From c3ad7ce3813b6f4dbe016152cbb8e29b3fc6b8d7 Mon Sep 17 00:00:00 2001
From: Idan Shilon <idan.shilon@quicklizard.com>
Date: Tue, 13 Jun 2023 16:59:01 +0200
Subject: [PATCH 01/15] Fix #1049 - add prior_scale and mode arguments to
 prophet model's add_seasonality

---
 darts/models/forecasting/prophet_model.py | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/darts/models/forecasting/prophet_model.py b/darts/models/forecasting/prophet_model.py
index 60890aacf3..6f26b44547 100644
--- a/darts/models/forecasting/prophet_model.py
+++ b/darts/models/forecasting/prophet_model.py
@@ -186,6 +186,8 @@ def _fit(self, series: TimeSeries, future_covariates: Optional[TimeSeries] = Non
                 name=seasonality_name,
                 period=attributes["seasonal_periods"] * interval_length,
                 fourier_order=attributes["fourier_order"],
+                prior_scale=attributes["prior_scale"],
+                mode=attributes["mode"],
             )
 
         # add covariates

From a6d53f08c5661c760b3be37c9bee9329f9e107f6 Mon Sep 17 00:00:00 2001
From: Idan Shilon <idan.shilon@quicklizard.com>
Date: Wed, 14 Jun 2023 19:24:57 +0200
Subject: [PATCH 02/15] Add option to treat seasonality as conditional

---
 darts/models/forecasting/prophet_model.py     | 37 ++++++++++++++++++-
 .../tests/models/forecasting/test_prophet.py  |  5 ++-
 2 files changed, 39 insertions(+), 3 deletions(-)

diff --git a/darts/models/forecasting/prophet_model.py b/darts/models/forecasting/prophet_model.py
index 6f26b44547..fc1cca76ce 100644
--- a/darts/models/forecasting/prophet_model.py
+++ b/darts/models/forecasting/prophet_model.py
@@ -5,6 +5,7 @@
 
 import logging
 import re
+import types
 from typing import Callable, List, Optional, Sequence, Union
 
 import numpy as np
@@ -182,12 +183,19 @@ def _fit(self, series: TimeSeries, future_covariates: Optional[TimeSeries] = Non
         # add user defined seasonalities (from model creation and/or pre-fit self.add_seasonalities())
         interval_length = self._freq_to_days(series.freq_str)
         for seasonality_name, attributes in self._add_seasonalities.items():
+            condition_name = None
+            if attributes["condition_func"] is not None:
+                fit_df[seasonality_name] = fit_df["ds"].apply(
+                    attributes["condition_func"]
+                )
+                condition_name = seasonality_name
             self.model.add_seasonality(
                 name=seasonality_name,
                 period=attributes["seasonal_periods"] * interval_length,
                 fourier_order=attributes["fourier_order"],
                 prior_scale=attributes["prior_scale"],
                 mode=attributes["mode"],
+                condition_name=condition_name,
             )
 
         # add covariates
@@ -226,6 +234,12 @@ def _predict(
 
         predict_df = self._generate_predict_df(n=n, future_covariates=future_covariates)
 
+        for seasonality_name, attributes in self._add_seasonalities.items():
+            if attributes["condition_func"] is not None:
+                predict_df[seasonality_name] = predict_df["ds"].apply(
+                    attributes["condition_func"]
+                )
+
         if num_samples == 1:
             forecast = self.model.predict(predict_df, vectorized=True)["yhat"].values
         else:
@@ -320,13 +334,20 @@ def add_seasonality(
         fourier_order: int,
         prior_scale: Optional[float] = None,
         mode: Optional[str] = None,
+        condition_func: Optional[types.FunctionType] = None,
     ) -> None:
         """Adds a custom seasonality to the model that repeats after every n `seasonal_periods` timesteps.
         An example for `seasonal_periods`: If you have hourly data (frequency='H') and your seasonal cycle repeats
         after 48 hours -> `seasonal_periods=48`.
 
-        Apart from `seasonal_periods`, this is very similar to how you would call Facebook Prophet's
-        `add_seasonality()` method. For information about the parameters see:
+        Apart from `seasonal_periods` and `conditional_name`, this is very similar to how you would call
+        Facebook Prophet's `add_seasonality()` method. In the case of a conditional seasonality,
+        the condition_name is set to be the seasonality name and the condition is supplied in the form of a function
+        that operates on the time index and returns a boolean mask indicating the timesteps
+        to include in the seasonality component.
+        For more details on conditional seasonalities see:
+        https://facebook.github.io/prophet/docs/seasonality,_holiday_effects,_and_regressors.html#seasonalities-that-depend-on-other-factors
+        For information about the parameters see:
         `The Prophet source code <https://github.com/facebook/prophet/blob/master/python/prophet/forecaster.py>`_.
 
         Parameters
@@ -341,6 +362,10 @@ def add_seasonality(
             optionally, a prior scale for this component
         mode
             optionally, 'additive' or 'multiplicative'
+        condition_func
+            optionally, a function that takes the time index and returns a boolean mask indicating the timesteps
+            to include in the seasonality component.
+            The masking function will be applied as follows: df[name] = df["ds"].apply(condition_func)
         """
         function_call = {
             "name": name,
@@ -348,6 +373,7 @@ def add_seasonality(
             "fourier_order": fourier_order,
             "prior_scale": prior_scale,
             "mode": mode,
+            "condition_func": condition_func,
         }
         self._store_add_seasonality_call(seasonality_call=function_call)
 
