Final deliverable

t_visa_trends_team1/.gitignore

@@ -0,0 +1 @@
+.ipynb_checkpoints

t_visa_trends_team1/Final delliverable/AAO_data_collection.py

@@ -0,0 +1,224 @@
+import re
+import requests
+import os
+from bs4 import BeautifulSoup
+import pandas as pd
+import subprocess
+from tqdm import tqdm
+
+import PyPDF2
+from pdfminer.pdfparser import PDFParser, PDFDocument
+from pdfminer.pdfinterp import PDFResourceManager, PDFPageInterpreter
+from pdfminer.converter import PDFPageAggregator
+from pdfminer.layout import LAParams, LTTextBox
+from pdfminer.pdfinterp import PDFTextExtractionNotAllowed
+
+# Global variable
+INDEX_FILE = './pdf_name_url.csv'
+PDF_DIR = './download_pdf'
+if not os.path.exists(PDF_DIR):
+    os.mkdir(PDF_DIR)
+RESULT_FILE = './result.csv'
+
+# spider
+def spider_url_and_download_pdf(update_index=True):
+    """loop through the target site
+
+    update_index: True: scrape again; False: no need update
+    """
+    PAGE_NUM = 1  # page number
+    NULL_PAGE_DISP = 'Sorry, no results found'
+    TITLE_MUST_DISP = 'Application for T Nonimmigrant Status'
+
+    if os.path.exists(INDEX_FILE):
+        DATA_HIST = [(na, url) for na, url in pd.read_csv(INDEX_FILE).values]
+    else:
+        DATA_HIST = []
+
+    # update_index == True, the crawl will loop through the site again, and check if there is any updates
+    if update_index:
+        DATA_NEW = []
+        while True:
+            url = f'https://search.usa.gov/search?affiliate=uscis-aao&dc=1847&page={PAGE_NUM}&query=form+I-914&search=Search&utf8=%E2%9C%93'
+            try:
+                res = requests.get(url)
+
+                # no more data
+                if NULL_PAGE_DISP in res.text:
+                    print('No more data, spider done.')
+                    break
+
+                soup = BeautifulSoup(res.text, 'html.parser')
+                divs = soup.find_all('div', attrs={'class': 'content-block-item result'})
+
+                title_urls = [re.split('\n+', d.text.strip())[1:3] for d in divs]
+                valid_title_urls = [(title, url) for title, url in title_urls
+                                    if TITLE_MUST_DISP in title]
+                new_data = [tup for tup in valid_title_urls if tup not in DATA_HIST]
+                print(f'Page {PAGE_NUM}, find {len(new_data)} new url.')
+
+                DATA_NEW.extend(new_data)
+                PAGE_NUM += 1
+
+            except Exception:
+                import traceback
+                traceback.print_exc()
+                import ipdb
+                ipdb.set_trace()
+
+        # update hist_file
+        DATA_HIST.extend(DATA_NEW)
+        pd.DataFrame(DATA_HIST, columns=['title', 'url']).drop_duplicates().to_csv(INDEX_FILE,
+                                                                                   index=None)
+
+    # download pdf file
+    df_hist = pd.read_csv(INDEX_FILE)
+    for title, url in tqdm(df_hist.values, desc='pdf download'):
+        try:
+            fname = os.path.basename(url)
+            fpath = os.path.join(PDF_DIR, fname)
+            if os.path.exists(fpath):
+                continue
+            subprocess.check_call(['wget', url], cwd=PDF_DIR, stderr=subprocess.DEVNULL)
+
+        except Exception:
+            print(f'{url} download failed')
+
+
+def parse_by_pdfminer(pdf_path):
+
+    praser = PDFParser(open(pdf_path, 'rb'))
+    doc = PDFDocument()
+    praser.set_document(doc)
+    doc.set_parser(praser)
+
+    doc.initialize()
+
+    if not doc.is_extractable:
+        raise PDFTextExtractionNotAllowed
+    else:
+        rsrcmgr = PDFResourceManager()
+        laparams = LAParams()
