Settings

dateparser’s parsing behavior can be configured by supplying settings as a dictionary to settings argument in dateparser.parse() or DateDataParser constructor.

Note

From dateparser 1.0.0 when a setting with a wrong value is provided, a SettingValidationError is raised.

All supported settings with their usage examples are given below:

Date Order

DATE_ORDER: specifies the order in which date components year, month and day are expected while parsing ambiguous dates. It defaults to MDY which translates to month first, day second and year last order. Characters M, D or Y can be shuffled to meet required order. For example, DMY specifies day first, month second and year last order:

>>> parse('15-12-18 06:00')  # assumes default order: MDY
datetime.datetime(2018, 12, 15, 6, 0)  # since 15 is not a valid value for Month, it is swapped with Day's
>>> parse('15-12-18 06:00', settings={'DATE_ORDER': 'YMD'})
datetime.datetime(2015, 12, 18, 6, 0)

PREFER_LOCALE_DATE_ORDER: defaults to True. Most languages have a default DATE_ORDER specified for them. For example, for French it is DMY:

>>> # parsing ambiguous date
>>> parse('02-03-2016')  # assumes english language, uses MDY date order
datetime.datetime(2016, 2, 3, 0, 0)
>>> parse('le 02-03-2016')  # detects french, hence, uses DMY date order
datetime.datetime(2016, 3, 2, 0, 0)

Note

There’s no language level default DATE_ORDER associated with en language. That’s why it assumes MDY which is :obj:settings <dateparser.conf.settings> default. If the language has a default DATE_ORDER associated, supplying custom date order will not be applied unless we set PREFER_LOCALE_DATE_ORDER to False:

>>> parse('le 02-03-2016', settings={'DATE_ORDER': 'MDY'})
datetime.datetime(2016, 3, 2, 0, 0)  # MDY didn't apply

>>> parse('le 02-03-2016', settings={'DATE_ORDER': 'MDY', 'PREFER_LOCALE_DATE_ORDER': False})
datetime.datetime(2016, 2, 3, 0, 0)  # MDY worked!

STRICT_DATE_ORDER: defaults to 'none'. DATE_ORDER is a preference: a number that does not fit where the order puts it is read as another part of the date. 13 is not a valid month, so it is read as the day where MDY puts the month, and any number that is not a valid day or month is a valid two-digit year, so 32 becomes the year 2032:

>>> parse('13/12/10', settings={'DATE_ORDER': 'MDY'})
datetime.datetime(2010, 12, 13, 0, 0)
>>> parse('32 DEC 10', settings={'DATE_ORDER': 'DMY'})
datetime.datetime(2032, 12, 10, 0, 0)

Set it to 'year' to read a two-digit year only from the position DATE_ORDER gives the year, or to 'all' to read the day, month and year only in the order DATE_ORDER gives. A date string that does not fit then gives None:

>>> parse('32 DEC 10', settings={'DATE_ORDER': 'DMY', 'STRICT_DATE_ORDER': 'year'})
>>> parse('13/12/10', settings={'DATE_ORDER': 'MDY', 'STRICT_DATE_ORDER': 'year'})
datetime.datetime(2010, 12, 13, 0, 0)
>>> parse('13/12/10', settings={'DATE_ORDER': 'MDY', 'STRICT_DATE_ORDER': 'all'})

'year' and 'all' raise a SettingValidationError unless DATE_ORDER is given too. Only the parts read from numbers are checked, and a four-digit year is read wherever it is, so '32 DEC' with DMY is still December 2032. Some dates that are read correctly without the setting are rejected, such as '95年12月10日' with MDY: once translated, it is 95-12-10, with nothing left to mark the year.

Timezone Related Configurations

TIMEZONE: defaults to local timezone. When specified, resultant datetime is localized with the given timezone. Can be timezone abbreviation or any of tz database name as given here.