@@ -379,12 +405,18 @@ def _store_add_seasonality_call(
             "fourier_order": {"default": None, "dtype": int},
             "prior_scale": {"default": None, "dtype": float},
             "mode": {"default": None, "dtype": str},
+            "condition_func": {"default": None, "dtype": types.FunctionType},
         }
         seasonality_default = {
             kw: seasonality_properties[kw]["default"] for kw in seasonality_properties
         }
 
         mandatory_keywords = ["name", "seasonal_periods", "fourier_order"]
+        if (
+            "condition_func" in seasonality_call.keys()
+            and seasonality_call["condition_func"] is not None
+        ):
+            mandatory_keywords.append("condition_func")
 
         add_seasonality_call = dict(seasonality_default, **seasonality_call)
 
@@ -428,6 +460,7 @@ def _store_add_seasonality_call(
             f'of type {[seasonality_properties[kw]["dtype"] for kw in invalid_types]}.',
             logger,
         )
+
         self._add_seasonalities[seasonality_name] = add_seasonality_call
 
     @staticmethod
diff --git a/darts/tests/models/forecasting/test_prophet.py b/darts/tests/models/forecasting/test_prophet.py
index fc363654a9..5913d06462 100644
--- a/darts/tests/models/forecasting/test_prophet.py
+++ b/darts/tests/models/forecasting/test_prophet.py
@@ -25,7 +25,10 @@ def test_add_seasonality_calls(self):
             "seasonal_periods": 24,
             "fourier_order": 1,
         }
-        kwargs_all = dict(kwargs_mandatory, **{"prior_scale": 1.0, "mode": "additive"})
+        kwargs_all = dict(
+            kwargs_mandatory,
+            **{"prior_scale": 1.0, "mode": "additive", "condition_func": lambda x: x}
+        )
         model1 = Prophet(add_seasonalities=kwargs_all)
         model2 = Prophet()
         model2.add_seasonality(**kwargs_all)

From 5713aba4fabd8c259e74cd412e39db9736546bb4 Mon Sep 17 00:00:00 2001
From: Idan Shilon <idan.shilon@quicklizard.com>
Date: Tue, 4 Jul 2023 14:09:37 +0200
Subject: [PATCH 03/15] Add seasonality conditions with a condition_name and
 future_covariates

---
 darts/models/forecasting/prophet_model.py | 86 ++++++++++++++---------
 1 file changed, 52 insertions(+), 34 deletions(-)

diff --git a/darts/models/forecasting/prophet_model.py b/darts/models/forecasting/prophet_model.py
index fc1cca76ce..9a7bd2d8e1 100644
--- a/darts/models/forecasting/prophet_model.py
+++ b/darts/models/forecasting/prophet_model.py
@@ -5,7 +5,6 @@
 
 import logging
 import re
-import types
 from typing import Callable, List, Optional, Sequence, Union
 
 import numpy as np
@@ -182,13 +181,27 @@ def _fit(self, series: TimeSeries, future_covariates: Optional[TimeSeries] = Non
 
         # add user defined seasonalities (from model creation and/or pre-fit self.add_seasonalities())
         interval_length = self._freq_to_days(series.freq_str)
+        conditional_seasonality_covariates = []
+        if future_covariates is not None:
+            future_covariates_columns = future_covariates.columns
+        else:
+            future_covariates_columns = []
+
         for seasonality_name, attributes in self._add_seasonalities.items():
-            condition_name = None
-            if attributes["condition_func"] is not None:
-                fit_df[seasonality_name] = fit_df["ds"].apply(
-                    attributes["condition_func"]
-                )
-                condition_name = seasonality_name
+            condition_name = attributes["condition_name"]
+            if condition_name is not None:
+                if condition_name in future_covariates_columns:
+                    conditional_seasonality_covariates.append(
+                        attributes["condition_name"]
+                    )
+                else:
+                    raise_if(
+                        True,
+                        f"Condition name '{attributes['condition_name']}' is required by the custom "
+                        f"seasonality '{seasonality_name}', but either it is not found in future_covariates "
+                        f"or future_covariates is None.",
+                    )
+
             self.model.add_seasonality(
                 name=seasonality_name,
                 period=attributes["seasonal_periods"] * interval_length,
@@ -198,7 +211,7 @@ def _fit(self, series: TimeSeries, future_covariates: Optional[TimeSeries] = Non
                 condition_name=condition_name,
             )
 
-        # add covariates
+        # add covariates as additional regressors
         if future_covariates is not None:
             fit_df = fit_df.merge(
                 future_covariates.pd_dataframe(),
@@ -206,8 +219,9 @@ def _fit(self, series: TimeSeries, future_covariates: Optional[TimeSeries] = Non
                 right_index=True,
                 how="left",
             )
-            for covariate in future_covariates.columns:
-                self.model.add_regressor(covariate)
+            for covariate in future_covariates_columns:
+                if covariate not in conditional_seasonality_covariates:
+                    self.model.add_regressor(covariate)
 
         # add built-in country holidays
         if self.country_holidays is not None:
@@ -230,16 +244,26 @@ def _predict(
         verbose: bool = False,
     ) -> TimeSeries:
 
+        for seasonality_name, attributes in self._add_seasonalities.items():
+            if attributes["condition_name"] is not None:
+                raise_if(
+                    future_covariates is None,
+                    f"Condition name '{attributes['condition_name']}' is required by "
+                    f"the custom seasonality '{seasonality_name}', but future_covariates is None. In addition, "
+                    f"the model should be re-trained with future_covariates.",
+                    logger,
+                )
+                raise_if(
+                    attributes["condition_name"] not in future_covariates.columns,
+                    f"Condition name '{attributes['condition_name']}' is required by "
+                    f"the custom seasonality '{seasonality_name}', but it is not found in future_covariates.",
+                    logger,
+                )
+
         super()._predict(n, future_covariates, num_samples)
 
         predict_df = self._generate_predict_df(n=n, future_covariates=future_covariates)
 
-        for seasonality_name, attributes in self._add_seasonalities.items():
-            if attributes["condition_func"] is not None:
-                predict_df[seasonality_name] = predict_df["ds"].apply(
-                    attributes["condition_func"]
-                )
-
         if num_samples == 1:
             forecast = self.model.predict(predict_df, vectorized=True)["yhat"].values
         else:
@@ -334,21 +358,20 @@ def add_seasonality(
         fourier_order: int,
         prior_scale: Optional[float] = None,
         mode: Optional[str] = None,
-        condition_func: Optional[types.FunctionType] = None,
+        condition_name: Optional[str] = None,
     ) -> None:
         """Adds a custom seasonality to the model that repeats after every n `seasonal_periods` timesteps.
         An example for `seasonal_periods`: If you have hourly data (frequency='H') and your seasonal cycle repeats
         after 48 hours -> `seasonal_periods=48`.
 
