Pular para conteúdo

Business Day Tools

count(start, end)

Count business days between start (inclusive) and end (exclusive) with Brazilian holiday adjustment and per-row holiday regime selection.

ORDER PRESERVATION (critical): The output order ALWAYS matches the element-wise order of the original inputs. No sorting, deduplication, alignment or reshaping is performed. If you pass arrays, the i-th result corresponds to the i-th pair of (start, end) after broadcasting. This guarantees safe assignment back to the originating DataFrame.

Holiday regime: For each start value, the holiday list (old vs. new) is chosen based on the transition date 2023-12-26 (TRANSITION_DATE). Starts before the transition use the old list for that row's count; starts on/after use the new list.

Null propagation: If any scalar argument is null, returns None. Nulls inside array inputs yield nulls in corresponding result positions.

Return type: If both inputs are scalars (non-null) an int is returned; otherwise a polars.Series of int counts (name: 'bday_count'). If a null scalar short-circuits, None is returned.

Parameters:

Name Type Description Default
start None | DateLike | ArrayLike

Single date or collection (inclusive boundary).

 required
end None | DateLike | ArrayLike

Single date or collection (exclusive boundary).

 required

Returns:

Type Description
None | int | Series

int | pl.Series | None: Returns an integer or None if start and end are single dates, or a Series if any of them is an array of dates.
Notes
  • This function is a wrapper around polars.business_day_count.
  • The holiday list is determined per-row based on the start date.

Examples:

>>> from pyield import bday
>>> bday.count("15-12-2023", "01-01-2024")
10

Total business days in January and February since the start of the year

>>> bday.count(start="01-01-2024", end=["01-02-2024", "01-03-2024"])
shape: (2,)
Series: 'bday_count' [i64]
[
    22
    41
]

The remaining business days from January/February until the end of the year

>>> bday.count(["01-01-2024", "01-02-2024"], "01-01-2025")
shape: (2,)
Series: 'bday_count' [i64]
[
    253
    231
]

The total business days in January and February of 2024

>>> bday.count(["01-01-2024", "01-02-2024"], ["01-02-2024", "01-03-2024"])
shape: (2,)
Series: 'bday_count' [i64]
[
    22
    19
]

Null values are propagated

>>> bday.count(None, "01-01-2024")  # None start

>>> bday.count("01-01-2024", None)  # None end

>>> bday.count("01-01-2024", ["01-02-2024", None])  # None in end array
shape: (2,)
Series: 'bday_count' [i64]
[
    22
    null
]

>>> start_dates = ["01-01-2024", "01-02-2024", "01-03-2024"]
>>> bday.count(start_dates, "01-01-2025")
shape: (3,)
Series: 'bday_count' [i64]
[
    253
    231
    212
]
Source code in pyield/bday/core.py 
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
def count(
    start: None | DateLike | ArrayLike,
    end: None | DateLike | ArrayLike,
) -> None | int | pl.Series:
    """
    Count business days between `start` (inclusive) and `end` (exclusive) with
    Brazilian holiday adjustment and per-row holiday regime selection.

    ORDER PRESERVATION (critical): The output order ALWAYS matches the element-wise
    order of the original inputs. No sorting, deduplication, alignment or reshaping is
    performed. If you pass arrays, the i-th result corresponds to the i-th pair of
    (`start`, `end`) after broadcasting. This guarantees safe assignment back to the
    originating DataFrame.

    Holiday regime: For each `start` value, the holiday list (old vs. new) is chosen
    based on the transition date 2023-12-26 (`TRANSITION_DATE`). Starts before the
    transition use the old list for that row's count; starts on/after use the new list.

    Null propagation: If any scalar argument is null, returns `None`. Nulls inside
    array inputs yield nulls in corresponding result positions.

    Return type: If both inputs are scalars (non-null) an `int` is returned; otherwise
    a `polars.Series` of int counts (name: 'bday_count').
    If a null scalar short-circuits, `None` is returned.

    Args:
        start: Single date or collection (inclusive boundary).
        end: Single date or collection (exclusive boundary).

    Returns:
        int | pl.Series | None: Returns an integer or None if `start` and `end` are
            single dates, or a Series if any of them is an array of dates.

