@@ -679,22 +679,23 @@ def dates(self,normalize=True,convert=True):
679679 return results
680680
681681 # extract times like 12:00 or délután 4
682- def times (self ,normalize = True ,convert = True ,current = - 1 ):
682+ def times (self ,normalize = True ,convert = True ,current = False ):
683683 if self .text :
684- matches = _re .findall (r'((?:ma\s?|holnap(?:\s?ut[aá]n)?\s?|tegnap(?:\s?el[oöő]t+)?\s?)?(?: reggel\s?|hajnal(?:i|ban)?\s?|d[eé]lel[oöő]t+\s?|d\.?e\.?\s?|d[eé]lut[aá]n\s?|d\.?u\.?\s?|este\s?|[eé]j+el\s?)?\,?\s?(?:[12345]?\d\s?perc+el\s)?(?:(?:h[aá]rom)?negyed\s?|f[eé]l\s?)?[012]?\d\s?(?:\:\s?|\-?kor\s?|[ oó]r[aá]\w{0,3}\s?)?(?:el[oöő]t+\s?|ut[aá]n\s?)?(?:[0123456]?\d[\-\s]?(?:kor|perc\w{0,3})?)?\,?\s?(?:ma\s?|holnap(?:\s?ut[aá]n)?\s?|tegnap(?:\s?el[oöő]t+)?\s?)?(?: reggel\s?|hajnal(?:i|ban)?\s?|d[eé]lel[oöő]t+\s?|d\.?e\.?\s?|d[eé]lut[aá]n\s?|d\.?u\.?\s?|este\s?|[eé]j+el\s?)?)' , re .IGNORECASE , self ._ntext_ if convert else self ._text_ )
684+ matches = _re .findall (r'((?:ma\s?|holnap(?:\s?ut[aá]n)?\s?|tegnap(?:\s?el[oöő]t+)?\s?)?(reggel\s?|hajnal(?:i|ban)?\s?|d[eé]lel[oöő]t+\s?|d\.?e\.?\s?|d[eé]lut[aá]n\s?|d\.?u\.?\s?|este\s?|[eé]j+el\s?)?\,?\s?(?:[12345]?\d\s?perc+el\s)?(?:(?:h[aá]rom)?negyed\s?|f[eé]l\s?)?[012]?\d\s?(?:\:\s?|\-?kor\s?|\-?t[oóöő]l|\-?ig|\-?r[ae]|[ oó]r[aá]\w{0,3}\s?)?(?:el[oöő]t+\s?|ut[aá]n\s?)?(?:[0123456]?\d[\-\s]?(?:kor|t[oóöő]l|ig|r[ae]| perc\w{0,3})?)?\,?\s?(?:ma\s?|holnap(?:\s?ut[aá]n)?\s?|tegnap(?:\s?el[oöő]t+)?\s?)?(?(1)( reggel\s?|hajnal(?:i|ban)?\s?|d[eé]lel[oöő]t+\s?|d\.?e\.?\s?|d[eé]lut[aá]n\s?|d\.?u\.?\s?|este\s?|[eé]j+el\s?) )?)' , re .IGNORECASE , self ._ntext_ if convert else self ._text_ )
685685 if normalize :
686686 results = []
687- for item in matches :
687+ for _item in matches :
688+ item = _item [0 ]
688689 if len (item .strip ())> 2 :
689- item = ' ' + item + ' '
690- hour = "00"
690+ item = ' ' + item . lower () + ' '
691+ hour = "00"
691692 minute = "00"
692693 pm = False
693- zero = False
694- elott = False
695- hour_matches = _re .findall (r'\D([012]?\d(?!\d))\D*?(?!perc)(?:\:|\-?kor|[ oó]r[aá])?' , re .IGNORECASE , item )
696- minute_matches = _re .findall (r'(?!negyed|f[eé]l)\D([0123456]?\d(?!\d))\D*?(?![oó]ra )(?:\-?kor|perc)?' , re .IGNORECASE , item )
697- quarter_matches = _re .findall (r'((?:h[aá]rom)?negyed|f[eé]l)' , re .IGNORECASE , item )
694+ zero = False
695+ elott = False
696+ hour_matches = _re .findall (r'\D([012]?\d(?!\d))\D*?(?!perc)(?:\:|\-?kor|\-?t[oóöő]l|\-?ig|\-?r[ae]|[ oó]r[aá]\w* )?' , re .IGNORECASE , item )
