upgini 1.1.262a3250.post3__py3-none-any.whl → 1.1.274a4__py3-none-any.whl

upgini/features_enricher.py

@@ -1,4 +1,5 @@
 import dataclasses
+import datetime
 import gc
 import hashlib
 import itertools
@@ -70,6 +71,7 @@ from upgini.utils.datetime_utils import (
     DateTimeSearchKeyConverter,
     is_blocked_time_series,
     is_time_series,
+    validate_dates_distribution,
 )
 from upgini.utils.deduplicate_utils import (
     clean_full_duplicates,
@@ -93,7 +95,7 @@ try:
 except Exception:
     from upgini.utils.fallback_progress_bar import CustomFallbackProgressBar as ProgressBar
 
-from upgini.utils.target_utils import define_task
+from upgini.utils.target_utils import calculate_psi, define_task
 from upgini.utils.warning_counter import WarningCounter
 from upgini.version_validator import validate_version
 
@@ -145,6 +147,7 @@ class FeaturesEnricher(TransformerMixin):
     """
 
     TARGET_NAME = "target"
+    CURRENT_DATE = "current_date"
     RANDOM_STATE = 42
     CALCULATE_METRICS_THRESHOLD = 50_000_000
     CALCULATE_METRICS_MIN_THRESHOLD = 500
@@ -206,6 +209,7 @@ class FeaturesEnricher(TransformerMixin):
         client_ip: Optional[str] = None,
         client_visitorid: Optional[str] = None,
         custom_bundle_config: Optional[str] = None,
+        add_date_if_missing: bool = True,
         **kwargs,
     ):
         self.bundle = get_custom_bundle(custom_bundle_config)
@@ -220,7 +224,7 @@ class FeaturesEnricher(TransformerMixin):
         if logs_enabled:
             self.logger = LoggerFactory().get_logger(endpoint, self._api_key, client_ip, client_visitorid)
         else:
-            self.logger = logging.getLogger()
+            self.logger = logging.getLogger("muted_logger")
             self.logger.setLevel("FATAL")
 
         if len(kwargs) > 0:
@@ -316,6 +320,7 @@ class FeaturesEnricher(TransformerMixin):
         self.raise_validation_error = raise_validation_error
         self.exclude_columns = exclude_columns
         self.baseline_score_column = baseline_score_column
+        self.add_date_if_missing = add_date_if_missing
 
     def _get_api_key(self):
         return self._api_key
@@ -423,7 +428,7 @@ class FeaturesEnricher(TransformerMixin):
                 self.X = X
                 self.y = y
                 self.eval_set = self._check_eval_set(eval_set, X, self.bundle)
-                self.dump_input(trace_id, X, y, eval_set)
+                self.dump_input(trace_id, X, y, self.eval_set)
                 self.__inner_fit(
                     trace_id,
                     X,
@@ -562,7 +567,7 @@ class FeaturesEnricher(TransformerMixin):
                 self.X = X
                 self.y = y
                 self.eval_set = self._check_eval_set(eval_set, X, self.bundle)
-                self.dump_input(trace_id, X, y, eval_set)
+                self.dump_input(trace_id, X, y, self.eval_set)
 
                 if _num_samples(drop_duplicates(X)) > Dataset.MAX_ROWS:
                     raise ValidationError(self.bundle.get("dataset_too_many_rows_registered").format(Dataset.MAX_ROWS))
@@ -822,12 +827,16 @@ class FeaturesEnricher(TransformerMixin):
                 print(msg)
 
             self.__validate_search_keys(self.search_keys, self.search_id)
+            effective_X = X if X is not None else self.X
+            effective_y = y if y is not None else self.y
+            effective_eval_set = eval_set if eval_set is not None else self.eval_set
+            effective_eval_set = self._check_eval_set(effective_eval_set, effective_X, self.bundle)
 
             try:
                 self.__log_debug_information(
-                    X if X is not None else self.X,
-                    y if y is not None else self.y,
-                    eval_set if eval_set is not None else self.eval_set,
+                    effective_X,
+                    effective_y,
+                    effective_eval_set,
                     exclude_features_sources=exclude_features_sources,
                     cv=cv if cv is not None else self.cv,
                     importance_threshold=importance_threshold,
@@ -841,17 +850,14 @@ class FeaturesEnricher(TransformerMixin):
                     self._search_task is None
                     or self._search_task.provider_metadata_v2 is None
                     or len(self._search_task.provider_metadata_v2) == 0
-                    or (self.X is None and X is None)
-                    or (self.y is None and y is None)
+                    or effective_X is None
+                    or effective_y is None
                 ):
                     raise ValidationError(self.bundle.get("metrics_unfitted_enricher"))
 
                 if X is not None and y is None:
                     raise ValidationError("X passed without y")
 
-                effective_X = X if X is not None else self.X
-                effective_eval_set = eval_set if eval_set is not None else self.eval_set
-
                 validate_scoring_argument(scoring)
 
                 self._validate_baseline_score(effective_X, effective_eval_set)
@@ -871,8 +877,7 @@ class FeaturesEnricher(TransformerMixin):
                 ):
                     cat_features = estimator.get_param("cat_features")
                     if len(cat_features) > 0 and isinstance(cat_features[0], int):
-                        effectiveX = X or self.X
-                        cat_features = [effectiveX.columns[i] for i in cat_features]
+                        cat_features = [effective_X.columns[i] for i in cat_features]
                         for cat_feature in cat_features:
                             if cat_feature in self.search_keys:
                                 if self.search_keys[cat_feature] in [SearchKey.COUNTRY, SearchKey.POSTAL_CODE]:
@@ -882,9 +887,9 @@ class FeaturesEnricher(TransformerMixin):
 
                 prepared_data = self._prepare_data_for_metrics(
                     trace_id=trace_id,
-                    X=X,
-                    y=y,
-                    eval_set=eval_set,
+                    X=effective_X,
+                    y=effective_y,
+                    eval_set=effective_eval_set,
                     exclude_features_sources=exclude_features_sources,
                     importance_threshold=importance_threshold,
                     max_features=max_features,
@@ -994,8 +999,6 @@ class FeaturesEnricher(TransformerMixin):
                         enriched_metric = None
                         uplift = None
 