-        Apart from `seasonal_periods` and `conditional_name`, this is very similar to how you would call
-        Facebook Prophet's `add_seasonality()` method. In the case of a conditional seasonality,
-        the condition_name is set to be the seasonality name and the condition is supplied in the form of a function
-        that operates on the time index and returns a boolean mask indicating the timesteps
-        to include in the seasonality component.
+        Apart from `seasonal_periods`, this is very similar to how you would call
+        Facebook Prophet's `add_seasonality()` method. To add conditional seasonalities,
+        a condition_name has to be given here and a future_covariates TimeSeries
+        with a component (column) named condition_name is expected to be passed to the 'fit()' and 'predict()' methods.
         For more details on conditional seasonalities see:
         https://facebook.github.io/prophet/docs/seasonality,_holiday_effects,_and_regressors.html#seasonalities-that-depend-on-other-factors
         For information about the parameters see:
-        `The Prophet source code <https://github.com/facebook/prophet/blob/master/python/prophet/forecaster.py>`_.
+        `The Prophet source code <https://github.com/facebook/prophet/blob/master/python/prophet/forecaster.py>`.
 
         Parameters
         ----------
@@ -362,10 +385,10 @@ def add_seasonality(
             optionally, a prior scale for this component
         mode
             optionally, 'additive' or 'multiplicative'
-        condition_func
-            optionally, a function that takes the time index and returns a boolean mask indicating the timesteps
-            to include in the seasonality component.
-            The masking function will be applied as follows: df[name] = df["ds"].apply(condition_func)
+        condition_name
+            optionally, the name of the condition on which the seasonality depends. If not None, a future_covariates
+            TimeSeries with a component (column) named condition_name is expected to be passed to the 'fit()' and
+            'predict()' methods.
         """
         function_call = {
             "name": name,
@@ -373,7 +396,7 @@ def add_seasonality(
             "fourier_order": fourier_order,
             "prior_scale": prior_scale,
             "mode": mode,
-            "condition_func": condition_func,
+            "condition_name": condition_name,
         }
         self._store_add_seasonality_call(seasonality_call=function_call)
 
@@ -405,18 +428,13 @@ def _store_add_seasonality_call(
             "fourier_order": {"default": None, "dtype": int},
             "prior_scale": {"default": None, "dtype": float},
             "mode": {"default": None, "dtype": str},
-            "condition_func": {"default": None, "dtype": types.FunctionType},
+            "condition_name": {"default": None, "dtype": str},
         }
         seasonality_default = {
             kw: seasonality_properties[kw]["default"] for kw in seasonality_properties
         }
 
         mandatory_keywords = ["name", "seasonal_periods", "fourier_order"]
-        if (
-            "condition_func" in seasonality_call.keys()
-            and seasonality_call["condition_func"] is not None
-        ):
-            mandatory_keywords.append("condition_func")
 
         add_seasonality_call = dict(seasonality_default, **seasonality_call)
 

From cf55740efb8cc14457453f771b41d89554b7ae73 Mon Sep 17 00:00:00 2001
From: Idan Shilon <idan.shilon@quicklizard.com>
Date: Tue, 4 Jul 2023 14:12:29 +0200
Subject: [PATCH 04/15] Add test for custom conditional seasonality

---
 .../tests/models/forecasting/test_prophet.py  | 42 ++++++++++++++++++-
 1 file changed, 41 insertions(+), 1 deletion(-)

diff --git a/darts/tests/models/forecasting/test_prophet.py b/darts/tests/models/forecasting/test_prophet.py
index 5913d06462..70814bfbcc 100644
--- a/darts/tests/models/forecasting/test_prophet.py
+++ b/darts/tests/models/forecasting/test_prophet.py
@@ -27,7 +27,11 @@ def test_add_seasonality_calls(self):
         }
         kwargs_all = dict(
             kwargs_mandatory,
-            **{"prior_scale": 1.0, "mode": "additive", "condition_func": lambda x: x}
+            **{
+                "prior_scale": 1.0,
+                "mode": "additive",
+                "condition_name": "custom_condition",
+            }
         )
         model1 = Prophet(add_seasonalities=kwargs_all)
         model2 = Prophet()
@@ -237,3 +241,39 @@ def helper_test_prophet_model(self, period, freq, compare_all_models=False):
         for pred in compare_preds:
             for val_i, pred_i in zip(val.univariate_values(), pred.univariate_values()):
                 self.assertAlmostEqual(val_i, pred_i, delta=0.1)
+
+    def test_conditional_seasonality(self):
+        """
+        Test that conditional seasonality is correctly incorporated by the model
+        """
+        df = pd.DataFrame()
+        df["ds"] = pd.date_range(start="2022-01-02", periods=395)
+        df["y"] = [i + 10 * (i % 7 == 0) for i in range(395)]
+        df["is_sunday"] = df["ds"].apply(lambda x: int(x.weekday() == 6))
+
+        ts = TimeSeries.from_dataframe(
+            df[:-30], time_col="ds", value_cols="y", freq="D"
+        )
+        future_covariates = TimeSeries.from_dataframe(
+            df, time_col="ds", value_cols=["is_sunday"], freq="D"
+        )
+        expected_result = TimeSeries.from_dataframe(
+            df[-30:], time_col="ds", value_cols="y", freq="D"
+        )
+
+        model = Prophet(seasonality_mode="additive")
+        model.add_seasonality(
+            name="weekly_sun",
+            seasonal_periods=7,
+            fourier_order=2,
+            condition_name="is_sunday",
+        )
+
+        model.fit(ts, future_covariates=future_covariates)
+
+        forecast = model.predict(30, future_covariates=future_covariates)
+
+        for val_i, pred_i in zip(
+            expected_result.univariate_values(), forecast.univariate_values()
+        ):
+            self.assertAlmostEqual(val_i, pred_i, delta=0.1)