697+ minute_matches = _re .findall (r'(?!negyed|f[eé]l)\D([0123456]?\d(?!\d))\D*?(?![oó]r[aá] )(?:\-?kor|\-?t[oóöő]l|\-?ig|\-?r[ae]| perc\w* )?' , re .IGNORECASE , item )
698+ quarter_matches = _re .findall (r'((?:h[aá]rom)?negyed|f[eé]l)' , re .IGNORECASE , item )
698699 am_matches = _re .findall (r'(reggel|hajnal|d[eé]lel[oöő]t|d\.?e\.?)' , re .IGNORECASE , item )
699700 pm_matches = _re .findall (r'(d[eé]lut[aá]n|d\.?u\.?|este|[eé]j+el)' , re .IGNORECASE , item )
700701 if len (hour_matches ) in (1 ,2 ):
@@ -763,15 +764,14 @@ def times(self,normalize=True,convert=True,current=-1):
763764 if pm_matches :
764765 pm = True
765766 elif not am_matches :
766- if current :
767- if current >= 0 :
768- now = current
769- else :
770- now = datetime .datetime .now ().hour
771- if 'holnap' in item and hour < 9 :
772- pm = True
773- elif hour < 12 and now > hour :
774- pm = True
767+ if current is not False :
768+ now = current
769+ else :
770+ now = datetime .datetime .now ().hour
771+ if 'holnap' in item and hour < 9 :
772+ pm = True
773+ elif hour < 12 and now > hour :
774+ pm = True
775775 if pm and hour <= 12 :
776776 hour += 12
777777 hour %= 24
@@ -780,7 +780,7 @@ def times(self,normalize=True,convert=True,current=-1):
780780 results .append (str (hour ).zfill (2 )+ ':' + str (minute ).zfill (2 ))
781781 return results
782782 else :
783- return [item .strip () for item in matches if len (item .strip ())> 2 ]
783+ return [item [ 0 ] .strip () for item in matches if len (item [ 0 ] .strip ())> 2 ]
784784 return []
785785
786786 # extract list of time durations
@@ -795,13 +795,13 @@ def durations(self,normalize=True,convert=True):
795795 val = 0
796796 for sub_item in sub_matches :
797797 match = sub_item .lower ().replace (',' ,'.' )
798- sval = ''
798+ sval = ''
799799 for char in match :
800800 if char .isdigit () or char == '.' :
801801 sval += char
802802 else :
803803 break
804- sval = float (sval )
804+ sval = float (sval )
805805 mpx = 1
806806 if 'tized' in match or 'tízed' in match :
807807 mpx = 0.1
@@ -953,11 +953,57 @@ def emails(self):
953953 return _re .findall (r'\b([\w\d\-\_\.]+\@[\w\d\-\_\.]+\.\w{2,4}(?:\.\w{2,4})?)\b' , re .IGNORECASE , self .text )
954954 return []
955955
956+ # extract relative dates like tomorrow or wednesday
957+ def relative_dates (self ,normalize = True ,current = False ):
958+ if self .text :
