class TaurusParser(): def __init__(self): self.base_xpath = '//html/body/center' self.date_utils = DateUtils() self.string_utils = StringUtils() def parse(self, content): html_object = self.__get_html_object(content) relative_html_object = self.__traverse_to_relative_html_object(html_object) column_name_list = self.__parse_column_name_list(relative_html_object) row_list = self.__parse_row_list(relative_html_object) release_date = self.__parse_release_date(relative_html_object) return column_name_list, row_list, release_date def __get_html_object(self, content): content = self.string_utils.normalize_string(content) content = content.replace(u'<br>', u'') return lxml.html.fromstring(content) def __traverse_to_relative_html_object(self, html_object): relative_html_object_list = html_object.xpath(self.base_xpath) assert len(relative_html_object_list) > 0, 'invalid base_xpath' return relative_html_object_list[0] def __parse_column_name_list(self, relative_html_object): # traverse and sanity check table_tags = relative_html_object.xpath('./table') assert len(table_tags) > 1, 'invalid table_tags' # skip first table of description about IFRS inner_table_tags = table_tags[1].xpath('./tr/td/table/tr/td/table') assert len(inner_table_tags) > 0, 'invalid inner_table_tags' tr_tags = inner_table_tags[0].xpath('./tr') assert len(tr_tags) > 1, 'invalid tr_tags' th_texts = tr_tags[1].xpath('./th/text()') return th_texts def __parse_row_list(self, relative_html_object): # traverse and sanity check table_tags = relative_html_object.xpath('./table') assert len(table_tags) > 1, 'invalid table_tags' # skip first table of description about IFRS inner_table_tags = table_tags[1].xpath('./tr/td/table/tr/td/table') assert len(inner_table_tags) > 0, 'invalid inner_table_tags' all_tr_tags = [] # every inner_table represents an industry for inner_table_tag in inner_table_tags: tr_tags = inner_table_tag.xpath('./tr') assert len(tr_tags) > 2, 'invalid tr_tags' # first two rows are headers # last row is u'合計' all_tr_tags += tr_tags[2:-1] return [self.__parse_row(tr_tag) for tr_tag in all_tr_tags] def __parse_row(self, relative_html_object): td_texts = relative_html_object.xpath('./td/text()') # record contains extra entry about comment assert len(td_texts) == 11, 'invalid td_texts size, should be 11' items = td_texts[:2] numbers = [] # skip the last entry about comment for td_text in td_texts[2:-1]: number = self.string_utils.normalize_number(td_text) numbers.append(number) return items + numbers def __parse_release_date(self, relative_html_object): div_tags = relative_html_object.xpath('./div') assert len(div_tags) > 0, 'invalid div_tags' groups = self.string_utils.match(u'^出表日期:(.*)$', div_tags[-1].text.strip()) assert len(groups) > 0, 'could not match ^出表日期:(.*)$' release_date = self.string_utils.from_local_string_to_date(groups[0]) return release_date
class StringUtilsTest(unittest.TestCase): def setUp(self): self.string_utils = StringUtils() def tearDown(self): self.string_utils = None def test_normalize_arabic_number(self): actual = self.string_utils.normalize_number("33,825,315") expected = 33825315 self.assertEqual(actual, expected) actual = self.string_utils.normalize_number("0") expected = 0 self.assertEqual(actual, expected) actual = self.string_utils.normalize_number("-115,859,592") expected = -115859592 self.assertEqual(actual, expected) actual = self.string_utils.normalize_number("(27,540)") expected = -27540 self.assertEqual(actual, expected) actual = self.string_utils.normalize_number("2.85") expected = 2.85 self.assertEqual(actual, expected) actual = self.string_utils.normalize_number("170,270,395.00") expected = 170270395 self.assertEqual(actual, expected) actual = self.string_utils.normalize_number("( 10,117,111)") expected = -10117111 self.assertEqual(actual, expected) def test_normalize_none_number(self): actual = self.string_utils.normalize_number(u"-") expected = None self.assertEqual(actual, expected) actual = self.string_utils.normalize_number(u"") expected = None self.assertEqual(actual, expected) actual = self.string_utils.normalize_number(u"不適用") expected = None self.assertEqual(actual, expected) def test_normalize_chinese_number(self): actual = self.string_utils.normalize_number(u"九十九") expected = 99 self.assertEqual(actual, expected) actual = self.string_utils.normalize_number(u"九十") expected = 90 self.assertEqual(actual, expected) actual = self.string_utils.normalize_number(u"三") expected = 3 self.assertEqual(actual, expected) def test_normalize_percentage(self): actual = self.string_utils.normalize_number(u"20.92%") expected = 0.2092 self.assertAlmostEqual(actual, expected) def test_from_local_string_to_date(self): actual = self.string_utils.from_local_string_to_date(u"2013年12月31日") expected = datetime.date(2013, 12, 31) self.assertEqual(actual, expected) actual = self.string_utils.from_local_string_to_date(u"2012年01月01日") expected = datetime.date(2012, 1, 1) self.assertEqual(actual, expected) actual = self.string_utils.from_local_string_to_date("1962/02/09") expected = datetime.date(1962, 2, 9) self.assertEqual(actual, expected) actual = self.string_utils.from_local_string_to_date(u"2015/08/13") expected = datetime.date(2015, 8, 13) self.assertEqual(actual, expected) actual = self.string_utils.from_local_string_to_date(u"民國103年09月") expected = datetime.date(2014, 9, 30) self.assertEqual(actual, expected) actual = self.string_utils.from_local_string_to_date(u"104") expected = datetime.date(2015, 12, 31) self.assertEqual(actual, expected) def test_roc_era_from_local_string_to_date(self): actual = self.string_utils.from_local_string_to_date(u"99年09月30日") expected = datetime.date(2010, 9, 30) self.assertEqual(actual, expected) actual = self.string_utils.from_local_string_to_date(u"102/05/07") expected = datetime.date(2013, 5, 7) self.assertEqual(actual, expected) def test_from_local_string_to_date_interval(self): actual = self.string_utils.from_local_string_to_date_period(u"2013年01月01日至2013年12月31日") expected = datetime.date(2013, 1, 1), datetime.date(2013, 12, 31) self.assertEqual(actual, expected) def test_roc_era_from_local_string_to_date_period(self): actual = self.string_utils.from_local_string_to_date_period(u"九十八年前三季") expected = datetime.date(2009, 1, 1), datetime.date(2009, 9, 30) self.assertEqual(actual, expected) actual = self.string_utils.from_local_string_to_date_period(u"九十八年第一季") expected = datetime.date(2009, 1, 1), datetime.date(2009, 3, 31) self.assertEqual(actual, expected) actual = self.string_utils.from_local_string_to_date_period(u"100年第一季") expected = datetime.date(2011, 1, 1), datetime.date(2011, 3, 31) self.assertEqual(actual, expected) actual = self.string_utils.from_local_string_to_date_period(u"100年上半年度") expected = datetime.date(2011, 1, 1), datetime.date(2011, 6, 30) self.assertEqual(actual, expected) actual = self.string_utils.from_local_string_to_date_period(u"99年上半年度") expected = datetime.date(2010, 1, 1), datetime.date(2010, 6, 30) self.assertEqual(actual, expected) actual = self.string_utils.from_local_string_to_date_period(u"100年前三季") expected = datetime.date(2011, 1, 1), datetime.date(2011, 9, 30) self.assertEqual(actual, expected) actual = self.string_utils.from_local_string_to_date_period(u"100年度") expected = datetime.date(2011, 1, 1), datetime.date(2011, 12, 31) self.assertEqual(actual, expected) def test_from_date_to_roc_era_string(self): actual = self.string_utils.from_date_to_roc_era_string(datetime.date(2001, 1, 1)) expected = "90" self.assertEqual(actual, expected) def test_from_date_to_2_digit_month_string(self): actual = self.string_utils.from_date_to_2_digit_month_string(datetime.date(2001, 1, 1)) expected = "01" self.assertEqual(actual, expected) actual = self.string_utils.from_date_to_2_digit_month_string(datetime.date(2001, 10, 31)) expected = "10" self.assertEqual(actual, expected) def test_from_date_to_2_digit_quarter_string(self): # spring actual = self.string_utils.from_date_to_2_digit_quarter_string(datetime.date(2001, 1, 1)) expected = "01" self.assertEqual(actual, expected) actual = self.string_utils.from_date_to_2_digit_quarter_string(datetime.date(2001, 3, 31)) expected = "01" self.assertEqual(actual, expected) # summer actual = self.string_utils.from_date_to_2_digit_quarter_string(datetime.date(2001, 4, 1)) expected = "02" self.assertEqual(actual, expected) actual = self.string_utils.from_date_to_2_digit_quarter_string(datetime.date(2001, 6, 30)) expected = "02" self.assertEqual(actual, expected) # fall