    Notes:
        - This function is a wrapper around `polars.business_day_count`.
        - The holiday list is determined per-row based on the `start` date.

    Examples:
        >>> from pyield import bday
        >>> bday.count("15-12-2023", "01-01-2024")
        10

        Total business days in January and February since the start of the year
        >>> bday.count(start="01-01-2024", end=["01-02-2024", "01-03-2024"])
        shape: (2,)
        Series: 'bday_count' [i64]
        [
            22
            41
        ]

        The remaining business days from January/February until the end of the year
        >>> bday.count(["01-01-2024", "01-02-2024"], "01-01-2025")
        shape: (2,)
        Series: 'bday_count' [i64]
        [
            253
            231
        ]

        The total business days in January and February of 2024
        >>> bday.count(["01-01-2024", "01-02-2024"], ["01-02-2024", "01-03-2024"])
        shape: (2,)
        Series: 'bday_count' [i64]
        [
            22
            19
        ]

        Null values are propagated
        >>> bday.count(None, "01-01-2024")  # None start

        >>> bday.count("01-01-2024", None)  # None end

        >>> bday.count("01-01-2024", ["01-02-2024", None])  # None in end array
        shape: (2,)
        Series: 'bday_count' [i64]
        [
            22
            null
        ]

        >>> start_dates = ["01-01-2024", "01-02-2024", "01-03-2024"]
        >>> bday.count(start_dates, "01-01-2025")
        shape: (3,)
        Series: 'bday_count' [i64]
        [
            253
            231
            212
        ]
    """
    # Coloca as séries em um DataFrame para trabalhar com expressões em colunas
    df = pl.DataFrame(
        data={"start": cv.convert_dates(start), "end": cv.convert_dates(end)},
        schema={"start": pl.Date, "end": pl.Date},
        nan_to_null=True,
    )

    count_expr = (
        pl.when(pl.col("start") < TRANSITION_DATE)
        .then(
            pl.business_day_count(
                start=pl.col("start"), end=pl.col("end"), holidays=OLD_HOLIDAYS_ARRAY
            ),
        )
        .otherwise(
            pl.business_day_count(
                start=pl.col("start"), end=pl.col("end"), holidays=NEW_HOLIDAYS_ARRAY
            )
        )
        .cast(pl.Int64)
        .alias("bday_count")
    )

    s = df.select(count_expr)["bday_count"]

    if not tp.has_array_like_args(start, end):
        return s.first()

    return s

generate(start=None, end=None, inclusive='both', holiday_option='new')

Generates a Series of business days between a start and end date, considering the list of Brazilian holidays. It supports customization of holiday lists and inclusion options for start and end dates. It wraps pandas.bdate_range.

Parameters:

Name Type Description Default
start DateLike | None

The start date for generating the dates. If None, the current date is used. Defaults to None.

 None
end DateLike | None

The end date for generating business days. If None, the current date is used. Defaults to None.

 None
inclusive Literal['both', 'neither', 'left', 'right']

Determines which of the start and end dates are included in the result. Valid options are 'both', 'neither', 'left', 'right'. Defaults to 'both'.

 'both'
holiday_option Literal['old', 'new', 'infer']

Specifies the list of holidays to consider. Defaults to "new". - 'old': Uses the holiday list effective before the transition date of 2023-12-26. - 'new': Uses the holiday list effective on and after the transition date of 2023-12-26. - 'infer': Automatically selects the holiday list ('old' or 'new') based on the start date relative to the transition date (2023-12-26). If start is before the transition, 'old' is used; otherwise, 'new' is used.

 'new'

Returns:

Type Description
Series

pl.Series: A Series representing a range of business days between the specified start and end dates, considering the specified holidays.

Examples:

>>> from pyield import bday
>>> bday.generate(start="22-12-2023", end="02-01-2024")
shape: (6,)
Series: 'bday' [date]
[
    2023-12-22
    2023-12-26
    2023-12-27
    2023-12-28
    2023-12-29
    2024-01-02
]
Note

For detailed information on parameters and error handling, refer to pandas.bdate_range documentation: https://pandas.pydata.org/docs/reference/api/pandas.bdate_range.html.