>>> parse('January 12, 2012 10:00 PM', settings={'TIMEZONE': 'US/Eastern'})
datetime.datetime(2012, 1, 12, 22, 0)

TO_TIMEZONE: defaults to None. When specified, resultant datetime converts according to the supplied timezone:

>>> settings = {'TIMEZONE': 'UTC', 'TO_TIMEZONE': 'US/Eastern'}
>>> parse('January 12, 2012 10:00 PM', settings=settings)
datetime.datetime(2012, 1, 12, 17, 0)

TIMEZONE_ABBREVIATIONS: maps timezone abbreviations found in date strings to the UTC offset to use for them, as a timedelta. Use it for abbreviations with several meanings, e.g. IST means Israel Standard Time by default. Keys are case-sensitive and must be abbreviations that dateparser supports, e.g. ChST. TIMEZONE and TO_TIMEZONE keep the default offsets.

New in version 1.5.0.

>>> from datetime import timedelta
>>> settings = {
...     'TIMEZONE_ABBREVIATIONS': {'IST': timedelta(hours=5, minutes=30)},
...     'TO_TIMEZONE': 'UTC',
... }
>>> parse('12 Feb 2015 10:56 PM IST', settings=settings)
datetime.datetime(2015, 2, 12, 17, 26, tzinfo=<StaticTzInfo 'UTC'>)

RETURN_AS_TIMEZONE_AWARE: if True returns tz aware datetime objects in case timezone is detected in the date string.

>>> parse('30 mins ago', settings={'RETURN_AS_TIMEZONE_AWARE': True})
datetime.datetime(2017, 3, 13, 1, 43, 10, 243565, tzinfo=<DstTzInfo 'Asia/Karachi' PKT+5:00:00 STD>)

>>> parse('12 Feb 2015 10:56 PM EST', settings={'RETURN_AS_TIMEZONE_AWARE': False})
datetime.datetime(2015, 2, 12, 22, 56)

Handling Incomplete Dates

PREFER_DAY_OF_MONTH: it comes handy when the date string is missing the day part. It defaults to current and can be first and last denoting first and last day of months respectively as values:

>>> from dateparser import parse
>>> parse('December 2015')  # default behavior
datetime.datetime(2015, 12, 16, 0, 0)
>>> parse('December 2015', settings={'PREFER_DAY_OF_MONTH': 'last'})
datetime.datetime(2015, 12, 31, 0, 0)
>>> parse('December 2015', settings={'PREFER_DAY_OF_MONTH': 'first'})
datetime.datetime(2015, 12, 1, 0, 0)

PREFER_MONTH_OF_YEAR: Similarly, another useful thing when the date string is missing the month part. It defaults to current and can be first and last denoting first and last month of year respectively as values:

>>> from dateparser import parse
>>> parse("2015") # default behavior
datetime.datetime(2015, 3, 27, 0, 0)
>>> parse("2015", settings={"PREFER_MONTH_OF_YEAR": "last"})
datetime.datetime(2015, 12, 27, 0, 0)
>>> parse("2015", settings={"PREFER_MONTH_OF_YEAR": "first"})
datetime.datetime(2015, 1, 27, 0, 0)
>>> parse("2015", settings={"PREFER_MONTH_OF_YEAR": "current"}) # it exactly behaves like default one
datetime.datetime(2015, 3, 27, 0, 0)

It also picks the month of a calendar quarter, such as 2015Q2 or Q2 2015, where current means the month that is as far into the quarter as the current month is into its own:

>>> parse("2015Q2", settings={"PREFER_MONTH_OF_YEAR": "last", "PREFER_DAY_OF_MONTH": "last"})
datetime.datetime(2015, 6, 30, 0, 0)

PREFER_DATES_FROM: defaults to current_period and can have past and future as values.