actual = self.string_utils.from_date_to_2_digit_quarter_string(datetime.date(2001, 7, 1)) expected = "03" self.assertEqual(actual, expected) actual = self.string_utils.from_date_to_2_digit_quarter_string(datetime.date(2001, 9, 30)) expected = "03" self.assertEqual(actual, expected) # winter actual = self.string_utils.from_date_to_2_digit_quarter_string(datetime.date(2001, 10, 1)) expected = "04" self.assertEqual(actual, expected) actual = self.string_utils.from_date_to_2_digit_quarter_string(datetime.date(2001, 12, 31)) expected = "04" self.assertEqual(actual, expected) def test_from_date_to_1_digit_quarter_string(self): # spring actual = self.string_utils.from_date_to_1_digit_quarter_string(datetime.date(2001, 1, 1)) expected = "1" self.assertEqual(actual, expected) actual = self.string_utils.from_date_to_1_digit_quarter_string(datetime.date(2001, 3, 31)) expected = "1" self.assertEqual(actual, expected) # summer actual = self.string_utils.from_date_to_1_digit_quarter_string(datetime.date(2001, 4, 1)) expected = "2" self.assertEqual(actual, expected) actual = self.string_utils.from_date_to_1_digit_quarter_string(datetime.date(2001, 6, 30)) expected = "2" self.assertEqual(actual, expected) # fall actual = self.string_utils.from_date_to_1_digit_quarter_string(datetime.date(2001, 7, 1)) expected = "3" self.assertEqual(actual, expected) actual = self.string_utils.from_date_to_1_digit_quarter_string(datetime.date(2001, 9, 30)) expected = "3" self.assertEqual(actual, expected) # winter actual = self.string_utils.from_date_to_1_digit_quarter_string(datetime.date(2001, 10, 1)) expected = "4" self.assertEqual(actual, expected) actual = self.string_utils.from_date_to_1_digit_quarter_string(datetime.date(2001, 12, 31)) expected = "4" self.assertEqual(actual, expected) def test_is_match_seperation(self): pattern = u"^(-| |=)*$" self.assertTrue(self.string_utils.is_match(pattern, u"====== ======")) self.assertTrue(self.string_utils.is_match(pattern, u"------ ------")) self.assertFalse(self.string_utils.is_match(pattern, u"同時影響現金及非現金項目之投資活動:")) def test_match_account(self): pattern = u"^([^\s]*):$" actual = self.string_utils.match(pattern, u"營業活動之現金流量:") expected = [u"營業活動之現金流量"] self.assertEqual(actual, expected)
class StockSymbolAssembler(): def __init__(self): self.base_xpath = '//html/body' self.string_utils = StringUtils() def assemble(self, param): content = self.string_utils.normalize_string(param['content']) html_object = lxml.html.fromstring(content) relative_html_object = self.__traverse_to_relative_html_object(html_object) release_date = self.__assemble_release_date(relative_html_object) column_name_list = self.__assemble_column_name_list(relative_html_object) row_list = self.__assemble_row_list(relative_html_object) return StockSymbolDao(column_name_list, row_list, release_date) def __traverse_to_relative_html_object(self, html_object): relative_html_object_list = html_object.xpath(self.base_xpath) assert len(relative_html_object_list) == 1, 'invalid base_xpath' return relative_html_object_list[0] def __assemble_release_date(self, relative_html_object): # try to get release date table_tags = relative_html_object.xpath('./table') assert len(table_tags) > 0, 'invalid table_tags' headline_tags = table_tags[0].xpath('./h2') assert len(headline_tags) > 0, 'invalid headline_tags' headline_texts = headline_tags[1].xpath('./strong/center') groups = self.string_utils.match(u'^最近更新日期:(.*)$', headline_texts[0].text.strip()) assert len(groups) > 0, 'could not match ^最近更新日期:(.*)$' release_date = self.string_utils.from_local_string_to_date(groups[0]) return release_date def __assemble_column_name_list(self, relative_html_object): # traverse and sanity check tr_tags = relative_html_object.xpath('./table[@class="h4"]/tr') assert len(tr_tags) > 0, 'invalid tr_tags' # traverse and sanity check original_column_name_list = tr_tags[0].xpath('./td/text()') # handle the first column name: '有價證券代號及名稱' combined_column_name = original_column_name_list[0].strip() assert combined_column_name == u'有價證券代號及名稱', 'should be 有價證券代號及名稱 in unicode' # the chinese character '及' means 'and' so we need to seperate this column name seperated_column_name_list = combined_column_name.split(u'及') assert len(seperated_column_name_list) == 2 column_name_list = seperated_column_name_list + original_column_name_list[1:] assert len(column_name_list) == 8, 'invalid column_name_list size, should be 8' return column_name_list def __assemble_row_list(self, relative_html_object): # skip one row of column name list tr_tags = relative_html_object.xpath('./table[@class="h4"]/tr')[1:] row_list = [] for tr_tag in tr_tags: row = self.