-                    effective_X = X if X is not None else self.X
-                    effective_y = y if y is not None else self.y
                     train_metrics = {
                         self.bundle.get("quality_metrics_segment_header"): self.bundle.get(
                             "quality_metrics_train_segment"
@@ -1256,6 +1259,7 @@ class FeaturesEnricher(TransformerMixin):
             ).get_cv_and_groups(X)
         else:
             from sklearn import __version__ as sklearn_version
+
             try:
                 from sklearn.model_selection._split import GroupsConsumerMixin
 
@@ -1684,6 +1688,9 @@ class FeaturesEnricher(TransformerMixin):
             df = validated_X.copy()
 
             df[TARGET] = validated_y
+
+            df = clean_full_duplicates(df, logger=self.logger, silent=True, bundle=self.bundle)
+
             num_samples = _num_samples(df)
             if num_samples > Dataset.FIT_SAMPLE_THRESHOLD:
                 self.logger.info(f"Downsampling from {num_samples} to {Dataset.FIT_SAMPLE_ROWS}")
@@ -1801,10 +1808,11 @@ class FeaturesEnricher(TransformerMixin):
         else:
             features_section = ""
 
-        api_example = f"""curl 'https://inference-upgini.azurewebsites.net/api/http_inference_trigger' \\
+        search_id = self._search_task.search_task_id
+        api_example = f"""curl 'https://search.upgini.com/online/api/http_inference_trigger?search_id={search_id}' \\
     -H 'Authorization: {self.api_key}' \\
     -H 'Content-Type: application/json' \\
-    -d '{{"search_id": "{self._search_task.search_task_id}", "search_keys": {keys}{features_section}}}'"""
+    -d '{{"search_keys": {keys}{features_section}}}'"""
         return api_example
 
     def _get_copy_of_runtime_parameters(self) -> RuntimeParameters:
@@ -1899,6 +1907,8 @@ class FeaturesEnricher(TransformerMixin):
                 generated_features.extend(converter.generated_features)
             else:
                 self.logger.info("Input dataset hasn't date column")
+                if self.add_date_if_missing:
+                    df = self._add_current_date_as_key(df, search_keys, self.logger, self.bundle)
             email_column = self._get_email_column(search_keys)
             hem_column = self._get_hem_column(search_keys)
             email_converted_to_hem = False
@@ -1918,6 +1928,7 @@ class FeaturesEnricher(TransformerMixin):
 
             meaning_types = {col: key.value for col, key in search_keys.items()}
             non_keys_columns = [column for column in df.columns if column not in search_keys.keys()]
+
             if email_converted_to_hem:
                 non_keys_columns.append(email_column)
 
@@ -1939,6 +1950,7 @@ class FeaturesEnricher(TransformerMixin):
             if add_fit_system_record_id:
                 df = self.__add_fit_system_record_id(df, dict(), search_keys)
                 df = df.rename(columns={SYSTEM_RECORD_ID: SORT_ID})
+                non_keys_columns.append(SORT_ID)
 
             columns_for_system_record_id = sorted(list(search_keys.keys()) + (original_features_for_transform or []))
 
@@ -2215,14 +2227,13 @@ class FeaturesEnricher(TransformerMixin):
         self.fit_search_keys = self.search_keys.copy()
         self.fit_search_keys = self.__prepare_search_keys(validated_X, self.fit_search_keys, is_demo_dataset)
 
-        has_date = self._get_date_column(self.fit_search_keys) is not None
+        validate_dates_distribution(validated_X, self.fit_search_keys, self.logger, self.bundle, self.warning_counter)
+
+        maybe_date_column = self._get_date_column(self.fit_search_keys)
+        has_date = maybe_date_column is not None
         model_task_type = self.model_task_type or define_task(validated_y, has_date, self.logger)
         self._validate_binary_observations(validated_y, model_task_type)
 
-        df = self.__handle_index_search_keys(df, self.fit_search_keys)
-
-        df = self.__correct_target(df)
-
         self.runtime_parameters = get_runtime_params_custom_loss(
             self.loss, model_task_type, self.runtime_parameters, self.logger
         )
@@ -2234,6 +2245,13 @@ class FeaturesEnricher(TransformerMixin):
                 eval_df[EVAL_SET_INDEX] = idx + 1
                 df = pd.concat([df, eval_df])
 
+        df = self.__correct_target(df)
+
+        df = self.__handle_index_search_keys(df, self.fit_search_keys)
+
+        if is_numeric_dtype(df[self.TARGET_NAME]) and has_date:
+            self._validate_PSI(df.sort_values(by=maybe_date_column))
+
         if DEFAULT_INDEX in df.columns:
             msg = self.bundle.get("unsupported_index_column")
             self.logger.info(msg)
@@ -2260,6 +2278,8 @@ class FeaturesEnricher(TransformerMixin):
             self.fit_generated_features.extend(converter.generated_features)
         else:
             self.logger.info("Input dataset hasn't date column")
+            if self.add_date_if_missing:
+                df = self._add_current_date_as_key(df, self.fit_search_keys, self.logger, self.bundle)
         email_column = self._get_email_column(self.fit_search_keys)
         hem_column = self._get_hem_column(self.fit_search_keys)
         email_converted_to_hem = False
@@ -2808,6 +2828,7 @@ class FeaturesEnricher(TransformerMixin):
 
             maybe_date_col = self._get_date_column(self.search_keys)
             if X is not None and maybe_date_col is not None and maybe_date_col in X.columns:
+                # TODO cast date column to single dtype
                 min_date = X[maybe_date_col].min()
                 max_date = X[maybe_date_col].max()
                 self.logger.info(f"Dates interval is ({min_date}, {max_date})")
@@ -2839,6 +2860,25 @@ class FeaturesEnricher(TransformerMixin):
             if t in [SearchKey.DATE, SearchKey.DATETIME]:
                 return col
 