Source code in pyield/bday/core.py 
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
def generate(
    start: DateLike | None = None,
    end: DateLike | None = None,
    inclusive: Literal["both", "neither", "left", "right"] = "both",
    holiday_option: Literal["old", "new", "infer"] = "new",
) -> pl.Series:
    """
    Generates a Series of business days between a `start` and `end` date, considering
    the list of Brazilian holidays. It supports customization of holiday lists and
    inclusion options for start and end dates. It wraps `pandas.bdate_range`.

    Args:
        start (DateLike | None, optional): The start date for generating the dates.
             If None, the current date is used. Defaults to None.
        end (DateLike | None, optional): The end date for generating business days.
            If None, the current date is used. Defaults to None.
        inclusive (Literal["both", "neither", "left", "right"], optional):
            Determines which of the start and end dates are included in the result.
            Valid options are 'both', 'neither', 'left', 'right'. Defaults to 'both'.
        holiday_option (Literal["old", "new", "infer"], optional):
            Specifies the list of holidays to consider. Defaults to "new".
            - **'old'**: Uses the holiday list effective *before* the transition date
            of 2023-12-26.
            - **'new'**: Uses the holiday list effective *on and after* the transition
            date of 2023-12-26.
            - **'infer'**: Automatically selects the holiday list ('old' or 'new') based
            on the `start` date relative to the transition date (2023-12-26). If `start`
            is before the transition, 'old' is used; otherwise, 'new' is used.

    Returns:
        pl.Series: A Series representing a range of business days between the specified
            start and end dates, considering the specified holidays.

    Examples:
        >>> from pyield import bday
        >>> bday.generate(start="22-12-2023", end="02-01-2024")
        shape: (6,)
        Series: 'bday' [date]
        [
            2023-12-22
            2023-12-26
            2023-12-27
            2023-12-28
            2023-12-29
            2024-01-02
        ]

    Note:
        For detailed information on parameters and error handling, refer to
        `pandas.bdate_range` documentation:
        https://pandas.pydata.org/docs/reference/api/pandas.bdate_range.html.
    """
    conv_start = cv.convert_dates(start)
    today = clock.today()
    if not conv_start:
        conv_start = today

    conv_end = cv.convert_dates(end)
    if not conv_end:
        conv_end = today

    applicable_holidays = br_holidays.get_holiday_series(
        dates=conv_start, holiday_option=holiday_option
    ).to_list()

    # Get the result as a DatetimeIndex (dti)
    result_dti = pd.bdate_range(
        start=conv_start,
        end=conv_end,
        freq="C",
        inclusive=inclusive,
        holidays=applicable_holidays,
    )
    s_pd = pd.Series(result_dti.values, name="bday").astype("date32[pyarrow]")
    return pl.from_pandas(s_pd)

is_business_day(dates)

Determine whether date(s) are Brazilian business days with per-element holiday regime selection.

PER-ROW HOLIDAY REGIME: For EACH input date the appropriate holiday list ("old" vs. "new") is selected by comparing to the transition date 2023-12-26 (TRANSITION_DATE). Dates strictly before the transition use the old list; dates on or after it use the new list. This mirrors the behavior of count and offset which apply regime logic element-wise.

ORDER & SHAPE PRESERVATION: The output preserves the original element order. No sorting, deduplication, reshaping or alignment is performed; the i-th result corresponds to the i-th provided date after broadcasting (if any broadcasting occurred from a scalar input elsewhere in the call chain).

NULL PROPAGATION: A null scalar argument short-circuits to None. Null values inside array-like inputs produce nulls at the corresponding output positions.

RETURN TYPE: If the (non-null) input resolves to a single element a Python bool is returned. If that lone element is null, None is returned. Otherwise a polars.Series of booleans named 'is_bday' is produced.

WEEKENDS: Saturdays and Sundays are never business days regardless of the holiday regime.

Parameters:

Name Type Description Default
dates None | DateLike | ArrayLike

Single date or collection (list/tuple/ndarray/Polars/Pandas Series). May include nulls which propagate. Null scalar input returns None.

 required

Returns:

Type Description
None | bool | Series

bool | pl.Series | None: True if business day, False otherwise
None | bool | Series

for scalar input; None for null scalar input; or a Polars boolean
None | bool | Series

Series (name: 'is_bday') for array inputs.