If date string is missing some part, this option ensures consistent results depending on the past or future preference, for example, assuming current date is June 16, 2015:

>>> from dateparser import parse
>>> parse('March')
datetime.datetime(2015, 3, 16, 0, 0)
>>> parse('March', settings={'PREFER_DATES_FROM': 'future'})
datetime.datetime(2016, 3, 16, 0, 0)
>>> # parsing with preference set for 'past'
>>> parse('August', settings={'PREFER_DATES_FROM': 'past'})
datetime.datetime(2014, 8, 16, 0, 0)

RELATIVE_BASE: allows setting the base datetime to use for interpreting partial or relative date strings. Defaults to the current date and time.

For example, assuming current date is June 16, 2015:

>>> from dateparser import parse
>>> parse('14:30')
datetime.datetime(2015, 6, 16, 14, 30)
>>> parse('14:30', settings={'RELATIVE_BASE': datetime.datetime(2020, 1, 1)})
datetime.datetime(2020, 1, 1, 14, 30)
>>> parse('tomorrow', settings={'RELATIVE_BASE': datetime.datetime(2020, 1, 1)})
datetime.datetime(2020, 1, 2, 0, 0)

RETURN_TIME_SPAN

default: False

When enabled, search_dates detects time span expressions (e.g., “past month”, “last week”) and returns start/end dates instead of single dates.

DEFAULT_START_OF_WEEK

default: 'monday'

Sets which day starts the week for time span calculations. Options: 'monday', 'sunday'.

DEFAULT_DAYS_IN_MONTH

default: 30

Sets the number of days to use for “past month” and similar expressions in time span detection.

STRICT_PARSING: defaults to False.

When set to True if missing any of day, month or year parts, it does not return any result altogether.:

>>> parse('March', settings={'STRICT_PARSING': True})
None

REQUIRE_PARTS: ensures results are dates that have all specified parts. It defaults to [] and can include day, month and/or year.

For example, assuming current date is June 16, 2019:

>>> parse('2012') # default behavior
datetime.datetime(2012, 6, 16, 0, 0)
>>> parse('2012', settings={'REQUIRE_PARTS': ['month']})
None
>>> parse('March 2012', settings={'REQUIRE_PARTS': ['day']})
None
>>> parse('March 12, 2012', settings={'REQUIRE_PARTS': ['day']})
datetime.datetime(2012, 3, 12, 0, 0)
>>> parse('March 12, 2012', settings={'REQUIRE_PARTS': ['day', 'month', 'year']})
datetime.datetime(2012, 3, 12, 0, 0)

IGNORE_SURROUNDING_TEXT: defaults to False. By default, the whole input string must be a date, so a date wrapped in surrounding text is not parsed. When this setting is enabled and the whole string could not be parsed as a date, dateparser retries ignoring the words at the beginning and at the end of the string that the tried language does not recognize as date-related:

>>> parse('Actualisé le 17 avril 2019', languages=['fr'])
None
>>> parse('Actualisé le 17 avril 2019', languages=['fr'], settings={'IGNORE_SURROUNDING_TEXT': True})
datetime.datetime(2019, 4, 17, 0, 0)
>>> parse('Published on 16/04/2019', settings={'IGNORE_SURROUNDING_TEXT': True})
datetime.datetime(2019, 4, 16, 0, 0)

Only the edges of the string are affected. An unrecognized word inside the date still prevents parsing, and strings that can be parsed as a whole are parsed exactly as if the setting was disabled:

>>> parse('17 foobar avril 2019', languages=['fr'], settings={'IGNORE_SURROUNDING_TEXT': True})
None

Stripping stops at the first token the language does recognize, and a number counts as recognized. So surrounding text that contains a number of its own (an id, a page number, a count) is not ignored, and the string is not parsed:

>>> parse('invoice 12345 paid on 3 March 2019', settings={'IGNORE_SURROUNDING_TEXT': True})
None

To extract a date out of arbitrary text that does not follow this shape, use dateparser.search.search_dates() instead, which scans the whole string rather than only trimming its edges.