+    @staticmethod
+    def _add_current_date_as_key(
+        df: pd.DataFrame, search_keys: Dict[str, SearchKey], logger: logging.Logger, bundle: ResourceBundle
+    ) -> pd.DataFrame:
+        if (
+            set(search_keys.values()) == {SearchKey.PHONE}
+            or set(search_keys.values()) == {SearchKey.EMAIL}
+            or set(search_keys.values()) == {SearchKey.HEM}
+            or set(search_keys.values()) == {SearchKey.COUNTRY, SearchKey.POSTAL_CODE}
+        ):
+            msg = bundle.get("current_date_added")
+            print(msg)
+            logger.warning(msg)
+            df[FeaturesEnricher.CURRENT_DATE] = datetime.date.today()
+            search_keys[FeaturesEnricher.CURRENT_DATE] = SearchKey.DATE
+            converter = DateTimeSearchKeyConverter(FeaturesEnricher.CURRENT_DATE, None, logger, bundle)
+            df = converter.convert(df)
+        return df
+
     @staticmethod
     def _get_group_columns(df: pd.DataFrame, search_keys: Dict[str, SearchKey]) -> List[str]:
         return [
@@ -2877,26 +2917,33 @@ class FeaturesEnricher(TransformerMixin):
 
         # order by date and idempotent order by other keys
         if self.cv not in [CVType.time_series, CVType.blocked_time_series]:
+            sort_exclude_columns = [original_order_name, ORIGINAL_INDEX, EVAL_SET_INDEX, TARGET, "__target"]
             if DateTimeSearchKeyConverter.DATETIME_COL in df.columns:
                 date_column = DateTimeSearchKeyConverter.DATETIME_COL
+                sort_exclude_columns.append(self._get_date_column(search_keys))
             else:
                 date_column = self._get_date_column(search_keys)
             sort_columns = [date_column] if date_column is not None else []
 
-            other_search_keys = sorted(
+            other_columns = sorted(
                 [
-                    sk
-                    for sk, key_type in search_keys.items()
-                    if key_type not in [SearchKey.DATE, SearchKey.DATETIME]
-                    and sk in df.columns
-                    and df[sk].nunique() > 1  # don't use constant keys for hash
+                    c
+                    for c in df.columns
+                    if c not in sort_columns and c not in sort_exclude_columns and df[c].nunique() > 1
                 ]
+                # [
+                #     sk
+                #     for sk, key_type in search_keys.items()
+                #     if key_type not in [SearchKey.DATE, SearchKey.DATETIME]
+                #     and sk in df.columns
+                #     and df[sk].nunique() > 1  # don't use constant keys for hash
+                # ]
             )
 
             search_keys_hash = "search_keys_hash"
-            if len(other_search_keys) > 0:
+            if len(other_columns) > 0:
                 sort_columns.append(search_keys_hash)
-                df[search_keys_hash] = pd.util.hash_pandas_object(df[sorted(other_search_keys)], index=False)
+                df[search_keys_hash] = pd.util.hash_pandas_object(df[other_columns], index=False)
 
             df = df.sort_values(by=sort_columns)
 
@@ -3185,22 +3232,21 @@ class FeaturesEnricher(TransformerMixin):
                 return None
             features_meta = self._search_task.get_all_features_metadata_v2()
 
-            def get_feature_by_display_index(idx, op):
+            def get_feature_by_name(name: str):
                 for m in features_meta:
-                    if m.name.endswith(f"_{op}_{idx}"):
+                    if m.name == name:
                         return m
 
             descriptions = []
             for m in autofe_meta:
                 autofe_feature = Feature.from_formula(m.formula)
+                autofe_feature.set_display_index(m.display_index)
                 if autofe_feature.op.is_vector:
                     continue
 
                 description = dict()
 
-                feature_meta = get_feature_by_display_index(
-                    m.display_index, autofe_feature.op.alias or autofe_feature.op.name
-                )
+                feature_meta = get_feature_by_name(autofe_feature.get_display_name(shorten=True))
                 if feature_meta is None:
                     self.logger.warning(f"Feature meta for display index {m.display_index} not found")
                     continue
@@ -3547,6 +3593,34 @@ class FeaturesEnricher(TransformerMixin):
             self.logger.warning(msg)
             print(msg)
 
+    def _validate_PSI(self, df: pd.DataFrame):
+        if EVAL_SET_INDEX in df.columns:
+            train = df.query(f"{EVAL_SET_INDEX} == 0")
+            eval1 = df.query(f"{EVAL_SET_INDEX} == 1")
+        else:
+            train = df
+            eval1 = None
+
+        # 1. Check train PSI
+        half_train = round(len(train) / 2)
+        part1 = train[:half_train]
+        part2 = train[half_train:]
+        train_psi = calculate_psi(part1[self.TARGET_NAME], part2[self.TARGET_NAME])
+        if train_psi > 0.2:
+            self.warning_counter.increment()
+            msg = self.bundle.get("train_unstable_target").format(train_psi)
+            print(msg)
+            self.logger.warning(msg)
+
+        # 2. Check train-test PSI
+        if eval1 is not None:
+            train_test_psi = calculate_psi(train[self.TARGET_NAME], eval1[self.TARGET_NAME])
+            if train_test_psi > 0.2:
+                self.warning_counter.increment()
+                msg = self.bundle.get("eval_unstable_target").format(train_test_psi)
+                print(msg)
+                self.logger.warning(msg)
+
     def _dump_python_libs(self):
         try:
             from pip._internal.operations.freeze import freeze
@@ -3613,7 +3687,7 @@ class FeaturesEnricher(TransformerMixin):
                     if y is not None:
                         with open(f"{tmp_dir}/y.pickle", "wb") as y_file:
                             pickle.dump(sample(y, xy_sample_index), y_file)
-                        if eval_set is not None:
+                        if eval_set:
                             eval_xy_sample_index = rnd.randint(0, _num_samples(eval_set[0][0]), size=1000)
                             with open(f"{tmp_dir}/eval_x.pickle", "wb") as eval_x_file:
                                 pickle.dump(sample(eval_set[0][0], eval_xy_sample_index), eval_x_file)