Examples:

>>> from pyield import bday
>>> bday.is_business_day("25-12-2023")  # Christmas (old calendar)
False
>>> bday.is_business_day("20-11-2024")  # National Zumbi Day (new holiday)
False
>>> bday.is_business_day(["22-12-2023", "26-12-2023"])  # Mixed periods
shape: (2,)
Series: 'is_bday' [bool]
[
    true
    true
]
Notes
  • Transition date defined in TRANSITION_DATE.
  • Mirrors per-row logic used in count and offset.
  • Weekends always evaluate to False.
  • Null elements propagate.
Source code in pyield/bday/core.py 
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
def is_business_day(dates: None | DateLike | ArrayLike) -> None | bool | pl.Series:
    """Determine whether date(s) are Brazilian business days with per-element
    holiday regime selection.

    PER-ROW HOLIDAY REGIME: For EACH input date the appropriate holiday list
    ("old" vs. "new") is selected by comparing to the transition date
    ``2023-12-26`` (``TRANSITION_DATE``). Dates strictly before the transition
    use the old list; dates on or after it use the new list. This mirrors the
    behavior of ``count`` and ``offset`` which apply regime logic element-wise.

    ORDER & SHAPE PRESERVATION: The output preserves the original element order.
    No sorting, deduplication, reshaping or alignment is performed; the i-th
    result corresponds to the i-th provided date after broadcasting (if any
    broadcasting occurred from a scalar input elsewhere in the call chain).

    NULL PROPAGATION: A null scalar argument short-circuits to ``None``. Null
    values inside array-like inputs produce nulls at the corresponding output
    positions.

    RETURN TYPE: If the (non-null) input resolves to a single element a Python
    ``bool`` is returned. If that lone element is null, ``None`` is returned.
    Otherwise a ``polars.Series`` of booleans named ``'is_bday'`` is produced.

    WEEKENDS: Saturdays and Sundays are never business days regardless of the
    holiday regime.

    Args:
        dates: Single date or collection (list/tuple/ndarray/Polars/Pandas
            Series). May include nulls which propagate. Null scalar input
            returns ``None``.

    Returns:
        bool | pl.Series | None: ``True`` if business day, ``False`` otherwise
        for scalar input; ``None`` for null scalar input; or a Polars boolean
        Series (name: ``'is_bday'``) for array inputs.

    Examples:
        >>> from pyield import bday
        >>> bday.is_business_day("25-12-2023")  # Christmas (old calendar)
        False
        >>> bday.is_business_day("20-11-2024")  # National Zumbi Day (new holiday)
        False
        >>> bday.is_business_day(["22-12-2023", "26-12-2023"])  # Mixed periods
        shape: (2,)
        Series: 'is_bday' [bool]
        [
            true
            true
        ]

    Notes:
        - Transition date defined in ``TRANSITION_DATE``.
        - Mirrors per-row logic used in ``count`` and ``offset``.
        - Weekends always evaluate to ``False``.
        - Null elements propagate.
    """
    # Build DataFrame to allow conditional expression selecting the right holiday list
    df = pl.DataFrame(
        {"dates": cv.convert_dates(dates)},
        schema={"dates": pl.Date},
        nan_to_null=True,
    )

    is_bday_expr = (
        pl.when(pl.col("dates") < TRANSITION_DATE)
        .then(
            pl.col("dates").dt.is_business_day(holidays=OLD_HOLIDAYS_ARRAY),
        )
        .otherwise(
            pl.col("dates").dt.is_business_day(holidays=NEW_HOLIDAYS_ARRAY),
        )
        .alias("is_bday")
    )

    s = df.select(is_bday_expr)["is_bday"]

    if not tp.has_array_like_args(dates):
        return s.first()

    return s

last_business_day()

Returns the last business day in Brazil. If the current date is a business day, it returns the current date. If it is a weekend or holiday, it returns the last business day before the current date.

