Added support for the Bloomberg equity screening BEQS() function, wit…

[creyl]

…h the ability to request a screen as of today, as of a date in the past and for an arbitrary set of dates. See section 15.7 of https://data.bloomberglp.com/professional/sites/4/BLPAPI-Core-Developer-Guide.pdf.

Tried to follow your style as closely as possible.

Let me know!
Charles

[matthewgilbert]

Great, this looks like useful functionality, thanks Charles. Just a couple of
quick comments. Could you run flake8 on your branch to check PEP8
compatibility, e.g. flake8 ./pdblp

In addition, I don't know if the date column makes sense in the context of beqs()? If I run

df = con.beqs('Insider Buyers', 'GLOBAL', 'Popular')

df.head()

   ticker        field           value  date
0  JPM US       Ticker          JPM US     4
1  JPM US   Short Name  JPMORGAN CHASE     4
2  JPM US   Market Cap     3.09662e+11     4
3  JPM US  GICS Sector      Financials     4
4  JPM US        Cntry   United States     4

the date columns is just a randomly generated CorrelationID. I understand the use in the context of beqs_hist() but maybe it should be removed from beqs(), thoughts?

[matthewgilbert]

Test is failing for me since I am getting a DataFrame of length 488. Bloombergs exact data returned is not always reliable, a test which just checks the known properties of the returned DataFrame would be better (e.g. a DataFrame with specific columns)

[BethTian]

Hello Matthew,

I have read your codes and also the codes from blpapi. I found it is impossible to get the data of backdates. Actually I want to realize the lookup function that use the market cap rank in 2019 to select stock. But I cannot find any api methods to do that. I have found yours codes is working, but it is now not accurate?

Beth

[matthewgilbert]

freq parameter here is unused and should be removed. longdata parameter should be documented.

[matthewgilbert]

This commented out section should be removed

[creyl]

Thanks for the thoughtful feedback. I am sidetracked at the moment, but should eventually get to it. Best, C

…

On Fri, May 12, 2017 at 11:36 AM, Matthew Gilbert ***@***.***> wrote: ***@***.**** commented on this pull request. Great, this looks like useful functionality, thanks Charles. Just a couple of quick comments. Could you run flake8 on your branch to check PEP8 compatibility, e.g. flake8 ./pdblp In addition, I don't know if the date column makes sense in the context of beqs()? If I run df = con.beqs('Insider Buyers', 'GLOBAL', 'Popular') df.head() ticker field value date 0 JPM US Ticker JPM US 4 1 JPM US Short Name JPMORGAN CHASE 4 2 JPM US Market Cap 3.09662e+11 4 3 JPM US GICS Sector Financials 4 4 JPM US Cntry United States 4 the date columns is just a randomly generated CorrelationID. I understand the use in the context of beqs_hist() but maybe it should be removed from beqs(), thoughts? ------------------------------ In pdblp/tests/test_pdblp.py <#9 (comment)>: > @@ -164,3 +164,19 @@ def test_hist_ref_one_ticker_one_field_pivoted_non_numeric(self): ) df_expect.index.names = ["date"] assert_frame_equal(df, df_expect) + + def test_beqs_current(self): + df = self.con.beqs('Insider Buyers', 'GLOBAL', 'Popular') + assert len(df) > 0 + + def test_beqs_asof(self): + asof_date = datetime.datetime(2016, 5, 4) + df = self.con.beqs('Insider Buyers', 'GLOBAL', 'Popular', asof_date=asof_date) + assert len(df) == 1145 + + def test_beqs_hist(self): Test is failing for me since I am getting a DataFrame of length 488. Bloombergs exact data returned is not always reliable, a test which just checks the known properties of the returned DataFrame would be better (e.g. a DataFrame with specific columns) ------------------------------ In pdblp/pdblp.py <#9 (comment)>: > + for elm in elms: + data.append([ticker, fld, elm.getValue(), corrID]) + else: + val = reqFldsData.getElement(fld).getValue() + data.append([ticker, fld, val, corrID]) + return data + + def _beqs_create_req(self, screen_name, screen_type, group, language_id): + request = self.refDataService.createRequest('BeqsRequest') + request.set('screenName', screen_name) + request.set('screenType', screen_type) + request.set('Group', group) + request.set('languageId', language_id) + return request + + def beqs_hist(self, screen_name, screen_type='PRIVATE', group='General', language_id='ENGLISH', freq parameter here is unused and should be removed. longdata parameter should be documented. ------------------------------ In pdblp/pdblp.py <#9 (comment)>: > + ovrd.setElement("value", dt.strftime('%Y%m%d')) + # CorrelationID used to keep track of which response coincides with which request + cid = blpapi.CorrelationId(dt) + logging.debug("Sending Request:\n %s" % request) + self.session.sendRequest(request, correlationId=cid) + data = [] + # Process received events + ev_counter = 0 + while (True): + ev = self.session.nextEvent(timeout) + data = self._beqs_parse_event(data, ev) + if ev.eventType() == blpapi.Event.RESPONSE: + ev_counter += 1 + if ev_counter == len(asof_dates): + break + # if ev.eventType() == blpapi.Event.TIMEOUT: This commented out section should be removed — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#9 (review)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AHlEN6qkvflefSh3YgcV6ggOk7BnByDdks5r5Hx7gaJpZM4NYptq> .