upgini/fingerprint.js

@@ -0,0 +1,8 @@
+/**
+ * FingerprintJS v3.4.2 - Copyright (c) FingerprintJS, Inc, 2023 (https://fingerprint.com)
+ * Licensed under the MIT (http://www.opensource.org/licenses/mit-license.php) license.
+ *
+ * This software contains code from open-source projects:
+ * MurmurHash3 by Karan Lyons (https://github.com/karanlyons/murmurHash3.js)
+ */
+var e=function(){return e=Object.assign||function(e){for(var n,t=1,r=arguments.length;t<r;t++)for(var o in n=arguments[t])Object.prototype.hasOwnProperty.call(n,o)&&(e[o]=n[o]);return e},e.apply(this,arguments)};function n(e,n,t,r){return new(t||(t=Promise))((function(o,a){function i(e){try{u(r.next(e))}catch(n){a(n)}}function c(e){try{u(r.throw(e))}catch(n){a(n)}}function u(e){var n;e.done?o(e.value):(n=e.value,n instanceof t?n:new t((function(e){e(n)}))).then(i,c)}u((r=r.apply(e,n||[])).next())}))}function t(e,n){var t,r,o,a,i={label:0,sent:function(){if(1&o[0])throw o[1];return o[1]},trys:[],ops:[]};return a={next:c(0),throw:c(1),return:c(2)},"function"==typeof Symbol&&(a[Symbol.iterator]=function(){return this}),a;function c(c){return function(u){return function(c){if(t)throw new TypeError("Generator is already executing.");for(;a&&(a=0,c[0]&&(i=0)),i;)try{if(t=1,r&&(o=2&c[0]?r.return:c[0]?r.throw||((o=r.return)&&o.call(r),0):r.next)&&!(o=o.call(r,c[1])).done)return o;switch(r=0,o&&(c=[2&c[0],o.value]),c[0]){case 0:case 1:o=c;break;case 4:return i.label++,{value:c[1],done:!1};case 5:i.label++,r=c[1],c=[0];continue;case 7:c=i.ops.pop(),i.trys.pop();continue;default:if(!(o=i.trys,(o=o.length>0&&o[o.length-1])||6!==c[0]&&2!==c[0])){i=0;continue}if(3===c[0]&&(!o||c[1]>o[0]&&c[1]<o[3])){i.label=c[1];break}if(6===c[0]&&i.label<o[1]){i.label=o[1],o=c;break}if(o&&i.label<o[2]){i.label=o[2],i.ops.push(c);break}o[2]&&i.ops.pop(),i.trys.pop();continue}c=n.call(e,i)}catch(u){c=[6,u],r=0}finally{t=o=0}if(5&c[0])throw c[1];return{value:c[0]?c[1]:void 0,done:!0}}([c,u])}}}function r(e,n,t){if(t||2===arguments.length)for(var r,o=0,a=n.length;o<a;o++)!r&&o in n||(r||(r=Array.prototype.slice.call(n,0,o)),r[o]=n[o]);return e.concat(r||Array.prototype.slice.call(n))}function o(e,n){return new Promise((function(t){return setTimeout(t,e,n)}))}function a(e){return!!e&&"function"==typeof e.then}function i(e,n){try{var t=e();a(t)?t.then((function(e){return n(!0,e)}),(function(e){return n(!1,e)})):n(!0,t)}catch(r){n(!1,r)}}function c(e,r,a){return void 0===a&&(a=16),n(this,void 0,void 0,(function(){var n,i,c,u;return t(this,(function(t){switch(t.label){case 0:n=Array(e.length),i=Date.now(),c=0,t.label=1;case 1:return c<e.length?(n[c]=r(e[c],c),(u=Date.now())>=i+a?(i=u,[4,o(0)]):[3,3]):[3,4];case 2:t.sent(),t.label=3;case 3:return++c,[3,1];case 4:return[2,n]}}))}))}function u(e){e.then(void 0,(function(){}))}function l(e,n){e=[e[0]>>>16,65535&e[0],e[1]>>>16,65535&e[1]],n=[n[0]>>>16,65535&n[0],n[1]>>>16,65535&n[1]];var t=[0,0,0,0];return t[3]+=e[3]+n[3],t[2]+=t[3]>>>16,t[3]&=65535,t[2]+=e[2]+n[2],t[1]+=t[2]>>>16,t[2]&=65535,t[1]+=e[1]+n[1],t[0]+=t[1]>>>16,t[1]&=65535,t[0]+=e[0]+n[0],t[0]&=65535,[t[0]<<16|t[1],t[2]<<16|t[3]]}function s(e,n){e=[e[0]>>>16,65535&e[0],e[1]>>>16,65535&e[1]],n=[n[0]>>>16,65535&n[0],n[1]>>>16,65535&n[1]];var t=[0,0,0,0];return t[3]+=e[3]*n[3],t[2]+=t[3]>>>16,t[3]&=65535,t[2]+=e[2]*n[3],t[1]+=t[2]>>>16,t[2]&=65535,t[2]+=e[3]*n[2],t[1]+=t[2]>>>16,t[2]&=65535,t[1]+=e[1]*n[3],t[0]+=t[1]>>>16,t[1]&=65535,t[1]+=e[2]*n[2],t[0]+=t[1]>>>16,t[1]&=65535,t[1]+=e[3]*n[1],t[0]+=t[1]>>>16,t[1]&=65535,t[0]+=e[0]*n[3]+e[1]*n[2]+e[2]*n[1]+e[3]*n[0],t[0]&=65535,[t[0]<<16|t[1],t[2]<<16|t[3]]}function d(e,n){return 32===(n%=64)?[e[1],e[0]]:n<32?