+        device = PDFPageAggregator(rsrcmgr, laparams=laparams)
+        interpreter = PDFPageInterpreter(rsrcmgr, device)
+
+        page_txts = []
+        for page in doc.get_pages():
+            interpreter.process_page(page)
+            layout = device.get_result()
+            layout_txts = []
+            for x in layout:
+                if isinstance(x, LTTextBox):
+                    layout_txts.append(x.get_text().strip())
+            page_txts.append('\n'.join(layout_txts))
+
+        return page_txts
+
+
+def parse_by_PyPDF2(pdf_path):
+
+    mypdf = open(pdf_path, mode='rb')
+
+    pdf_document = PyPDF2.PdfFileReader(mypdf)
+
+    page_num = pdf_document.numPages
+    page_txts = []
+    for page in range(page_num):
+        txt = pdf_document.getPage(page).extractText()
+        page_txts.append(txt)
+
+    return page_txts
+
+
+def extract_info(page_txts):
+
+    # 1. get id
+    # Get match in first page: "In Re: 9435010" or last page: "ID# 1940904"
+    match1 = re.search('In Re: (\d+)', page_txts[0], re.IGNORECASE)
+    match2 = re.search('# (\d+)', page_txts[-1], re.IGNORECASE)
+
+    if match1:
+        ID = match1.group(1)
+    elif match2:
+        ID = match2.group(1)
+    else:
+        ID = None
+
+    # 2. date decision status
+    # match_date = re.search('DA\s*TE.*?:(.*?\d{4})', page_txts[0], re.IGNORECASE)
+    # date = match_date.group(1).strip() if match_date else None
+
+    match_decision = re.search('(appeal|motion).*?decision', page_txts[0], re.IGNORECASE)
+    decision = match_decision.group() if match_decision else None
+
+    match_status = re.search('FORM [i1l]-914.*?status', page_txts[0], re.IGNORECASE)
+    status = match_status.group() if match_status else None
+    status = re.sub('[il1]-', 'I-', status, re.IGNORECASE)
+
+    match_order = re.search('ORDER:.*?(The.*?\.)', page_txts[-1], re.IGNORECASE)
+    order = match_order.group(1).strip() if match_order else None
+
+    match_desc = re.search('(The Applicant.*?)(\s+I\.|$)', page_txts[0], re.IGNORECASE | re.S)
+    desc = match_desc.group(1).strip() if match_desc else None
+    # if not desc:
+    #     raise ValueError
+
+    match_is_family = re.search('FAMILY MEMBER', page_txts[0], re.IGNORECASE)
+    is_family = True if match_is_family else False
+
+    return {
+        'ID': ID,
+        'decision': decision,
+        'status': status,
+        'order': order,
+        'is_family': is_family,
+        'desc': desc,
+    }
+
+
+def parse_pdf_info(tool='PyPDF2'):
+
+    tool2func = {'pdfminer': parse_by_pdfminer, 'PyPDF2': parse_by_PyPDF2}
+
+    BASE_INFO = {
+        'url': None,
+        'file_name': None,
+        'ID': None,
+        'date': None,
+        'decision': None,
+        'status': None,
+        'order': None,
+        'is_family': None,
+    }
+    ALL_DATA = []
+    urls = pd.read_csv(INDEX_FILE).url
+    for url in tqdm(urls, desc='Parse pdf file'):
+
+        data = BASE_INFO.copy()
+
+        fname = os.path.basename(url)
+        data['file_name'] = fname
+        data['url'] = url
+        data['date'] = fname.split('_')[0]
+
+        fp = os.path.join(PDF_DIR, fname)
+        try:
+            page_txts = tool2func[tool](fp)
+            extrat_dict = extract_info(page_txts)
+            data.update(extrat_dict)
+
+            ALL_DATA.append(data)
+
+        except:
+            import traceback
+            traceback.print_exc()
+            print(f'{fname}failed')
+            import ipdb
+            ipdb.set_trace()
+
+    pd.DataFrame(ALL_DATA).to_csv(RESULT_FILE, index=None)
+
+
+if __name__ == "__main__":
+    spider_url_and_download_pdf(update_index=True)
+    parse_pdf_info()