Python 如何将来自 Bloomberg API 的数据存储到 Pandas 数据框中？

Question

提问by Danny

I recently started using Python so I could interact with the Bloomberg API, and I'm having some trouble storing the data into a Pandas dataframe (or a panel). I can get the output in the command prompt just fine, so that's not an issue.

我最近开始使用 Python，因此我可以与 Bloomberg API 进行交互，但在将数据存储到 Pandas 数据框（或面板）中时遇到了一些问题。我可以在命令提示符中获得输出就好了，所以这不是问题。

A very similar question was asked here: Pandas wrapper for Bloomberg api?

这里提出了一个非常相似的问题： Pandas wrapper for Bloomberg api?

The referenced code in the accepted answer for that question is for the old API, however, and it doesn't work for the new open API. Apparently the user who asked the question was able to easily modify that code to work with the new API, but I'm used to having my hand held in R, and this is my first endeavor with Python.

但是，该问题的已接受答案中引用的代码适用于旧 API，并且不适用于新的开放 API。显然，提出问题的用户能够轻松修改该代码以使用新的 API，但我已经习惯了将手放在 R 中，这是我第一次尝试使用 Python。

Could some benevolent user show me how to get this data into Pandas? There is an example in the Python API (available here: http://www.openbloomberg.com/open-api/) called SimpleHistoryExample.py that I've been working with that I've included below. I believe I'll need to modify mostly around the 'while(True)' loop toward the end of the 'main()' function, but everything I've tried so far has had issues.

一些仁慈的用户可以告诉我如何将这些数据导入 Pandas 吗？Python API 中有一个示例（可在此处获得：http: //www.openbloomberg.com/open-api/），名为 SimpleHistoryExample.py，我一直在使用它，我已将其包含在下面。我相信我需要在 'main()' 函数结束时主要围绕 'while(True)' 循环进行修改，但到目前为止我尝试过的所有内容都有问题。

Thanks in advance, and I hope this can be of help to anyone using Pandas for finance.

提前致谢，我希望这对使用 Pandas 进行融资的任何人都有帮助。

# SimpleHistoryExample.py

import blpapi
from optparse import OptionParser


def parseCmdLine():
    parser = OptionParser(description="Retrieve reference data.")
    parser.add_option("-a",
                      "--ip",
                      dest="host",
                      help="server name or IP (default: %default)",
                      metavar="ipAddress",
                      default="localhost")
    parser.add_option("-p",
                      dest="port",
                      type="int",
                      help="server port (default: %default)",
                      metavar="tcpPort",
                      default=8194)

    (options, args) = parser.parse_args()

    return options


def main():
    options = parseCmdLine()

    # Fill SessionOptions
    sessionOptions = blpapi.SessionOptions()
    sessionOptions.setServerHost(options.host)
    sessionOptions.setServerPort(options.port)

    print "Connecting to %s:%s" % (options.host, options.port)
    # Create a Session
    session = blpapi.Session(sessionOptions)

    # Start a Session
    if not session.start():
        print "Failed to start session."
        return

    try:
        # Open service to get historical data from
        if not session.openService("//blp/refdata"):
            print "Failed to open //blp/refdata"
            return

        # Obtain previously opened service
        refDataService = session.getService("//blp/refdata")

        # Create and fill the request for the historical data
        request = refDataService.createRequest("HistoricalDataRequest")
        request.getElement("securities").appendValue("IBM US Equity")
        request.getElement("securities").appendValue("MSFT US Equity")
        request.getElement("fields").appendValue("PX_LAST")
        request.getElement("fields").appendValue("OPEN")
        request.set("periodicityAdjustment", "ACTUAL")
        request.set("periodicitySelection", "DAILY")
        request.set("startDate", "20061227")
        request.set("endDate", "20061231")
        request.set("maxDataPoints", 100)

        print "Sending Request:", request
        # Send the request
        session.sendRequest(request)

        # Process received events
        while(True):
            # We provide timeout to give the chance for Ctrl+C handling:
            ev = session.nextEvent(500)
            for msg in ev:
                print msg

            if ev.eventType() == blpapi.Event.RESPONSE:
                # Response completly received, so we could exit
                break
    finally:
        # Stop the session
        session.stop()

if __name__ == "__main__":
    print "SimpleHistoryExample"
    try:
        main()
    except KeyboardInterrupt:
        print "Ctrl+C pressed. Stopping..."