From 3d5fc4a506cf9e160d16c47fd741c11b18aa2b3a Mon Sep 17 00:00:00 2001
From: Idan Shilon <idan.shilon@quicklizard.com>
Date: Tue, 4 Jul 2023 14:13:26 +0200
Subject: [PATCH 05/15] Add entry for pr #1829

---
 CHANGELOG.md | 1 +
 1 file changed, 1 insertion(+)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 03489ff964..544672408e 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -28,6 +28,7 @@ but cannot always guarantee backwards compatibility. Changes that may **break co
 - Fixed a bug when loading the weights of a `TorchForecastingModel` trained with encoders or a Likelihood. [#1744](https://github.com/unit8co/darts/pull/1744) by [Antoine Madrona](https://github.com/madtoinou).
 - Fixed a bug when using selected `target_components` with `ShapExplainer. [#1803](https://github.com/unit8co/darts/pull/#1803) by [Dennis Bader](https://github.com/dennisbader).
 - Fixed `TimeSeries.__getitem__()` for series with a RangeIndex with start != 0 and freq != 1. [#1868](https://github.com/unit8co/darts/pull/#1868) by [Dennis Bader](https://github.com/dennisbader).
+- Fixed an issue with `prophet_model.Prophet.add_seasonality()` to allow proper use of all passed parameters. [#1829](https://github.com/unit8co/darts/pull/#1829) by [Idan Shilon](https://github.com/id5h).
 
 **Removed**
 - Removed support for Python 3.7 [#1864](https://github.com/unit8co/darts/pull/#1864) by [Dennis Bader](https://github.com/dennisbader).

From 801e80a8cfbd37c0a9f30bce88d68a4f3eca0fb6 Mon Sep 17 00:00:00 2001
From: id5h <shilonphys@gmail.com>
Date: Wed, 5 Jul 2023 14:43:08 +0200
Subject: [PATCH 06/15] Update darts/models/forecasting/prophet_model.py

Co-authored-by: Dennis Bader <dennis.bader@gmx.ch>
---
 darts/models/forecasting/prophet_model.py | 14 ++++++++------
 1 file changed, 8 insertions(+), 6 deletions(-)

diff --git a/darts/models/forecasting/prophet_model.py b/darts/models/forecasting/prophet_model.py
index a0be8993c4..0dc21747f9 100644
--- a/darts/models/forecasting/prophet_model.py
+++ b/darts/models/forecasting/prophet_model.py
@@ -368,14 +368,16 @@ def add_seasonality(
         An example for `seasonal_periods`: If you have hourly data (frequency='H') and your seasonal cycle repeats
         after 48 hours -> `seasonal_periods=48`.
 
-        Apart from `seasonal_periods`, this is very similar to how you would call
-        Facebook Prophet's `add_seasonality()` method. To add conditional seasonalities,
-        a condition_name has to be given here and a future_covariates TimeSeries
-        with a component (column) named condition_name is expected to be passed to the 'fit()' and 'predict()' methods.
-        For more details on conditional seasonalities see:
-        https://facebook.github.io/prophet/docs/seasonality,_holiday_effects,_and_regressors.html#seasonalities-that-depend-on-other-factors
+        Apart from `seasonal_periods`, this is very similar to how you would call Facebook Prophet's
+        `add_seasonality()` method.
+
+        To add conditional seasonalities, provide `condition_name` here, and add a boolean (binary) component/column
+        named `condition_name` to the `future_covariates` series passed to `fit()` and `predict()`.
+
         For information about the parameters see:
         `The Prophet source code <https://github.com/facebook/prophet/blob/master/python/prophet/forecaster.py>`.
+        For more details on conditional seasonalities see:
+        https://facebook.github.io/prophet/docs/seasonality,_holiday_effects,_and_regressors.html#seasonalities-that-depend-on-other-factors
 
         Parameters
         ----------

From 5b0321c04927bcc580d570ead2ef0b55f4332571 Mon Sep 17 00:00:00 2001
From: id5h <shilonphys@gmail.com>
Date: Wed, 5 Jul 2023 14:43:43 +0200
Subject: [PATCH 07/15] Update darts/models/forecasting/prophet_model.py

Co-authored-by: Dennis Bader <dennis.bader@gmx.ch>
---
 darts/models/forecasting/prophet_model.py | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/darts/models/forecasting/prophet_model.py b/darts/models/forecasting/prophet_model.py
index 0dc21747f9..e8721edb4f 100644
--- a/darts/models/forecasting/prophet_model.py
+++ b/darts/models/forecasting/prophet_model.py
@@ -392,9 +392,9 @@ def add_seasonality(
         mode
             optionally, 'additive' or 'multiplicative'
         condition_name
-            optionally, the name of the condition on which the seasonality depends. If not None, a future_covariates
-            TimeSeries with a component (column) named condition_name is expected to be passed to the 'fit()' and
-            'predict()' methods.
+            optionally, the name of the condition on which the seasonality depends. If not `None`, expects a 
+            `future_covariates` time series with a component/column named `condition_name` to be passed to `fit()` 
+            and `predict()`.
         """
         function_call = {
             "name": name,