[e[0]<<n|e[1]>>>32-n,e[1]<<n|e[0]>>>32-n]:(n-=32,[e[1]<<n|e[0]>>>32-n,e[0]<<n|e[1]>>>32-n])}function m(e,n){return 0===(n%=64)?e:n<32?[e[0]<<n|e[1]>>>32-n,e[1]<<n]:[e[1]<<n-32,0]}function f(e,n){return[e[0]^n[0],e[1]^n[1]]}function v(e){return e=f(e,[0,e[0]>>>1]),e=f(e=s(e,[4283543511,3981806797]),[0,e[0]>>>1]),e=f(e=s(e,[3301882366,444984403]),[0,e[0]>>>1])}function h(e,n){n=n||0;var t,r=(e=e||"").length%16,o=e.length-r,a=[0,n],i=[0,n],c=[0,0],u=[0,0],h=[2277735313,289559509],p=[1291169091,658871167];for(t=0;t<o;t+=16)c=[255&e.charCodeAt(t+4)|(255&e.charCodeAt(t+5))<<8|(255&e.charCodeAt(t+6))<<16|(255&e.charCodeAt(t+7))<<24,255&e.charCodeAt(t)|(255&e.charCodeAt(t+1))<<8|(255&e.charCodeAt(t+2))<<16|(255&e.charCodeAt(t+3))<<24],u=[255&e.charCodeAt(t+12)|(255&e.charCodeAt(t+13))<<8|(255&e.charCodeAt(t+14))<<16|(255&e.charCodeAt(t+15))<<24,255&e.charCodeAt(t+8)|(255&e.charCodeAt(t+9))<<8|(255&e.charCodeAt(t+10))<<16|(255&e.charCodeAt(t+11))<<24],c=d(c=s(c,h),31),a=l(a=d(a=f(a,c=s(c,p)),27),i),a=l(s(a,[0,5]),[0,1390208809]),u=d(u=s(u,p),33),i=l(i=d(i=f(i,u=s(u,h)),31),a),i=l(s(i,[0,5]),[0,944331445]);switch(c=[0,0],u=[0,0],r){case 15:u=f(u,m([0,e.charCodeAt(t+14)],48));case 14:u=f(u,m([0,e.charCodeAt(t+13)],40));case 13:u=f(u,m([0,e.charCodeAt(t+12)],32));case 12:u=f(u,m([0,e.charCodeAt(t+11)],24));case 11:u=f(u,m([0,e.charCodeAt(t+10)],16));case 10:u=f(u,m([0,e.charCodeAt(t+9)],8));case 9:u=s(u=f(u,[0,e.charCodeAt(t+8)]),p),i=f(i,u=s(u=d(u,33),h));case 8:c=f(c,m([0,e.charCodeAt(t+7)],56));case 7:c=f(c,m([0,e.charCodeAt(t+6)],48));case 6:c=f(c,m([0,e.charCodeAt(t+5)],40));case 5:c=f(c,m([0,e.charCodeAt(t+4)],32));case 4:c=f(c,m([0,e.charCodeAt(t+3)],24));case 3:c=f(c,m([0,e.charCodeAt(t+2)],16));case 2:c=f(c,m([0,e.charCodeAt(t+1)],8));case 1:c=s(c=f(c,[0,e.charCodeAt(t)]),h),a=f(a,c=s(c=d(c,31),p))}return a=l(a=f(a,[0,e.length]),i=f(i,[0,e.length])),i=l(i,a),a=l(a=v(a),i=v(i)),i=l(i,a),("00000000"+(a[0]>>>0).toString(16)).slice(-8)+("00000000"+(a[1]>>>0).toString(16)).slice(-8)+("00000000"+(i[0]>>>0).toString(16)).slice(-8)+("00000000"+(i[1]>>>0).toString(16)).slice(-8)}function p(e){return parseInt(e)}function b(e){return parseFloat(e)}function y(e,n){return"number"==typeof e&&isNaN(e)?n:e}function g(e){return e.reduce((function(e,n){return e+(n?1:0)}),0)}function w(e,n){if(void 0===n&&(n=1),Math.abs(n)>=1)return Math.round(e/n)*n;var t=1/n;return Math.round(e*t)/t}function L(e){return e&&"object"==typeof e&&"message"in e?e:{message:e}}function k(e){return"function"!=typeof e}function V(e,r,o){var a=Object.keys(e).filter((function(e){return!function(e,n){for(var t=0,r=e.length;t<r;++t)if(e[t]===n)return!0;return!1}(o,e)})),l=c(a,(function(n){return function(e,n){var t=new Promise((function(t){var r=Date.now();i(e.bind(null,n),(function(){for(var e=[],n=0;n<arguments.length;n++)e[n]=arguments[n];var o=Date.now()-r;if(!e[0])return t((function(){return{error:L(e[1]),duration:o}}));var a=e[1];if(k(a))return t((function(){return{value:a,duration:o}}));t((function(){return new Promise((function(e){var n=Date.now();i(a,(function(){for(var t=[],r=0;r<arguments.length;r++)t[r]=arguments[r];var a=o+Date.now()-n;if(!t[0])return e({error:L(t[1]),duration:a});e({value:t[1],duration:a})}))}))}))}))}));return u(t),function(){return t.then((function(e){return e()}))}}(e[n],r)}));return u(l),function(){return n(this,void 0,void 0,(function(){var e,n,r,o;return t(this,(function(t){switch(t.label){case 0:return[4,l];case 1:return[4,c(t.sent(),(function(e){var n=e();return u(n),n}))];case 2:return e=t.sent(),[4,Promise.all(e)];case 3:for(n=t.sent(),r={},o=0;o<a.length;++o)r[a[o]]=n[o];return[2,r]}}))}))}}function Z(e,n){var t=function(e){return