Answer 1

回答by Luciano

I've been using pybbg to do this sort of stuff. You can get it here:

我一直在使用 pybbg 来做这种事情。你可以在这里得到它：

https://github.com/bpsmith/pybbg

Import the package and you can then do (this is in the source code, bbg.py file):

导入包，然后你就可以做（这是在源代码中，bbg.py 文件）：

banner('ReferenceDataRequest: single security, single field, frame response')
req = ReferenceDataRequest('msft us equity', 'px_last', response_type='frame')
print req.execute().response

The advantages:

优点：

Easy to use; minimal boilerplate, and parses indices and dates for you.
It's blocking. Since you mention R, I assume you are using this in some type of an interactive environment, like IPython. So this is what you want , rather than having to mess around with callbacks.
It can also do historical (i.e. price series), intraday and bulk data request (no tick data yet).

便于使用; 最小的样板，并为您解析索引和日期。
它在阻塞。既然你提到了 R，我假设你在某种类型的交互式环境中使用它，比如 IPython。所以这就是你想要的，而不是必须处理回调。
它还可以做历史（即价格系列）、日内和批量数据请求（还没有报价数据）。

Disadvantages:

缺点：

Only works in Windows, as far as I know (you must have BB workstationg installed and running).
Following on the above, it depends on the 32 bit OLE api for Python. It only works with the 32 bit version - so you will need 32 bit python and 32 bit OLE bindings
There are some bugs. In my experience, when retrieving data for a number of instruments, it tends to hang IPython. Not sure what causes this.

据我所知，仅适用于 Windows（您必须安装并运行 BB 工作站）。
继上述之后，它取决于 Python 的 32 位 OLE api。它只适用于 32 位版本 - 所以你需要 32 位 python 和 32 位 OLE 绑定
有一些错误。根据我的经验，在为许多仪器检索数据时，它往往会挂起 IPython。不知道是什么原因造成的。

Based on the last point, I would suggest that if you are getting large amounts of data, you retrieve and store these in an excel sheet (one instrument per sheet), and then import these. read_excelisn't efficient for doing this; you need to use the ExcelReader (?) object, and then iterate over the sheets. Otherwise, using read_excel will reopen the file each time you read a sheet; this can take ages.

基于最后一点，我建议如果您要获取大量数据，则将这些数据检索并存储在 Excel 表格中（每张表格一个仪器），然后导入这些数据。read_excel这样做效率不高；您需要使用 ExcelReader (?) 对象，然后遍历工作表。否则，每次阅读工作表时，使用 read_excel 都会重新打开文件；这可能需要很长时间。

Answer 2

回答by Thomas Browne

A proper Bloomberg API for python now exists which does not use COM. It has all of the hooks to allow you to replicate the functionality of the Excel addin, with the obvious advantage of a proper programming language endpoint. The request and response objects are fairly poorly documented, and are quite obtuse. Still, the examples in the API are good, and some playing around using the inspect module and printing of response messages should get you up to speed. Sadly, the standard terminal licence only works on Windows. For *nix you will need a server licence (even more expensive). I have used it quite extensively.

现在存在一个不使用 COM 的适用于 python 的 Bloomberg API。它具有所有钩子，允许您复制 Excel 插件的功能，并具有适当的编程语言端点的明显优势。请求和响应对象的文档相当差，而且相当迟钝。尽管如此，API 中的示例还是不错的，使用检查模块和打印响应消息的一些尝试应该可以让您快速上手。遗憾的是，标准终端许可证仅适用于 Windows。对于 *nix，您需要一个服务器许可证（甚至更贵）。我已经非常广泛地使用它。

https://www.bloomberg.com/professional/support/api-library/

Answer 3

回答by alex314159

I've just published this which might help

我刚刚发布了这可能会有所帮助

http://github.com/alex314159/blpapiwrapper

It's basically not very intuitive to unpack the message, but this is what works for me, where strData is a list of bloomberg fields, for instance ['PX_LAST','PX_OPEN']:

解包消息基本上不是很直观，但这对我有用，其中 strData 是bloomberg字段的列表，例如['PX_LAST','PX_OPEN']：

fieldDataArray = msg.getElement('securityData').getElement('fieldData')
size = fieldDataArray.numValues()
fieldDataList = [fieldDataArray.getValueAsElement(i) for i in range(0,size)]
outDates = [x.getElementAsDatetime('date') for x in fieldDataList]
output = pandas.DataFrame(index=outDates,columns=strData)
for strD in strData:
    outData = [x.getElementAsFloat(strD) for x in fieldDataList]
    output[strD] = outData
output.replace('#N/A History',pandas.np.nan,inplace=True)
output.index = output.index.to_datetime()
return output