Returns:

Type Description
date

dt.date: The last business day in Brazil.
Notes
  • The determination of the last business day considers the correct Brazilian holiday list (before or after the 2023-12-26 transition) applicable to the current date.
Source code in pyield/bday/core.py 
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
def last_business_day() -> dt.date:
    """
    Returns the last business day in Brazil. If the current date is a business day, it
    returns the current date. If it is a weekend or holiday, it returns the last
    business day before the current date.

    Returns:
        dt.date: The last business day in Brazil.

    Notes:
        - The determination of the last business day considers the correct Brazilian
        holiday list (before or after the 2023-12-26 transition) applicable to
        the current date.

    """
    # Get the current date in Brazil without timezone information
    bz_today = clock.today()
    result = offset(bz_today, 0, roll="backward")
    assert isinstance(result, dt.date), (
        "Assumption violated: offset did not return a date for the current date."
    )
    return result

offset(dates, offset, roll='forward')

Offset date(s) by a number of business days with per-row Brazilian holiday regime selection. The operation is performed in two steps per element: 1) ROLL: If the original date falls on a weekend or holiday, move it according to roll ("forward" -> next business day; "backward" -> previous). 2) ADD: Apply the signed business-day offset (positive forward, negative backward, zero = stay on the rolled date).

ORDER PRESERVATION (critical): Output ordering strictly matches the element-wise pairing after broadcasting between dates and offset. No sorting, deduplication or shape changes occur. The i-th result corresponds to the i-th (date, offset) pair, enabling safe assignment back into the originating DataFrame.

Holiday regime: For EACH date the appropriate holiday list (old vs. new) is chosen based on the transition date 2023-12-26 (TRANSITION_DATE). Dates before the transition use the old list; dates on/after use the new list.

Roll semantics: roll only acts when the original date is not already a business day under its regime. After rolling, the subsequent business-day addition is applied from that rolled anchor. An offset of 0 therefore returns either the original date (if already a business day) or the rolled business day.

Null propagation: If any scalar argument is null, the function short-circuits to None. Nulls inside array inputs propagate to their corresponding output positions.

Broadcasting: dates and offset may be scalars or array-like. Standard Polars broadcasting rules apply when constructing the per-row pairs.

Return type: If both inputs are non-null scalars a datetime.date is returned. Otherwise a polars.Series of dates named 'adjusted_date' is produced. Null scalar inputs yield None.

Parameters:

Name Type Description Default
dates DateLike | ArrayLike | None

Single date or collection of dates to be rolled (if needed) and then offset. Each date independently selects the holiday regime.

 required
offset int | ArrayLike | None

Signed count of business days to apply after rolling. Positive moves forward, negative backward, zero keeps the rolled anchor.

 required
roll Literal['forward', 'backward']

Direction to roll a non-business starting date ("forward" or "backward"). Defaults to "forward".

 'forward'

Returns:

Type Description
date | Series | None

dt.date | pl.Series | None: A Python date for scalar inputs, a Polars
date | Series | None

Series of dates for any array input, or None if a null scalar argument
date | Series | None

was provided.
Notes
  • Wrapper around polars.Expr.dt.add_business_days applied conditionally.
  • Holiday regime is decided per element by comparing to TRANSITION_DATE.
  • Weekends are always treated as non-business days.

Examples:

>>> from pyield import bday

Offset Saturday before Christmas to the next b. day (Tuesday after Christmas)

>>> bday.offset("23-12-2023", 0)
datetime.date(2023, 12, 26)

Offset Friday before Christmas (no offset because it's a business day)

>>> bday.offset("22-12-2023", 0)
datetime.date(2023, 12, 22)

Offset to the previous business day if not a bday (offset=0 and roll="backward")

No offset because it's a business day

>>> bday.offset("22-12-2023", 0, roll="backward")
datetime.date(2023, 12, 22)

Offset to the first business day before "23-12-2023"

>>> bday.offset("23-12-2023", 0, roll="backward")
datetime.date(2023, 12, 22)

Jump to the next business day (1 offset and roll="forward")

Offset Friday to the next business day (Friday is jumped -> Monday)

>>> bday.offset("27-09-2024", 1)
datetime.date(2024, 9, 30)

Offset Saturday to the next business day (Monday is jumped -> Tuesday)