k(e)?n(e):function(){var t=e();return a(t)?t.then(n):n(t)}};return function(n){var r=e(n);return a(r)?r.then(t):t(r)}}function W(){var e=window,n=navigator;return g(["MSCSSMatrix"in e,"msSetImmediate"in e,"msIndexedDB"in e,"msMaxTouchPoints"in n,"msPointerEnabled"in n])>=4}function C(){var e=window,n=navigator;return g(["msWriteProfilerMark"in e,"MSStream"in e,"msLaunchUri"in n,"msSaveBlob"in n])>=3&&!W()}function S(){var e=window,n=navigator;return g(["webkitPersistentStorage"in n,"webkitTemporaryStorage"in n,0===n.vendor.indexOf("Google"),"webkitResolveLocalFileSystemURL"in e,"BatteryManager"in e,"webkitMediaStream"in e,"webkitSpeechGrammar"in e])>=5}function x(){var e=window,n=navigator;return g(["ApplePayError"in e,"CSSPrimitiveValue"in e,"Counter"in e,0===n.vendor.indexOf("Apple"),"getStorageUpdates"in n,"WebKitMediaKeys"in e])>=4}function F(){var e=window;return g(["safari"in e,!("DeviceMotionEvent"in e),!("ongestureend"in e),!("standalone"in navigator)])>=3}function Y(){var e,n,t=window;return g(["buildID"in navigator,"MozAppearance"in(null!==(n=null===(e=document.documentElement)||void 0===e?void 0:e.style)&&void 0!==n?n:{}),"onmozfullscreenchange"in t,"mozInnerScreenX"in t,"CSSMozDocumentRule"in t,"CanvasCaptureMediaStream"in t])>=4}function M(){var e=document;return e.fullscreenElement||e.msFullscreenElement||e.mozFullScreenElement||e.webkitFullscreenElement||null}function G(){var e=S(),n=Y();if(!e&&!n)return!1;var t=window;return g(["onorientationchange"in t,"orientation"in t,e&&!("SharedWorker"in t),n&&/android/i.test(navigator.appVersion)])>=2}function R(e){var n=new Error(e);return n.name=e,n}function X(e,r,a){var i,c,u;return void 0===a&&(a=50),n(this,void 0,void 0,(function(){var n,l;return t(this,(function(t){switch(t.label){case 0:n=document,t.label=1;case 1:return n.body?[3,3]:[4,o(a)];case 2:return t.sent(),[3,1];case 3:l=n.createElement("iframe"),t.label=4;case 4:return t.trys.push([4,,10,11]),[4,new Promise((function(e,t){var o=!1,a=function(){o=!0,e()};l.onload=a,l.onerror=function(e){o=!0,t(e)};var i=l.style;i.setProperty("display","block","important"),i.position="absolute",i.top="0",i.left="0",i.visibility="hidden",r&&"srcdoc"in l?l.srcdoc=r:l.src="about:blank",n.body.appendChild(l);var c=function(){var e,n;o||("complete"===(null===(n=null===(e=l.contentWindow)||void 0===e?void 0:e.document)||void 0===n?void 0:n.readyState)?a():setTimeout(c,10))};c()}))];case 5:t.sent(),t.label=6;case 6:return(null===(c=null===(i=l.contentWindow)||void 0===i?void 0:i.document)||void 0===c?void 0:c.body)?[3,8]:[4,o(a)];case 7:return t.sent(),[3,6];case 8:return[4,e(l,l.contentWindow)];case 9:return[2,t.sent()];case 10:return null===(u=l.parentNode)||void 0===u||u.removeChild(l),[7];case 11:return[2]}}))}))}function A(e){for(var n=function(e){for(var n,t,r="Unexpected syntax '".concat(e,"'"),o=/^\s*([a-z-]*)(.*)$/i.exec(e),a=o[1]||void 0,i={},c=/([.:#][\w-]+|\[.+?\])/gi,u=function(e,n){i[e]=i[e]||[],i[e].push(n)};;){var l=c.exec(o[2]);if(!l)break;var s=l[0];switch(s[0]){case".":u("class",s.slice(1));break;case"#":u("id",s.slice(1));break;case"[":var d=/^\[([\w-]+)([~|^$*]?=("(.*?)"|([\w-]+)))?(\s+[is])?\]$/.exec(s);if(!d)throw new Error(r);u(d[1],null!==(t=null!==(n=d[4])&&void 0!==n?n:d[5])&&void 0!==t?t:"");break;default:throw new Error(r)}}return[a,i]}(e),t=n[0],r=n[1],o=document.createElement(null!=t?t:"div"),a=0,i=Object.keys(r);a<i.length;a++){var c=i[a],u=r[c].join(" ");"style"===c?j(o.style,u):o.setAttribute(c,u)}return o}function j(e,n){for(var t=0,r=n.split(";");t<r.length;t++){var o=r[t],a=/^\s*([\w-]+)\s*:\s*(.+?)(\s*!([\w-]+))?