4
\$\begingroup\$

I have a Class that:

  1. goes to a url
  2. grabs a link from that page, and a date (filing_date)
  3. navigates to the link, and
  4. writes the table from that page to a dataframe.

I am trying to add the respective filing_date from step 2 to the dataframe from step 4, but rather than pass the multiple filing_dates, like so:

                     nameOfIssuer                cik Filing Date
0    Agilent Technologies, Inc. (A)  ...  0000846222  2020-01-10
1                 Adient PLC (ADNT)  ...  0000846222  2020-01-10
..                             ...   ...         ...         ...
662            Whirlpool Corp (WHR)  ...  0000846222  2010-07-08

it only passes the last scraped date from the prior page to all rows:

                     nameOfIssuer                cik Filing Date
0    Agilent Technologies, Inc. (A)  ...  0000846222  2010-07-08
1                 Adient PLC (ADNT)  ...  0000846222  2010-07-08
..                             ...   ...         ...         ...
662            Whirlpool Corp (WHR)  ...  0000846222  2010-07-08

I've tried storing the dates to an empty list and then appending to the output data frame, but because the length of the list doesn't match the list of the dataframe, I get ValueError: Length of values does not match length of index.

Can someone advise on what the best approach would be (e.g., making another function to solely handle filing_date or perhaps returning a data frame instead)?

import pandas as pd
from urllib.parse import urljoin
from bs4 import BeautifulSoup, SoupStrainer
import requests

class Scraper:
    BASE_URL = "https://www.sec.gov"
    FORMS_URL_TEMPLATE = "/cgi-bin/browse-edgar?action=getcompany&CIK={cik}&type=13F"

    def __init__(self):
        self.session = requests.Session()

    def get_holdings(self, cik):
        """
        Main function that first finds the most recent 13F form and then passes
        it to scrapeForm to get the holdings for a particular institutional investor.
        """
        # get the form urls
        forms_url = urljoin(self.BASE_URL, self.FORMS_URL_TEMPLATE.format(cik=cik))
        parse_only = SoupStrainer('a', {"id": "documentsbutton"})
        soup = BeautifulSoup(self.session.get(forms_url).content, 'lxml', parse_only=parse_only)
        urls = soup.find_all('a', href=True)

        # get form document URLs
        form_urls = []
        for url in urls:
            url = url.get("href")
            url = urljoin(self.BASE_URL, str(url))

            headers = {'User-Agent': 'Mozilla/5.0'}
            page = requests.get(url, headers=headers)
            soup = BeautifulSoup(page.content, 'html.parser')

            # Get filing date and "period date"
            dates = soup.find("div", {"class": "formContent"})
            filing_date = dates.find_all("div", {"class": "formGrouping"})[0]
            filing_date = filing_date.find_all("div", {"class": "info"})[0]
            filing_date = filing_date.text

            # get form table URLs
            parse_only = SoupStrainer('tr', {"class": 'blueRow'})
            soup = BeautifulSoup(self.session.get(url).content,'lxml', parse_only=parse_only)
            form_url = soup.find_all('tr', {"class": 'blueRow'})[-1].find('a')['href']
            if ".txt" in form_url:
                pass
            else:
                form_url = urljoin(self.BASE_URL, form_url)
                # print(form_url)
                form_urls.append(form_url)

        return self.scrape_document(form_urls, cik, filing_date)

    def scrape_document(self, urls, cik, filing_date):
        """This function scrapes holdings from particular document URL"""

        cols = ['nameOfIssuer', 'titleOfClass', 'cusip', 'value', 'sshPrnamt',
                'sshPrnamtType', 'putCall', 'investmentDiscretion',
                'otherManager', 'Sole', 'Shared', 'None']

        data = []

        for url in urls:
            soup = BeautifulSoup(self.session.get(url).content, 'lxml')

            for info_table in soup.find_all(['ns1:infotable', 'infotable']):
                row = []
                for col in cols:
                    d = info_table.find([col.lower(), 'ns1:' + col.lower()])
                    row.append(d.text.strip() if d else 'NaN')
                data.append(row)

            df = pd.DataFrame(data, columns=cols)
            df['cik'] = cik
            df['Filing Date'] = filing_date

        return df

holdings = Scraper()
holdings = holdings.get_holdings("0000846222")
print(holdings)

| improve this question | | | | |
\$\endgroup\$
  • 2
    \$\begingroup\$ What version of python is this written in? \$\endgroup\$ – Linny Mar 11 at 2:16
  • 1
    \$\begingroup\$ Sorry, it's Python 3.7 \$\endgroup\$ – MSD Mar 11 at 2:44
5
\$\begingroup\$

Seems like you have as many filing_dates as you have URLs, so you should have those together and handle them similarly.