>>> bday.offset("28-09-2024", 1)
datetime.date(2024, 10, 1)

Jump to the previous business day (-1 offset and roll="backward")

Offset Friday to the previous business day (Friday is jumped -> Thursday)

>>> bday.offset("27-09-2024", -1, roll="backward")
datetime.date(2024, 9, 26)

Offset Saturday to the previous business day (Friday is jumped -> Thursday)

>>> bday.offset("28-09-2024", -1, roll="backward")
datetime.date(2024, 9, 26)

List of dates and offsets

>>> bday.offset(["19-09-2024", "20-09-2024"], 1)
shape: (2,)
Series: 'adjusted_date' [date]
[
    2024-09-20
    2024-09-23
]

>>> bday.offset("19-09-2024", [1, 2])  # a list of offsets
shape: (2,)
Series: 'adjusted_date' [date]
[
    2024-09-20
    2024-09-23
]

Scalar nulls propagate to None

>>> print(bday.offset(None, 1))
None

Scalar null propagates inside arrays

>>> bday.offset(None, [1, 2])
shape: (2,)
Series: 'adjusted_date' [date]
[
    null
    null
]

Nulls inside arrays are preserved

>>> bday.offset(["19-09-2024", None], 1)
shape: (2,)
Series: 'adjusted_date' [date]
[
    2024-09-20
    null
]

>>> dates = ["19-09-2024", "20-09-2024", "21-09-2024"]
>>> bday.offset(dates, 1)
shape: (3,)
Series: 'adjusted_date' [date]
[
    2024-09-20
    2024-09-23
    2024-09-24
]
Note

This function uses polars.Expr.dt.add_business_days under the hood. For detailed information, refer to the Polars documentation.

Source code in pyield/bday/core.py 
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
def offset(
    dates: DateLike | ArrayLike | None,
    offset: int | ArrayLike | None,
    roll: Literal["forward", "backward"] = "forward",
) -> dt.date | pl.Series | None:
    """
    Offset date(s) by a number of business days with per-row Brazilian holiday
    regime selection. The operation is performed in two steps per element:
    1) ROLL: If the original date falls on a weekend or holiday, move it according
       to ``roll`` ("forward" -> next business day; "backward" -> previous).
    2) ADD: Apply the signed business-day ``offset`` (positive forward, negative
       backward, zero = stay on the rolled date).

    ORDER PRESERVATION (critical): Output ordering strictly matches the element-wise
    pairing after broadcasting between ``dates`` and ``offset``. No sorting,
    deduplication or shape changes occur. The i-th result corresponds to the i-th
    (date, offset) pair, enabling safe assignment back into the originating DataFrame.

    Holiday regime: For EACH date the appropriate holiday list (old vs. new) is
    chosen based on the transition date ``2023-12-26`` (``TRANSITION_DATE``). Dates
    before the transition use the *old* list; dates on/after use the *new* list.

    Roll semantics: ``roll`` only acts when the original date is not already a
    business day under its regime. After rolling, the subsequent business-day
    addition is applied from that rolled anchor. An ``offset`` of 0 therefore
    returns either the original date (if already a business day) or the rolled
    business day.

    Null propagation: If any scalar argument is null, the function short-circuits
    to ``None``. Nulls inside array inputs propagate to their corresponding output
    positions.

    Broadcasting: ``dates`` and ``offset`` may be scalars or array-like. Standard
    Polars broadcasting rules apply when constructing the per-row pairs.

    Return type: If both inputs are non-null scalars a ``datetime.date`` is returned.
    Otherwise a ``polars.Series`` of dates named ``'adjusted_date'`` is produced.
    Null scalar inputs yield ``None``.

    Args:
        dates: Single date or collection of dates to be rolled (if needed) and then
            offset. Each date independently selects the holiday regime.
        offset: Signed count of business days to apply after rolling. Positive moves
            forward, negative backward, zero keeps the rolled anchor.
        roll: Direction to roll a non-business starting date ("forward" or
            "backward"). Defaults to "forward".

    Returns:
        dt.date | pl.Series | None: A Python ``date`` for scalar inputs, a Polars
        Series of dates for any array input, or ``None`` if a null scalar argument
        was provided.