From 50a70d411d98d9d412fea10ebb04e8a745cdfe58 Mon Sep 17 00:00:00 2001
From: Idan Shilon <idan.shilon@quicklizard.com>
Date: Wed, 5 Jul 2023 23:31:30 +0200
Subject: [PATCH 08/15] Validate seasonality considitions through a private
 method when calling fit() and predict()

---
 darts/models/forecasting/prophet_model.py | 114 ++++++++++++++--------
 1 file changed, 75 insertions(+), 39 deletions(-)

diff --git a/darts/models/forecasting/prophet_model.py b/darts/models/forecasting/prophet_model.py
index e8721edb4f..00dd0361a8 100644
--- a/darts/models/forecasting/prophet_model.py
+++ b/darts/models/forecasting/prophet_model.py
@@ -181,34 +181,17 @@ def _fit(self, series: TimeSeries, future_covariates: Optional[TimeSeries] = Non
 
         # add user defined seasonalities (from model creation and/or pre-fit self.add_seasonalities())
         interval_length = self._freq_to_days(series.freq_str)
-        conditional_seasonality_covariates = []
-        if future_covariates is not None:
-            future_covariates_columns = future_covariates.columns
-        else:
-            future_covariates_columns = []
-
+        conditional_seasonality_covariates = self._check_seasonality_conditions(
+            future_covariates=future_covariates
+        )
         for seasonality_name, attributes in self._add_seasonalities.items():
-            condition_name = attributes["condition_name"]
-            if condition_name is not None:
-                if condition_name in future_covariates_columns:
-                    conditional_seasonality_covariates.append(
-                        attributes["condition_name"]
-                    )
-                else:
-                    raise_if(
-                        True,
-                        f"Condition name '{attributes['condition_name']}' is required by the custom "
-                        f"seasonality '{seasonality_name}', but either it is not found in future_covariates "
-                        f"or future_covariates is None.",
-                    )
-
             self.model.add_seasonality(
                 name=seasonality_name,
                 period=attributes["seasonal_periods"] * interval_length,
                 fourier_order=attributes["fourier_order"],
                 prior_scale=attributes["prior_scale"],
                 mode=attributes["mode"],
-                condition_name=condition_name,
+                condition_name=attributes["condition_name"],
             )
 
         # add covariates as additional regressors
@@ -219,7 +202,7 @@ def _fit(self, series: TimeSeries, future_covariates: Optional[TimeSeries] = Non
                 right_index=True,
                 how="left",
             )
-            for covariate in future_covariates_columns:
+            for covariate in future_covariates.columns:
                 if covariate not in conditional_seasonality_covariates:
                     self.model.add_regressor(covariate)
 
@@ -244,21 +227,7 @@ def _predict(
         verbose: bool = False,
     ) -> TimeSeries:
 
-        for seasonality_name, attributes in self._add_seasonalities.items():
-            if attributes["condition_name"] is not None:
-                raise_if(
-                    future_covariates is None,
-                    f"Condition name '{attributes['condition_name']}' is required by "
-                    f"the custom seasonality '{seasonality_name}', but future_covariates is None. In addition, "
-                    f"the model should be re-trained with future_covariates.",
-                    logger,
-                )
-                raise_if(
-                    attributes["condition_name"] not in future_covariates.columns,
-                    f"Condition name '{attributes['condition_name']}' is required by "
-                    f"the custom seasonality '{seasonality_name}', but it is not found in future_covariates.",
-                    logger,
-                )
+        _ = self._check_seasonality_conditions(future_covariates=future_covariates)
 
         super()._predict(n, future_covariates, num_samples)
 
@@ -307,6 +276,73 @@ def _generate_predict_df(
             )
         return predict_df
 
+    def _check_seasonality_conditions(
+        self, future_covariates: Optional[TimeSeries] = None
+    ) -> List[str]:
+        """
+        Checks if the conditions for custom conditional seasonalities are met. Each custom seasonality that has a
+        `condition_name` other than None is checked. If the `condition_name` is not a column in the `future_covariates`
+        or if the values in the column are not all True or False, an error is raised.
+        Returns a list of the `condition_name`s of the conditional seasonalities that have been checked.
+
+        Parameters
+        ----------
+        future_covariates
+            optionally, a TimeSeries containing the future covariates and including the columns that are used as
+            conditions for the conditional seasonalities when necessary
+
+        Raises
+        ------
+        ValueError
+            if a seasonality has a `condition_name` and a column named `condition_name` is missing in
+            the `future_covariates`
+
+            if a seasonality has a `condition_name` and the values in the corresponding column in `future_covariates`
+            are not binary values (True or False, 1 or 0)
+        """
+
+        conditional_seasonality_covariates = []
+        invalid_conditional_seasonalities = []
+        if future_covariates is not None:
+            future_covariates_columns = future_covariates.columns
+        else:
+            future_covariates_columns = []
+
+        for seasonality_name, attributes in self._add_seasonalities.items():
+            condition_name = attributes["condition_name"]
+            if condition_name is not None:
+                if condition_name not in future_covariates_columns:
+                    invalid_conditional_seasonalities.append(
+                        (seasonality_name, condition_name, "column missing")
+                    )
+                    continue
+                if (
+                    not future_covariates[condition_name]
+                    .pd_series()
+                    .isin([True, False])
+                    .all()
+                ):
+                    invalid_conditional_seasonalities.append(
+                        (seasonality_name, condition_name, "invalid values")
+                    )
+                    continue
+                conditional_seasonality_covariates.append(condition_name)
+
+        formatted_issues_str = ", ".join(
+            f"'{name}' (condition_name: '{cond}'; issue: {reason})"
+            for name, cond, reason in invalid_conditional_seasonalities
+        )
+        raise_if(
+            len(invalid_conditional_seasonalities) > 0,
+            f"The following seasonalities have invalid conditions: "
+            f"{formatted_issues_str}. "
+            f"Each conditional seasonality must be accompanied by a binary component/column in the future_covariates "
+            f"with the same name as the condition_name. These components must only contain "
+            f"True or False values (or 1 or 0).",
+            logger,
+        )
+        return conditional_seasonality_covariates
+
     @property
     def supports_multivariate(self) -> bool:
         return False
@@ -392,8 +428,8 @@ def add_seasonality(
         mode
             optionally, 'additive' or 'multiplicative'
         condition_name
-            optionally, the name of the condition on which the seasonality depends. If not `None`, expects a 
-            `future_covariates` time series with a component/column named `condition_name` to be passed to `fit()` 
+            optionally, the name of the condition on which the seasonality depends. If not `None`, expects a
+            `future_covariates` time series with a component/column named `condition_name` to be passed to `fit()`
             and `predict()`.
         """
         function_call = {