\s*$/.exec(o);if(a){var i=a[1],c=a[2],u=a[4];e.setProperty(i,c,u||"")}}}var I=["monospace","sans-serif","serif"],J=["sans-serif-thin","ARNO PRO","Agency FB","Arabic Typesetting","Arial Unicode MS","AvantGarde Bk BT","BankGothic Md BT","Batang","Bitstream Vera Sans Mono","Calibri","Century","Century Gothic","Clarendon","EUROSTILE","Franklin Gothic","Futura Bk BT","Futura Md BT","GOTHAM","Gill Sans","HELV","Haettenschweiler","Helvetica Neue","Humanst521 BT","Leelawadee","Letter Gothic","Levenim MT","Lucida Bright","Lucida Sans","Menlo","MS Mincho","MS Outlook","MS Reference Specialty","MS UI Gothic","MT Extra","MYRIAD PRO","Marlett","Meiryo UI","Microsoft Uighur","Minion Pro","Monotype Corsiva","PMingLiU","Pristina","SCRIPTINA","Segoe UI Light","Serifa","SimHei","Small Fonts","Staccato222 BT","TRAJAN PRO","Univers CE 55 Medium","Vrinda","ZWAdobeF"];function H(e){return e.toDataURL()}var P,N;function z(){var e=this;return function(){if(void 0===N){var e=function(){var n=D();E(n)?N=setTimeout(e,2500):(P=n,N=void 0)};e()}}(),function(){return n(e,void 0,void 0,(function(){var e;return t(this,(function(n){switch(n.label){case 0:return E(e=D())?P?[2,r([],P,!0)]:M()?[4,(t=document,(t.exitFullscreen||t.msExitFullscreen||t.mozCancelFullScreen||t.webkitExitFullscreen).call(t))]:[3,2]:[3,2];case 1:n.sent(),e=D(),n.label=2;case 2:return E(e)||(P=e),[2,e]}var t}))}))}}function D(){var e=screen;return[y(b(e.availTop),null),y(b(e.width)-b(e.availWidth)-y(b(e.availLeft),0),null),y(b(e.height)-b(e.availHeight)-y(b(e.availTop),0),null),y(b(e.availLeft),null)]}function E(e){for(var n=0;n<4;++n)if(e[n])return!1;return!0}function T(e){var r;return n(this,void 0,void 0,(function(){var n,a,i,c,u,l,s;return t(this,(function(t){switch(t.label){case 0:for(n=document,a=n.createElement("div"),i=new Array(e.length),c={},B(a),s=0;s<e.length;++s)"DIALOG"===(u=A(e[s])).tagName&&u.show(),B(l=n.createElement("div")),l.appendChild(u),a.appendChild(l),i[s]=u;t.label=1;case 1:return n.body?[3,3]:[4,o(50)];case 2:return t.sent(),[3,1];case 3:n.body.appendChild(a);try{for(s=0;s<e.length;++s)i[s].offsetParent||(c[e[s]]=!0)}finally{null===(r=a.parentNode)||void 0===r||r.removeChild(a)}return[2,c]}}))}))}function B(e){e.style.setProperty("display","block","important")}function _(e){return matchMedia("(inverted-colors: ".concat(e,")")).matches}function O(e){return matchMedia("(forced-colors: ".concat(e,")")).matches}function U(e){return matchMedia("(prefers-contrast: ".concat(e,")")).matches}function Q(e){return matchMedia("(prefers-reduced-motion: ".concat(e,")")).matches}function K(e){return matchMedia("(dynamic-range: ".concat(e,")")).matches}var q=Math,$=function(){return 0};var ee={default:[],apple:[{font:"-apple-system-body"}],serif:[{fontFamily:"serif"}],sans:[{fontFamily:"sans-serif"}],mono:[{fontFamily:"monospace"}],min:[{fontSize:"1px"}],system:[{fontFamily:"system-ui"}]};var ne={fonts:function(){return X((function(e,n){var t=n.document,r=t.body;r.style.fontSize="48px";var o=t.createElement("div"),a={},i={},c=function(e){var n=t.createElement("span"),r=n.style;return r.position="absolute",r.top="0",r.left="0",r.fontFamily=e,n.textContent="mmMwWLliI0O&1",o.appendChild(n),n},u=I.map(c),l=function(){for(var e={},n=function(n){e[n]=I.map((function(e){return function(e,n){return c("'".concat(e,"',").concat(n))}(n,e)}))},t=0,r=J;t<r.length;t++){n(r[t])}return e}();r.appendChild(o);for(var s=0;s<I.length;s++)a[I[s]]=u[s].offsetWidth,i[I[s]]=u[s].offsetHeight;return J.filter((function(e){return n=l[e],I.some((function(e,t){return n[t].offsetWidth!==a[e]||n[t].offsetHeight!==i[e]}));var n}))}))},domBlockers:function(e){var r=(void 0===e?