    Notes:
        - Wrapper around ``polars.Expr.dt.add_business_days`` applied conditionally.
        - Holiday regime is decided per element by comparing to ``TRANSITION_DATE``.
        - Weekends are always treated as non-business days.

    Examples:
        >>> from pyield import bday

        Offset Saturday before Christmas to the next b. day (Tuesday after Christmas)
        >>> bday.offset("23-12-2023", 0)
        datetime.date(2023, 12, 26)

        Offset Friday before Christmas (no offset because it's a business day)
        >>> bday.offset("22-12-2023", 0)
        datetime.date(2023, 12, 22)

        Offset to the previous business day if not a bday (offset=0 and roll="backward")

        No offset because it's a business day
        >>> bday.offset("22-12-2023", 0, roll="backward")
        datetime.date(2023, 12, 22)

        Offset to the first business day before "23-12-2023"
        >>> bday.offset("23-12-2023", 0, roll="backward")
        datetime.date(2023, 12, 22)

        Jump to the next business day (1 offset and roll="forward")

        Offset Friday to the next business day (Friday is jumped -> Monday)
        >>> bday.offset("27-09-2024", 1)
        datetime.date(2024, 9, 30)

        Offset Saturday to the next business day (Monday is jumped -> Tuesday)
        >>> bday.offset("28-09-2024", 1)
        datetime.date(2024, 10, 1)

        Jump to the previous business day (-1 offset and roll="backward")

        Offset Friday to the previous business day (Friday is jumped -> Thursday)
        >>> bday.offset("27-09-2024", -1, roll="backward")
        datetime.date(2024, 9, 26)

        Offset Saturday to the previous business day (Friday is jumped -> Thursday)
        >>> bday.offset("28-09-2024", -1, roll="backward")
        datetime.date(2024, 9, 26)

        # List of dates and offsets
        >>> bday.offset(["19-09-2024", "20-09-2024"], 1)
        shape: (2,)
        Series: 'adjusted_date' [date]
        [
            2024-09-20
            2024-09-23
        ]

        >>> bday.offset("19-09-2024", [1, 2])  # a list of offsets
        shape: (2,)
        Series: 'adjusted_date' [date]
        [
            2024-09-20
            2024-09-23
        ]

        # Scalar nulls propagate to None
        >>> print(bday.offset(None, 1))
        None

        # Scalar null propagates inside arrays
        >>> bday.offset(None, [1, 2])
        shape: (2,)
        Series: 'adjusted_date' [date]
        [
            null
            null
        ]

        # Nulls inside arrays are preserved
        >>> bday.offset(["19-09-2024", None], 1)
        shape: (2,)
        Series: 'adjusted_date' [date]
        [
            2024-09-20
            null
        ]

        >>> dates = ["19-09-2024", "20-09-2024", "21-09-2024"]
        >>> bday.offset(dates, 1)
        shape: (3,)
        Series: 'adjusted_date' [date]
        [
            2024-09-20
            2024-09-23
            2024-09-24
        ]

    Note:
        This function uses `polars.Expr.dt.add_business_days` under the hood. For
        detailed information, refer to the Polars documentation.
    """
    # Coloca as entradas em um DataFrame para trabalhar com expressões em colunas
    df = pl.DataFrame(
        data={"dates": cv.convert_dates(dates), "offset": offset},
        schema={"dates": pl.Date, "offset": pl.Int64},
        nan_to_null=True,
    )

    # Cria a expressão condicional para aplicar a lista de feriados correta
    offset_expr = (
        pl.when(pl.col("dates") < TRANSITION_DATE)
        .then(
            pl.col("dates").dt.add_business_days(
                n=pl.col("offset"),
                roll=roll,
                holidays=OLD_HOLIDAYS_ARRAY,
            )
        )
        .otherwise(
            pl.col("dates").dt.add_business_days(
                n=pl.col("offset"),
                roll=roll,
                holidays=NEW_HOLIDAYS_ARRAY,
            )
        )
        .alias("adjusted_date")
    )

    # Executa a expressão e obtém a série de resultados
    s = df.select(offset_expr)["adjusted_date"]

    if not tp.has_array_like_args(dates, offset):
        return s.first()

    return s