959+ matches = _re .findall (r'\b((?:(?:meg)?el[oöő]z[oöő]|m[uú]lt|(?:r[aá])?k[oö]vetkez[oöő]|j[oö]v[oöő])?\s?(?:h[eé]t(?:i|en)?\s?)?(?:tegnap(?:el[oöő]t+)?|holnap(?:ut[aá]n)?|m[aá](?:i nap)?|h[eé]tf[oöő]|ked+|szerd[aá]|cs[uü]t[oö]rt[oö]k|p[eé]ntek|szo[nm]bat|vas[aá]rnap))(?:[aáeoö][dm])?(?:ig|r[ae]|t[oóöő]l|[aáeoöő]?t|[dkmnptv][ae][lk]|[aáeoö]?n)?\b' , re .IGNORECASE , self .text )
960+ if normalize :
961+ if current is not False :
962+ _now = datetime .datetime .strptime (current ,"%Y-%m-%d" )
963+ else :
964+ _now = datetime .datetime .now ()
965+ results = []
966+ for item in matches :
967+ item = item .lower ()
968+ now = _now
969+ if 'holnap' in item :
970+ if 'ut' in item :
971+ now += datetime .timedelta (days = 2 )
972+ else :
973+ now += datetime .timedelta (days = 1 )
974+ elif 'tegnap' in item :
975+ if 'el' in item :
976+ now += datetime .timedelta (days = - 2 )
977+ else :
978+ now += datetime .timedelta (days = - 1 )
979+ elif 'ma' not in item and 'má' not in item :
980+ now -= datetime .timedelta (days = now .weekday ())
981+ if _re .findall (r'((?:meg)?el[oöő]z[oöő]|m[uú]lt)' , re .IGNORECASE , item ):
982+ now -= datetime .timedelta (weeks = 1 )
983+ elif _re .findall (r'((?:r[aá])?k[oö]vetkez[oöő]|j[oö]v[oöő])' , re .IGNORECASE , item ):
984+ now += datetime .timedelta (weeks = 1 )
985+ if 'ked' in item :
986+ now += datetime .timedelta (days = 1 )
987+ elif 'szerd' in item :
988+ now += datetime .timedelta (days = 2 )
989+ elif _re .findall (r'cs[uü]t[oö]rt[oö]k' , re .IGNORECASE , item ):
990+ now += datetime .timedelta (days = 3 )
991+ elif _re .findall (r'p[eé]ntek' , re .IGNORECASE , item ):
992+ now += datetime .timedelta (days = 4 )
993+ elif _re .findall (r'szo[mn]bat' , re .IGNORECASE , item ):
994+ now += datetime .timedelta (days = 5 )
995+ elif _re .findall (r'vas[aá]rnap' , re .IGNORECASE , item ):
996+ now += datetime .timedelta (days = 6 )
997+ results .append (now .strftime ('%Y-%m-%d' ))
998+ return results
999+ else :
1000+ return [item .strip () for item in matches ]
1001+ return []
1002+
9561003 # Converts text representation of numbers to digits
9571004 def _convert_numbers (self ,text ):
9581005 if text :
959- #fix = _re.sub(r'(?<=\d)\s+(?=\d)',re.IGNORECASE,'',text.lower())
960- matches = _re .findall (r'((?:m[ií]n[uú]sz\s?|negat[ií]v\s?)?(?:(?:(?:(?:(?:t[ií]z|h[uú]sz|harminc)(?:[eo]n)?)?(?:nulla|egy|els[eoöő]|k[eé]t+[oöő]?|m[aá]sod(?:ik)?|h[aá]rom|harmadik|n[eé]gy|[oö]t|hat|h[eé]t|nyolc|kilenc)(?:v[ae]n)?)(?:milli[aá]rd|milli[oó]|ezer|sz[aá]z)?\W*)|(?:milli[aá]rd|milli[oó]|ezer|sz[aá]z|t[ií]z|h[uú]sz|harminc|nulla|z[eé]r[oó])\W*)+(?:[aeoö]dik)?(?:j?[aáeéi]+n?)?)\b' , re .IGNORECASE , text .lower ())