__assemble_row(tr_tag) # if there is only one cell '股票' in row, skip it if row: row_list.append(row) return row_list def __assemble_row(self, relative_html_object): td_tags = relative_html_object.xpath('./td') # we could not handle empty string between td tag if we use xpath './td/text()' # so we need to check each td.text one by one. td_texts = self.__get_lxml_text_list(td_tags) # if there is only one cell '股票', return None if len(td_texts) == 1: return None # sanity check assert len(td_texts) == 7 # handle the first cell: '有價證券代號及名稱' # it should be seperated as stock symbol and stock name combined_cell = td_texts[0].strip() seperated_cell_list = combined_cell.split() assert len(seperated_cell_list) == 2 # convert to datetime.date type listing_date = self.string_utils.from_local_string_to_date(td_texts[2]) row = seperated_cell_list + [td_texts[1]] + [listing_date] + td_texts[3:] return row def __get_lxml_text_list(self, tag_list): text_list = [] for tag in tag_list: if tag.text is None: text_list.append('') else: text_list.append(tag.text) return text_list
class GeminiParser: def __init__(self): self.base_xpath = "//html/body/center" self.date_utils = DateUtils() self.string_utils = StringUtils() def parse(self, content): html_object = self.__get_html_object(content) relative_html_object = self.__traverse_to_relative_html_object(html_object) column_name_list = self.__parse_column_name_list(relative_html_object) row_list = self.__parse_row_list(relative_html_object) release_date = self.__parse_release_date(relative_html_object) return column_name_list, row_list, release_date def __get_html_object(self, content): content = self.string_utils.normalize_string(content) content = content.replace(u"<br>", u"") return lxml.html.fromstring(content) def __traverse_to_relative_html_object(self, html_object): relative_html_object_list = html_object.xpath(self.base_xpath) assert len(relative_html_object_list) > 0, "invalid base_xpath" return relative_html_object_list[0] def __parse_column_name_list(self, relative_html_object): # traverse and sanity check table_tags = relative_html_object.xpath("./table") assert len(table_tags) > 1, "invalid table_tags" # skip first table of description about IFRS inner_table_tags = table_tags[1].xpath("./tr/td/table/tr/td/table") assert len(inner_table_tags) > 0, "invalid inner_table_tags" tr_tags = inner_table_tags[0].xpath("./tr") assert len(tr_tags) > 1, "invalid tr_tags" th_texts = tr_tags[1].xpath("./th/text()") return th_texts def __parse_row_list(self, relative_html_object): # traverse and sanity check table_tags = relative_html_object.xpath("./table") assert len(table_tags) > 1, "invalid table_tags" # skip first table of description about IFRS inner_table_tags = table_tags[1].xpath("./tr/td/table/tr/td/table") assert len(inner_table_tags) > 0, "invalid inner_table_tags" all_tr_tags = [] # every inner_table represents an industry for inner_table_tag in inner_table_tags: tr_tags = inner_table_tag.xpath("./tr") assert len(tr_tags) > 2, "invalid tr_tags" # first two rows are headers # last row is u'合計' all_tr_tags += tr_tags[2:-1] return [self.__parse_row(tr_tag) for tr_tag in all_tr_tags] def __parse_row(self, relative_html_object): td_texts = relative_html_object.xpath("./td/text()") assert len(td_texts) == 10, "invalid td_texts size, should be 10" items = td_texts[:2] numbers = [] for td_text in td_texts[2:]: number = self.string_utils.normalize_number(td_text) numbers.append(number) return items + numbers def __parse_release_date(self, relative_html_object): div_tags = relative_html_object.xpath("./div") assert len(div_tags) > 0, "invalid div_tags" groups = self.string_utils.match(u"^出表日期:(.*)$", div_tags[-1].text.strip()) assert len(groups) > 0, "could not match ^出表日期:(.*)$" release_date = self.string_utils.from_local_string_to_date(groups[0]) return release_date