
1206 | |||
1207 | Notes |
||
1208 | ----- |
||
1209 | Data is forward-filled. If there is no fetcher data defined for day |
||
1210 | N, we use day N-1's data (if available, otherwise we keep going back). |
||
1211 | |||
1212 | Returns |
||
1213 | ------- |
||
1214 | list: a list of Asset objects. |
||
1215 | """ |
||
1216 | # return a list of assets for the current date, as defined by the |
||
1217 | # fetcher source |
||
1218 | if self._extra_source_df is None: |
||
1219 | return [] |
||
1220 | |||
1221 | if day in self._extra_source_df.index: |
||
1222 | date_to_use = day |
||
1223 | else: |
||
1224 | # current day isn't in the fetcher df, go back the last |
||
1225 | # available day |
||
1226 | idx = self._extra_source_df.index.searchsorted(day) |
||
1227 | if idx == 0: |
||
1228 | return [] |
||
1229 | |||
1230 | date_to_use = self._extra_source_df.index[idx - 1] |
||
1231 | |||
1232 | asset_list = self._extra_source_df.loc[date_to_use]["sid"] |
||
1233 | |||
1234 | # make sure they're actually assets |
||
1235 | asset_list = [asset for asset in asset_list |
||
1236 | if isinstance(asset, Asset)] |
||
1237 | |||
1238 | return asset_list |
||
1239 |