Your problem seems to come from the fact that you're losing the intel of which row comes from which URL, and so your only option becomes to set one date for the full dataframe.

Here's an updated version saving the dates at the same time as the URLs and using and using a new res_df dataframe in scrape_document to aggregate the dataframes retrieved from each URL.

import pandas as pd
from urllib.parse import urljoin
from bs4 import BeautifulSoup, SoupStrainer
import requests

class Scraper:
    BASE_URL = "https://www.sec.gov"
    FORMS_URL_TEMPLATE = "/cgi-bin/browse-edgar?action=getcompany&CIK={cik}&type=13F"

    def __init__(self):
        self.session = requests.Session()

    def get_holdings(self, cik):
        """
        Main function that first finds the most recent 13F form and then passes
        it to scrapeForm to get the holdings for a particular institutional investor.
        """
        # get the form urls
        forms_url = urljoin(self.BASE_URL, self.FORMS_URL_TEMPLATE.format(cik=cik))
        parse_only = SoupStrainer('a', {"id": "documentsbutton"})
        soup = BeautifulSoup(self.session.get(forms_url).content, 'lxml', parse_only=parse_only)
        urls = soup.find_all('a', href=True)

        # get form document URLs
        form_urls = []
        filing_dates = []
        for url in urls:
            url = url.get("href")
            url = urljoin(self.BASE_URL, str(url))

            headers = {'User-Agent': 'Mozilla/5.0'}
            page = requests.get(url, headers=headers)
            soup = BeautifulSoup(page.content, 'html.parser')

            # Get filing date and "period date"
            dates = soup.find("div", {"class": "formContent"})
            filing_date = dates.find_all("div", {"class": "formGrouping"})[0]
            filing_date = filing_date.find_all("div", {"class": "info"})[0]
            filing_date = filing_date.text

            # get form table URLs
            parse_only = SoupStrainer('tr', {"class": 'blueRow'})
            soup = BeautifulSoup(self.session.get(url).content,'lxml', parse_only=parse_only)
            form_url = soup.find_all('tr', {"class": 'blueRow'})[-1].find('a')['href']
            if ".txt" in form_url:
                pass
            else:
                form_url = urljoin(self.BASE_URL, form_url)
                # print(form_url)
                form_urls.append(form_url)
                # Save the filing date too
                filing_dates.append(filing_date)

        # Pass the dates list rather than the last one
        return self.scrape_document(form_urls, cik, filing_dates)

    def scrape_document(self, urls, cik, filing_dates):
        """This function scrapes holdings from particular document URL"""

        cols = ['nameOfIssuer', 'titleOfClass', 'cusip', 'value', 'sshPrnamt',
                'sshPrnamtType', 'putCall', 'investmentDiscretion',
                'otherManager', 'Sole', 'Shared', 'None']

        res_df = pd.DataFrame(columns=cols+["Filing Date"])

        # Iterate over both list at the same time
        for url, date in zip(urls, filing_dates):
            data = []
            soup = BeautifulSoup(self.session.get(url).content, 'lxml')

            for info_table in soup.find_all(['ns1:infotable', 'infotable']):
                row = []
                for col in cols:
                    d = info_table.find([col.lower(), 'ns1:' + col.lower()])
                    row.append(d.text.strip() if d else 'NaN')
                data.append(row)
            url_df = pd.DataFrame(data, columns=cols)
            url_df["Filing Date"] = date
            res_df = res_df.append(url_df, ignore_index=True)

        # CIK seems common to the whole DF, if not follow the example of dates
        res_df['cik'] = cik

        return res_df

holdings = Scraper()
holdings = holdings.get_holdings("0000846222")
print(holdings)
```
| improve this answer | | | | |
\$\endgroup\$
  • 1
    \$\begingroup\$ Thanks I appreciate this. Made a couple edits on some minor typos but I think this will work. Out of curiosity, is there a better way I could have organized/structured this from the beginning? \$\endgroup\$ – MSD Mar 11 at 18:06
  • 1
    \$\begingroup\$ It's pretty clean already. Separating the data processing from the data fetching (get requests) might prove useful if you plan to write unit tests. It'll make mocking easier. \$\endgroup\$ – Cal Mar 12 at 17:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.