Answer 4

回答by Gabriel

I use tia (https://github.com/bpsmith/tia/blob/master/examples/datamgr.ipynb)

我使用 tia ( https://github.com/bpsmith/tia/blob/master/examples/datamgr.ipynb)

It already downloads data as a panda dataframe from bloomberg. You can download history for multiple tickers in one single call and even download some bloombergs reference data (Central Bank date meetings, holidays for a certain country, etc)

它已经从彭博社下载数据作为熊猫数据框。您可以在一次调用中下载多个股票的历史记录，甚至可以下载一些彭博参考数据（中央银行日期会议、某个国家的假期等）

And you just install it with pip. This link is full of examples but to download historical data is as easy as:

你只需用 pip 安装它。此链接充满了示例，但下载历史数据非常简单：

import pandas as pd
import tia.bbg.datamgr as dm

mgr = dm.BbgDataManager()
sids = mgr['MSFT US EQUITY', 'IBM US EQUITY', 'CSCO US EQUITY']
df = sids.get_historical('PX_LAST', '1/1/2014', '11/12/2014')

and df is a pandas dataframe.

和 df 是一个熊猫数据框。

Hope it helps

希望能帮助到你

Answer 5

回答by citynorman

Tia https://github.com/bpsmith/tiais the best I've found, and I've tried them all... It allows you to do:

Tia https://github.com/bpsmith/tia是我发现的最好的，我已经尝试了所有......它允许你做：

import pandas as pd
import datetime
import tia.bbg.datamgr as dm
mgr = dm.BbgDataManager()
sids = mgr['BAC US EQUITY', 'JPM US EQUITY']
df = sids.get_historical(['BEST_PX_BPS_RATIO','BEST_ROE'],
                         datetime.date(2013,1,1),
                         datetime.date(2013,2,1),
                         BEST_FPERIOD_OVERRIDE="1GY",
                         non_trading_day_fill_option="ALL_CALENDAR_DAYS",
                         non_trading_day_fill_method="PREVIOUS_VALUE")
print df

#and you'll probably want to carry on with something like this
df1=df.unstack(level=0).reset_index()
df1.columns = ('ticker','field','date','value')
df1.pivot_table(index=['date','ticker'],values='value',columns='field')
df1.pivot_table(index=['date','field'],values='value',columns='ticker')

The caching is nice too.

缓存也不错。

Both https://github.com/alex314159/blpapiwrapperand https://github.com/kyuni22/pybbgdo the basic job (thanks guys!) but have trouble with multiple securities/fields as well as overrides which you will inevitably need.

无论https://github.com/alex314159/blpapiwrapper和https://github.com/kyuni22/pybbg做基本的工作（谢谢你们！），但有多个证券/领域以及覆盖麻烦你必然需要.

The one thing this https://github.com/kyuni22/pybbghas that tia doesn't have is bds(security, field).

这个https://github.com/kyuni22/pybbg有而 tia 没有的一件事是 bds(security, field)。

Answer 6

回答by mgilbert

You can also use pdblpfor this (Disclaimer: I'm the author). There is a tutorial showing similar functionality available here https://matthewgilbert.github.io/pdblp/tutorial.html, the functionality could be achieved using something like

您也可以为此使用pdblp（免责声明：我是作者）。有一个教程显示了这里可用的类似功能https://matthewgilbert.github.io/pdblp/tutorial.html，该功能可以使用类似的东西来实现

import pdblp
con = pdblp.BCon()
con.start()
con.bdh(['IBM US Equity', 'MSFT US Equity'], ['PX_LAST', 'OPEN'],
        '20061227', '20061231', elms=[("periodicityAdjustment", "ACTUAL")])

Python 如何将来自 Bloomberg API 的数据存储到 Pandas 数据框中？

提问by Danny

回答by Luciano

回答by Thomas Browne

回答by alex314159

回答by Gabriel

回答by citynorman

回答by mgilbert

相关推荐

最近更新

标签

Python 如何将来自 Bloomberg API 的数据存储到 Pandas 数据框中？

提问by Danny

回答by Luciano

回答by Thomas Browne

回答by alex314159

回答by Gabriel

回答by citynorman

回答by mgilbert

相关推荐

改进 Python 回文代码

Python TypeError: __init__() 在 argparse 中有一个意外的关键字参数“type”

Python 限制一次运行的最大线程数的正确方法？

Python 使用 conda 安装 gdal？

相关推荐

最近更新

标签

Python TypeError: init() 在 argparse 中有一个意外的关键字参数“type”