From 0fe11d2cf8910e58f1410488aefb18d4ad21a7bc Mon Sep 17 00:00:00 2001
From: Idan Shilon <idan.shilon@quicklizard.com>
Date: Wed, 5 Jul 2023 23:33:01 +0200
Subject: [PATCH 09/15] Reduce predict horizon to 7. Add tests for missing and
 invalid conditions

---
 .../tests/models/forecasting/test_prophet.py  | 24 +++++++++++++++----
 1 file changed, 20 insertions(+), 4 deletions(-)

diff --git a/darts/tests/models/forecasting/test_prophet.py b/darts/tests/models/forecasting/test_prophet.py
index 70814bfbcc..0486665fb2 100644
--- a/darts/tests/models/forecasting/test_prophet.py
+++ b/darts/tests/models/forecasting/test_prophet.py
@@ -2,6 +2,7 @@
 
 import numpy as np
 import pandas as pd
+import pytest
 
 from darts import TimeSeries
 from darts.logging import get_logger
@@ -246,19 +247,21 @@ def test_conditional_seasonality(self):
         """
         Test that conditional seasonality is correctly incorporated by the model
         """
+        duration = 395
+        horizon = 7
         df = pd.DataFrame()
         df["ds"] = pd.date_range(start="2022-01-02", periods=395)
-        df["y"] = [i + 10 * (i % 7 == 0) for i in range(395)]
+        df["y"] = [i + 10 * (i % 7 == 0) for i in range(duration)]
         df["is_sunday"] = df["ds"].apply(lambda x: int(x.weekday() == 6))
 
         ts = TimeSeries.from_dataframe(
-            df[:-30], time_col="ds", value_cols="y", freq="D"
+            df[:-horizon], time_col="ds", value_cols="y", freq="D"
         )
         future_covariates = TimeSeries.from_dataframe(
             df, time_col="ds", value_cols=["is_sunday"], freq="D"
         )
         expected_result = TimeSeries.from_dataframe(
-            df[-30:], time_col="ds", value_cols="y", freq="D"
+            df[-horizon:], time_col="ds", value_cols="y", freq="D"
         )
 
         model = Prophet(seasonality_mode="additive")
@@ -271,9 +274,22 @@ def test_conditional_seasonality(self):
 
         model.fit(ts, future_covariates=future_covariates)
 
-        forecast = model.predict(30, future_covariates=future_covariates)
+        forecast = model.predict(horizon, future_covariates=future_covariates)
 
         for val_i, pred_i in zip(
             expected_result.univariate_values(), forecast.univariate_values()
         ):
             self.assertAlmostEqual(val_i, pred_i, delta=0.1)
+
+        invalid_future_covariates = future_covariates.with_values(
+            np.reshape(np.random.randint(0, 3, duration), (-1, 1, 1)).astype("float")
+        )
+
+        with pytest.raises(ValueError):
+            model.fit(ts, future_covariates=invalid_future_covariates)
+
+        with pytest.raises(ValueError):
+            model.fit(
+                ts,
+                future_covariates=invalid_future_covariates.drop_columns("is_sunday"),
+            )

From a96a15f1a29346683ef774e7ffa241eab4dd9ccf Mon Sep 17 00:00:00 2001
From: Idan Shilon <idan.shilon@quicklizard.com>
Date: Wed, 5 Jul 2023 23:33:37 +0200
Subject: [PATCH 10/15] Move entry to models improvements section