Warning

When this setting is enabled, discarded edge text is dropped even if it could have changed the meaning of the date, and what remains is parsed as if it were the whole input. As a result, strings that merely contain date-like words can produce a date, e.g. 'Chapter 12 March of the Penguins'. A recognized timezone is kept and applied only when it is the last token of the string (e.g. '... 1:21 PM EST'); a timezone before the date, one followed by further unrecognized text, or any timezone name the parser does not recognize, is discarded with the rest of the surrounding text (use dateparser.search.search_dates() for such input). Enable this setting only when the input is expected to contain a date, and consider combining it with STRICT_PARSING or REQUIRE_PARTS to discard remainders that are not complete dates.

Language Detection

SKIP_TOKENS: it is a list of tokens to discard while detecting language. Defaults to ['t'] which skips T in iso format datetime string .e.g. 2015-05-02T10:20:19+0000.:

>>> from dateparser.date import DateDataParser
>>> DateDataParser(settings={'SKIP_TOKENS': ['de']}).get_date_data(u'27 Haziran 1981 de')  # Turkish (at 27 June 1981)
DateData(date_obj=datetime.datetime(1981, 6, 27, 0, 0), period='day', locale='tr', parts=('year', 'month', 'day'))

NORMALIZE: applies unicode normalization (removing accents, diacritics…) when parsing the words. Defaults to True.

>>> dateparser.parse('4 decembre 2015', settings={'NORMALIZE': False})
# It doesn't work as the expected input should be '4 décembre 2015'

>>> dateparser.parse('4 decembre 2015', settings={'NORMALIZE': True})
datetime.datetime(2015, 12, 4, 0, 0)

Default languages

DEFAULT_LANGUAGES: a list of ISO 639 language codes to fall back to when parsing fails with every other language, i.e. those given through the languages or locales arguments of dateparser.parse() or those detected. These fallback languages are tried even if the input contains no word of theirs:

>>> from dateparser import parse
>>> parse('3 de marzo de 2020', languages=['en'])
>>> parse('3 de marzo de 2020', languages=['en'], settings={'DEFAULT_LANGUAGES': ['es']})
datetime.datetime(2020, 3, 3, 0, 0)

It is especially useful with detect_languages_function, to cover input for which language detection fails.

Language Order

USE_GIVEN_LANGUAGE_ORDER: defaults to False. By default, the languages and locales passed through the languages and locales arguments are tried in the order of the most common languages, regardless of the order in which they are given. Set this to True to instead try them in the order in which they are given, which is useful when that order already reflects a preference (for example, the output of a language detector ordered by confidence):

>>> import dateparser
>>> dateparser.parse('11/12/2020', languages=['es', 'en'])
datetime.datetime(2020, 11, 12, 0, 0)
>>> dateparser.parse('11/12/2020', languages=['es', 'en'], settings={'USE_GIVEN_LANGUAGE_ORDER': True})
datetime.datetime(2020, 12, 11, 0, 0)

It also applies to the languages given through the DEFAULT_LANGUAGES setting when they are used as a fallback.

Note

This setting only reorders the languages and locales that are given. If neither languages nor locales is provided, it has no effect and the default order is used:

>>> dateparser.parse('11/12/2020', settings={'USE_GIVEN_LANGUAGE_ORDER': True})
datetime.datetime(2020, 11, 12, 0, 0)

Note

This setting does not affect dateparser.search.search_dates(), which detects a single most likely language and normalizes the text before parsing, so it does not use the given language order to disambiguate values such as 11/12.

Optional language detection

LANGUAGE_DETECTION_CONFIDENCE_THRESHOLD: defaults to 0.5. It is a float of minimum required confidence for the custom language detection:

>>> from dateparser import parse
>>> parse('3 de marzo de 2020', settings={'LANGUAGE_DETECTION_CONFIDENCE_THRESHOLD': 0.5}, detect_languages_function=detect_languages)

Other settings

RETURN_TIME_AS_PERIOD: returns time as period in date object, if time component is present in date string. Defaults to False.