1006+ matches = _re .findall (r'((?:m[ií]n[uú]sz\s?|negat[ií]v\s?)?(?:(?:(?:(?:(?:t[ií]z|h[uú]sz|harminc)(?:[eo]n)?)?(?:nulla|eg+y|els[eoöő]|k[eé]t+[oöő]?|m[aá]sod(?:ik)?|h[aá]rom|harmadik|n[eé]g+y|[oö]t|hat|h[eé]t|nyolc|kilenc)(?:v[ae]n)?)(?:milli[aá]rd|milli[oó]|ezer|sz[aá]z)?\W*)|(?:milli[aá]rd|milli[oó]|ezer|sz[aá]z|t[ií]z|h[uú]sz|harminc|nulla|z[eé]r[oó])\W*)+(?:[aeoö]dik)?(?:j?[aáeéi]+[gnt]?|[aáeéoöő]?t|kor|t[oóöő]l|r[ae]|[ckmrtvz]?[ae]l)?)\b' , re .IGNORECASE , text )
9611007 results = {}
9621008 for match in matches :
9631009 value = 0
@@ -968,7 +1014,7 @@ def _convert_numbers(self,text):
9681014 else :
9691015 minus = 1
9701016 minusm = ''
971- parts = _re .findall (r'((?:(?:(?:(?:t[ií]z|h[uú]sz|harminc)(?:[eo]n)?)?(?:nulla|egy |els[eoöő]|k[eé]t+[oöő]?|m[aá]sod(?:ik)?|h[aá]rom|harmadik|n[eé]gy |[oö]t|hat|h[eé]t|nyolc|kilenc)(?:v[ae]n)?)(?:milli[aá]rd|milli[oó]|ezer|sz[aá]z)?|(?:milli[aá]rd|milli[oó]|ezer|sz[aá]z|t[ií]z|h[uú]sz|harminc|nulla|z[eé]r[oó]))\W*)' , re .IGNORECASE , match )
1017+ parts = _re .findall (r'((?:(?:(?:(?:t[ií]z|h[uú]sz|harminc)(?:[eo]n)?)?(?:nulla|eg+y |els[eoöő]|k[eé]t+[oöő]?|m[aá]sod(?:ik)?|h[aá]rom|harmadik|n[eé]g+y |[oö]t|hat|h[eé]t|nyolc|kilenc)(?:v[ae]n)?)(?:milli[aá]rd|milli[oó]|ezer|sz[aá]z)?|(?:milli[aá]rd|milli[oó]|ezer|sz[aá]z|t[ií]z|h[uú]sz|harminc|nulla|z[eé]r[oó]))\W*)' , re .IGNORECASE , match )
9721018 values = []
9731019 for part in parts :
9741020 val = 0
@@ -1045,7 +1091,7 @@ def _convert_numbers(self,text):
10451091
10461092 swap = sorted (results .items (), key = lambda x : x [1 ], reverse = True )
10471093 for item in swap :
1048- text = _re .sub (r'\b(' + re .escape (item [0 ])+ r')(?:[aeoö]dik?)?(?:j?[aáeéi]+n? )?\b' , re .IGNORECASE , str (item [1 ]), text )
1094+ text = _re .sub (r'\b(' + re .escape (item [0 ])+ r')(?:[aeoö]dik?)?(?:j?[aáeéi]+[gnt]?|[aáeéoöő]?t|kor|t[oóöő]l|r[ae]|[ckmrtvz]?[ae]l )?\b' , re .IGNORECASE , str (item [1 ]), text )
10491095 return text
10501096 return ''
10511097
@@ -1063,9 +1109,9 @@ def _convert_numbers_helper(self,match,default):
10631109 return 2
10641110 elif _re .findall (r'(harmadik|h[aá]rom)' , re .IGNORECASE , match ):
10651111 return 3
1066- elif _re .findall (r'n[eé]gy ' , re .IGNORECASE , match ):
1112+ elif _re .findall (r'n[eé]g+y ' , re .IGNORECASE , match ):
10671113 return 4
1068- elif 'egy' in match or 'els' in match :
1114+ elif 'egy' in match or 'els' in match or 'eggy' in match :
10691115 return 1
10701116 elif _re .findall (r'[oö]t' , re .IGNORECASE , match ):
10711117 return 5
0 commit comments