---
 CHANGELOG.md | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index bbc192efac..f389f76b66 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -15,6 +15,8 @@ but cannot always guarantee backwards compatibility. Changes that may **break co
   - Added support for `PathLike` to the `save()` and `load()` functions of all non-deep learning based models. [#1754](https://github.com/unit8co/darts/pull/1754) by [Simon Sudrich](https://github.com/sudrich).
   - Improved efficiency of `historical_forecasts()` and `backtest()` for all models giving significant process time reduction for larger number of predict iterations and series. [#1801](https://github.com/unit8co/darts/pull/1801) by [Dennis Bader](https://github.com/dennisbader).
   - Added model property `ForecastingModel.supports_multivariate` to indicate whether the model supports multivariate forecasting. [#1848](https://github.com/unit8co/darts/pull/1848) by [Felix Divo](https://github.com/felixdivo).
+  - `Prophet` now supports conditional seasonalities, and properly handles all parameters passed to `Prophet.add_seasonality()` and model creation parameter `add_seasonalities` [#1829](https://github.com/unit8co/darts/pull/#1829) by [Idan Shilon](https://github.com/id5h).
+
 - Improvements to `EnsembleModel`:
   - Model creation parameter `forecasting_models` now supports a mix of `LocalForecastingModel` and `GlobalForecastingModel` (single `TimeSeries` training/inference only, due to the local models). [#1745](https://github.com/unit8co/darts/pull/1745) by [Antoine Madrona](https://github.com/madtoinou).
   - Future and past covariates can now be used even if `forecasting_models` have different covariates support. The covariates passed to `fit()`/`predict()` are used only by models that support it. [#1745](https://github.com/unit8co/darts/pull/1745) by [Antoine Madrona](https://github.com/madtoinou).
@@ -29,7 +31,6 @@ but cannot always guarantee backwards compatibility. Changes that may **break co
 - Fixed a bug when loading the weights of a `TorchForecastingModel` trained with encoders or a Likelihood. [#1744](https://github.com/unit8co/darts/pull/1744) by [Antoine Madrona](https://github.com/madtoinou).
 - Fixed a bug when using selected `target_components` with `ShapExplainer. [#1803](https://github.com/unit8co/darts/pull/#1803) by [Dennis Bader](https://github.com/dennisbader).
 - Fixed `TimeSeries.__getitem__()` for series with a RangeIndex with start != 0 and freq != 1. [#1868](https://github.com/unit8co/darts/pull/#1868) by [Dennis Bader](https://github.com/dennisbader).
-- Fixed an issue with `prophet_model.Prophet.add_seasonality()` to allow proper use of all passed parameters. [#1829](https://github.com/unit8co/darts/pull/#1829) by [Idan Shilon](https://github.com/id5h).
 
 **Removed**
 - Removed support for Python 3.7 [#1864](https://github.com/unit8co/darts/pull/#1864) by [Dennis Bader](https://github.com/dennisbader).

From e9418a6c07a35f81341bcb8c9ce8d78f93e64465 Mon Sep 17 00:00:00 2001
From: id5h <shilonphys@gmail.com>
Date: Fri, 7 Jul 2023 21:30:10 +0200
Subject: [PATCH 11/15] Update err msg in _check_seasonality_conditions

Co-authored-by: Dennis Bader <dennis.bader@gmx.ch>
---
 darts/models/forecasting/prophet_model.py | 22 +++++++++++++---------
 1 file changed, 13 insertions(+), 9 deletions(-)

diff --git a/darts/models/forecasting/prophet_model.py b/darts/models/forecasting/prophet_model.py
index 00dd0361a8..a951171634 100644
--- a/darts/models/forecasting/prophet_model.py
+++ b/darts/models/forecasting/prophet_model.py
@@ -332,15 +332,19 @@ def _check_seasonality_conditions(
             f"'{name}' (condition_name: '{cond}'; issue: {reason})"
             for name, cond, reason in invalid_conditional_seasonalities
         )
-        raise_if(
-            len(invalid_conditional_seasonalities) > 0,
-            f"The following seasonalities have invalid conditions: "
-            f"{formatted_issues_str}. "
-            f"Each conditional seasonality must be accompanied by a binary component/column in the future_covariates "
-            f"with the same name as the condition_name. These components must only contain "
-            f"True or False values (or 1 or 0).",
-            logger,
-        )
+        if len(invalid_conditional_seasonalities) > 0:
+            formatted_issues_str = ", ".join(
+                f"'{name}' (condition_name: '{cond}'; issue: {reason})"
+                for name, cond, reason in invalid_conditional_seasonalities
+            )
+            raise_log(
+                ValueError(
+                    f"The following seasonalities have invalid conditions: {formatted_issues_str}. "
+                    f"Each conditional seasonality must be accompanied by a binary component/column in the "
+                    f"`future_covariates` with the same name as the `condition_name`"
+                ),
+                logger
+            )
         return conditional_seasonality_covariates
 
     @property

From 9057345aa1660ee2fa7517dbb16150d2e5b11f85 Mon Sep 17 00:00:00 2001
From: Idan Shilon <idan.shilon@quicklizard.com>
Date: Fri, 7 Jul 2023 21:48:14 +0200
Subject: [PATCH 12/15] Import raise_log. Initialize formatted str when
 necessary.

---
 darts/models/forecasting/prophet_model.py | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

diff --git a/darts/models/forecasting/prophet_model.py b/darts/models/forecasting/prophet_model.py
index a951171634..83b5ef1aef 100644
--- a/darts/models/forecasting/prophet_model.py
+++ b/darts/models/forecasting/prophet_model.py
@@ -11,7 +11,7 @@
 import pandas as pd
 import prophet
 
-from darts.logging import execute_and_suppress_output, get_logger, raise_if
+from darts.logging import execute_and_suppress_output, get_logger, raise_if, raise_log
 from darts.models.forecasting.forecasting_model import (
     FutureCovariatesLocalForecastingModel,
 )
@@ -328,10 +328,6 @@ def _check_seasonality_conditions(
                     continue
                 conditional_seasonality_covariates.append(condition_name)
 
-        formatted_issues_str = ", ".join(
-            f"'{name}' (condition_name: '{cond}'; issue: {reason})"
-            for name, cond, reason in invalid_conditional_seasonalities
-        )
         if len(invalid_conditional_seasonalities) > 0:
             formatted_issues_str = ", ".join(
                 f"'{name}' (condition_name: '{cond}'; issue: {reason})"
@@ -343,7 +339,7 @@ def _check_seasonality_conditions(
                     f"Each conditional seasonality must be accompanied by a binary component/column in the "
                     f"`future_covariates` with the same name as the `condition_name`"
                 ),
-                logger
+                logger,
             )
         return conditional_seasonality_covariates
 

From 157f206d6cb85f9206c00812d3d147dfc2d8f154 Mon Sep 17 00:00:00 2001
From: Idan Shilon <idan.shilon@quicklizard.com>
Date: Fri, 7 Jul 2023 21:55:30 +0200
Subject: [PATCH 13/15] Accept float seasonalities as well. Update test