>>> ddp = DateDataParser(settings={'RETURN_TIME_AS_PERIOD': True})
>>> ddp.get_date_data('vr jan 24, 2014 12:49')
DateData(date_obj=datetime.datetime(2014, 1, 24, 12, 49), period='time', locale='nl', parts=('year', 'month', 'day', 'time'))

PARTS_OF_DAY: a PartsOfDay object that sets the time to use for each part of the day, e.g. for tonight. Defaults to PartsOfDay().

New in version 1.5.0:

>>> from datetime import time
>>> from dateparser import PartsOfDay
>>> ddp = DateDataParser(settings={'PARTS_OF_DAY': PartsOfDay(night=time(22)), 'RETURN_TIME_AS_PERIOD': True})
>>> ddp.get_date_data('5 January 2026 at night')
DateData(date_obj=datetime.datetime(2026, 1, 5, 22, 0), period='part_of_day', locale='en', part_of_day=<PartOfDay.NIGHT: 'night'>)

class dateparser.PartsOfDay(early_morning: time = datetime.time(3, 0), morning: time = datetime.time(9, 0), afternoon: time = datetime.time(15, 0), evening: time = datetime.time(18, 0), night: time = datetime.time(20, 0))

Time of the day to use for each PartOfDay, for the PARTS_OF_DAY setting.

New in version 1.5.0.

afternoon: time = datetime.time(15, 0)

early_morning: time = datetime.time(3, 0)

evening: time = datetime.time(18, 0)

morning: time = datetime.time(9, 0)

night: time = datetime.time(20, 0)

class dateparser.PartOfDay(value, names=None, *, module=None, qualname=None, type=None, start=1, boundary=None)

Part of the day that a date string refers to, e.g. "tonight".

See DateData.part_of_day.

New in version 1.5.0.

AFTERNOON = 'afternoon'

EARLY_MORNING = 'early_morning'

Between midnight and dawn, e.g. "in the small hours" or Spanish "de madrugada".

EVENING = 'evening'

MORNING = 'morning'

NIGHT = 'night'

PARSERS: it is a list of names of parsers to try, allowing to customize which parsers are tried against the input date string, and in which order they are tried.

The following parsers exist:

• 'timestamp': If the input string starts with 10 digits, optionally followed by additional digits or a period (.), those first 10 digits are interpreted as Unix time.

• 'negative-timestamp': 'timestamp' for negative timestamps. For example, parses -186454800000 as 1964-02-03T23:00:00.

• 'relative-time': Parses dates and times expressed in relation to the current date and time (e.g. “1 day ago”, “in 2 weeks”).

• 'custom-formats': Parses dates that match one of the date formats in the list of the date_formats parameter of dateparser.parse() or DateDataParser.get_date_data. When a date string matches none of the formats, the other parsers still try it. To accept only the given formats, set PARSERS to ['custom-formats']: no other parser is tried then, and without date_formats nothing is parsed.

• 'absolute-time': Parses dates and times expressed in absolute form (e.g. “May 4th”, “1991-05-17”). It takes into account settings such as DATE_ORDER or PREFER_LOCALE_DATE_ORDER.

• 'no-spaces-time': Parses dates and times that consist in only digits or a combination of digits and non-digits where the first non-digit it’s a colon (e.g. “121994”, “11:052020”). It’s not included in the default parsers and it can produce false positives frequently.

dateparser.settings.default_parsers contains the default value of PARSERS (the list above, in that order) and can be used to write code that changes the parsers to try without skipping parsers that may be added to Dateparser in the future. For example, to ignore relative times:

>>> from dateparser_data.settings import default_parsers
>>> parsers = [parser for parser in default_parsers if parser != 'relative-time']
>>> parse('today', settings={'PARSERS': parsers})

CACHE_SIZE_LIMIT: limits the size of caches, that store data for already processed dates. Default to 1000, but you can set 0 for turning off the limit.