def insert_extract_info(vol,
image_url, image_name, image_author, image_cita,
article_title, article_author, article_content,
qa_q_title, qa_q_content, qa_a_title, qa_a_content):
res = True
logger = get_logger()
sql = '''
insert into TJournalExtractInfo(
vol,
image_url, image_name, image_author, image_cita,
article_title, article_author, article_content,
qa_q_title, qa_q_content, qa_a_title, qa_a_content
)
values(?,?,?,?,?,?,?,?,?,?,?,?)'''
conn = get_db_connect()
try:
logstr = 'sql: {} \t vol:{}'.format(sql, vol)
logstr = ' '.join(logstr.split())
logger.info(logstr)
params = (vol,
image_url, image_name, image_author, image_cita,
article_title, article_author, article_content,
qa_q_title, qa_q_content, qa_a_title, qa_a_content)
conn.execute(sql, params)
conn.commit()
except Exception as ex:
logger.error(ex)
res = False
finally:
conn.close()
return res
def get_one_journal(num):
logger = get_logger()
url = BASE_URL + 'vol.' + str(num)
logger.info('url:%s', url)
request = urllib.request.Request(url)
request = fill_request(request)
html = get_html(request)
return html
def get_bytes(req):
logger = get_logger()
res_bytes = None
try:
response = urllib.request.urlopen(req, timeout=15)
res_bytes = response.read()
except Exception as ex:
logger.error(ex)
res_bytes = None
return res_bytes
def download_image(url, path):
logger = get_logger()
logger.info('get image url:%s', url)
request = urllib.request.Request(url)
bytes = get_bytes(request)
if bytes is None:
logger.error('get image error')
return False
f = open(path, mode='bw')
f.write(bytes)
f.close()
return True
def insert_journal_src(vol, html):
res = True
logger = get_logger()
sql = 'insert into TJournalSrc(vol, html) values(?,?)'
conn = get_db_connect()
try:
logger.info('sql: %s\t vol:%s', sql, vol)
conn.execute(sql, (vol, html))
conn.commit()
except Exception as ex:
logger.error(ex)
res = False
finally:
conn.close()
return res
logger = get_logger()
url = BASE_URL + 'vol.' + str(num)
logger.info('url:%s', url)
request = urllib.request.Request(url)
request = fill_request(request)
html = get_html(request)
return html
def fill_request(request):
for k in HEADERS:
request.add_header(k, HEADERS[k])
return request
if __name__ == '__main__':
logger = get_logger()
start_day = datetime.date(2012, 10, 8)
today = datetime.date.today()
vol = today - start_day + datetime.timedelta(days=2)
vol = vol.days
logger.info('range from %d to %d', 1, vol)
for i in range(1, vol):
logger.info('begin get vol %d', i)
html = get_one_journal(i)
if html is None:
logger.error('get html error')
else:
insert_journal_src(i, html)
time.sleep(5)