---
 darts/models/forecasting/prophet_model.py      | 4 ++--
 darts/tests/models/forecasting/test_prophet.py | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/darts/models/forecasting/prophet_model.py b/darts/models/forecasting/prophet_model.py
index 83b5ef1aef..2e0f6bb278 100644
--- a/darts/models/forecasting/prophet_model.py
+++ b/darts/models/forecasting/prophet_model.py
@@ -394,7 +394,7 @@ def predict_raw(
     def add_seasonality(
         self,
         name: str,
-        seasonal_periods: int,
+        seasonal_periods: Union[int, float],
         fourier_order: int,
         prior_scale: Optional[float] = None,
         mode: Optional[str] = None,
@@ -466,7 +466,7 @@ def _store_add_seasonality_call(
 
         seasonality_properties = {
             "name": {"default": None, "dtype": str},
-            "seasonal_periods": {"default": None, "dtype": int},
+            "seasonal_periods": {"default": None, "dtype": Union[int, float]},
             "fourier_order": {"default": None, "dtype": int},
             "prior_scale": {"default": None, "dtype": float},
             "mode": {"default": None, "dtype": str},
diff --git a/darts/tests/models/forecasting/test_prophet.py b/darts/tests/models/forecasting/test_prophet.py
index 0486665fb2..af8bffca41 100644
--- a/darts/tests/models/forecasting/test_prophet.py
+++ b/darts/tests/models/forecasting/test_prophet.py
@@ -23,7 +23,7 @@ def test_add_seasonality_calls(self):
         }
         kwargs_mandatory2 = {
             "name": "custom2",
-            "seasonal_periods": 24,
+            "seasonal_periods": 24.9,
             "fourier_order": 1,
         }
         kwargs_all = dict(

From 7c03f75c81e586532ddfb9ca33f0afc6d76477fb Mon Sep 17 00:00:00 2001
From: Idan Shilon <idan.shilon@quicklizard.com>
Date: Mon, 10 Jul 2023 13:35:44 +0200
Subject: [PATCH 14/15] Fix dtype of seasonal_periods. Update docstrings.

---
 darts/models/forecasting/prophet_model.py | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)

diff --git a/darts/models/forecasting/prophet_model.py b/darts/models/forecasting/prophet_model.py
index 2e0f6bb278..53338e0a70 100644
--- a/darts/models/forecasting/prophet_model.py
+++ b/darts/models/forecasting/prophet_model.py
@@ -53,7 +53,7 @@ def __init__(
 
                 dict({
                 'name': str  # (name of the seasonality component),
-                'seasonal_periods': int  # (nr of steps composing a season),
+                'seasonal_periods': Union[int, float]  # (nr of steps composing a season),
                 'fourier_order': int  # (number of Fourier components to use),
                 'prior_scale': Optional[float]  # (a prior scale for this component),
                 'mode': Optional[str]  # ('additive' or 'multiplicative')
@@ -61,7 +61,9 @@ def __init__(
             ..
 
             An example for `seasonal_periods`: If you have hourly data (frequency='H') and your seasonal cycle repeats
-            after 48 hours then set `seasonal_periods=48`.
+            after 48 hours then set `seasonal_periods=48`. Notice that this value will be multiplied by the inferred
+            number of days for the TimeSeries frequency (1 / 24 in this example) to be consistent with the
+            `add_seasonality()` method of Facebook Prophet, where the `period` parameter is specified in days.
 
             Apart from `seasonal_periods`, this is very similar to how you would call Facebook Prophet's
             `add_seasonality()` method.
@@ -420,7 +422,9 @@ def add_seasonality(
         name
             name of the seasonality component
         seasonal_periods
-            number of timesteps after which the seasonal cycle repeats
+            number of timesteps after which the seasonal cycle repeats. This value will be multiplied by the inferred
+            number of days for the TimeSeries frequency (e.g. 365.25 for a yearly frequency) to be consistent with the
+            `add_seasonality()` method of Facebook Prophet.
         fourier_order
             number of Fourier components to use
         prior_scale
@@ -466,7 +470,7 @@ def _store_add_seasonality_call(
 
         seasonality_properties = {
             "name": {"default": None, "dtype": str},
-            "seasonal_periods": {"default": None, "dtype": Union[int, float]},
+            "seasonal_periods": {"default": None, "dtype": (int, float)},
             "fourier_order": {"default": None, "dtype": int},
             "prior_scale": {"default": None, "dtype": float},
             "mode": {"default": None, "dtype": str},

From 61abeea594870f48caeacfc5e0567a29dac46c36 Mon Sep 17 00:00:00 2001
From: dennisbader <dennis.bader@gmx.ch>
Date: Mon, 17 Jul 2023 15:08:35 +0200
Subject: [PATCH 15/15] update docstring

---
 darts/models/forecasting/prophet_model.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/darts/models/forecasting/prophet_model.py b/darts/models/forecasting/prophet_model.py
index 53338e0a70..9ded2cba78 100644
--- a/darts/models/forecasting/prophet_model.py
+++ b/darts/models/forecasting/prophet_model.py
@@ -424,7 +424,8 @@ def add_seasonality(
         seasonal_periods
             number of timesteps after which the seasonal cycle repeats. This value will be multiplied by the inferred
             number of days for the TimeSeries frequency (e.g. 365.25 for a yearly frequency) to be consistent with the
-            `add_seasonality()` method of Facebook Prophet.
+            `add_seasonality()` method of Facebook Prophet. The inferred number of days can be obtained with
+            `model._freq_to_days(series.freq)`, where `model` is the `Prophet` model and `series` is the target series.
         fourier_order
             number of Fourier components to use
         prior_scale