{}:e).debug;return n(this,void 0,void 0,(function(){var e,n,o,a,i;return t(this,(function(t){switch(t.label){case 0:return x()||G()?(c=atob,e={abpIndo:["#Iklan-Melayang","#Kolom-Iklan-728","#SidebarIklan-wrapper",'[title="ALIENBOLA" i]',c("I0JveC1CYW5uZXItYWRz")],abpvn:[".quangcao","#mobileCatfish",c("LmNsb3NlLWFkcw=="),'[id^="bn_bottom_fixed_"]',"#pmadv"],adBlockFinland:[".mainostila",c("LnNwb25zb3JpdA=="),".ylamainos",c("YVtocmVmKj0iL2NsaWNrdGhyZ2guYXNwPyJd"),c("YVtocmVmXj0iaHR0cHM6Ly9hcHAucmVhZHBlYWsuY29tL2FkcyJd")],adBlockPersian:["#navbar_notice_50",".kadr",'TABLE[width="140px"]',"#divAgahi",c("YVtocmVmXj0iaHR0cDovL2cxLnYuZndtcm0ubmV0L2FkLyJd")],adBlockWarningRemoval:["#adblock-honeypot",".adblocker-root",".wp_adblock_detect",c("LmhlYWRlci1ibG9ja2VkLWFk"),c("I2FkX2Jsb2NrZXI=")],adGuardAnnoyances:[".hs-sosyal","#cookieconsentdiv",'div[class^="app_gdpr"]',".as-oil",'[data-cypress="soft-push-notification-modal"]'],adGuardBase:[".BetterJsPopOverlay",c("I2FkXzMwMFgyNTA="),c("I2Jhbm5lcmZsb2F0MjI="),c("I2NhbXBhaWduLWJhbm5lcg=="),c("I0FkLUNvbnRlbnQ=")],adGuardChinese:[c("LlppX2FkX2FfSA=="),c("YVtocmVmKj0iLmh0aGJldDM0LmNvbSJd"),"#widget-quan",c("YVtocmVmKj0iLzg0OTkyMDIwLnh5eiJd"),c("YVtocmVmKj0iLjE5NTZobC5jb20vIl0=")],adGuardFrench:["#pavePub",c("LmFkLWRlc2t0b3AtcmVjdGFuZ2xl"),".mobile_adhesion",".widgetadv",c("LmFkc19iYW4=")],adGuardGerman:['aside[data-portal-id="leaderboard"]'],adGuardJapanese:["#kauli_yad_1",c("YVtocmVmXj0iaHR0cDovL2FkMi50cmFmZmljZ2F0ZS5uZXQvIl0="),c("Ll9wb3BJbl9pbmZpbml0ZV9hZA=="),c("LmFkZ29vZ2xl"),c("Ll9faXNib29zdFJldHVybkFk")],adGuardMobile:[c("YW1wLWF1dG8tYWRz"),c("LmFtcF9hZA=="),'amp-embed[type="24smi"]',"#mgid_iframe1",c("I2FkX2ludmlld19hcmVh")],adGuardRussian:[c("YVtocmVmXj0iaHR0cHM6Ly9hZC5sZXRtZWFkcy5jb20vIl0="),c("LnJlY2xhbWE="),'div[id^="smi2adblock"]',c("ZGl2W2lkXj0iQWRGb3hfYmFubmVyXyJd"),"#psyduckpockeball"],adGuardSocial:[c("YVtocmVmXj0iLy93d3cuc3R1bWJsZXVwb24uY29tL3N1Ym1pdD91cmw9Il0="),c("YVtocmVmXj0iLy90ZWxlZ3JhbS5tZS9zaGFyZS91cmw/Il0="),".etsy-tweet","#inlineShare",".popup-social"],adGuardSpanishPortuguese:["#barraPublicidade","#Publicidade","#publiEspecial","#queTooltip",".cnt-publi"],adGuardTrackingProtection:["#qoo-counter",c("YVtocmVmXj0iaHR0cDovL2NsaWNrLmhvdGxvZy5ydS8iXQ=="),c("YVtocmVmXj0iaHR0cDovL2hpdGNvdW50ZXIucnUvdG9wL3N0YXQucGhwIl0="),c("YVtocmVmXj0iaHR0cDovL3RvcC5tYWlsLnJ1L2p1bXAiXQ=="),"#top100counter"],adGuardTurkish:["#backkapat",c("I3Jla2xhbWk="),c("YVtocmVmXj0iaHR0cDovL2Fkc2Vydi5vbnRlay5jb20udHIvIl0="),c("YVtocmVmXj0iaHR0cDovL2l6bGVuemkuY29tL2NhbXBhaWduLyJd"),c("YVtocmVmXj0iaHR0cDovL3d3dy5pbnN0YWxsYWRzLm5ldC8iXQ==")],bulgarian:[c("dGQjZnJlZW5ldF90YWJsZV9hZHM="),"#ea_intext_div",".lapni-pop-over","#xenium_hot_offers"],easyList:[".yb-floorad",c("LndpZGdldF9wb19hZHNfd2lkZ2V0"),c("LnRyYWZmaWNqdW5reS1hZA=="),".textad_headline",c("LnNwb25zb3JlZC10ZXh0LWxpbmtz")],easyListChina:[c("LmFwcGd1aWRlLXdyYXBbb25jbGljayo9ImJjZWJvcy5jb20iXQ=="),c("LmZyb250cGFnZUFkdk0="),"#taotaole","#aafoot.top_box",".cfa_popup"],easyListCookie:[".ezmob-footer",".cc-CookieWarning","[data-cookie-number]",c("LmF3LWNvb2tpZS1iYW5uZXI="),".sygnal24-gdpr-modal-wrap"],easyListCzechSlovak:["#onlajny-stickers",c("I3Jla2xhbW5pLWJveA=="),c("LnJla2xhbWEtbWVnYWJvYXJk"),".sklik",c("W2lkXj0ic2tsaWtSZWtsYW1hIl0=")],easyListDutch:[c("I2FkdmVydGVudGll"),c("I3ZpcEFkbWFya3RCYW5uZXJCbG9jaw=="),".adstekst",c("YVtocmVmXj0iaHR0cHM6Ly94bHR1YmUubmwvY2xpY2svIl0="),"#semilo-lrectangle"],easyListGermany:["#SSpotIMPopSlider",c("LnNwb25zb3JsaW5rZ3J1ZW4="),c("I3dlcmJ1bmdza3k="),c("I3Jla2xhbWUtcmVjaHRzLW1pdHRl"),c("YVtocmVmXj0iaHR0cHM6Ly9iZDc0Mi5jb20vIl0=")],easyListItaly:[c("LmJveF9hZHZfYW5udW5jaQ=="),".sb-box-pubbliredazionale",c("YVtocmVmXj0iaHR0cDovL2FmZmlsaWF6aW9uaWFkcy5zbmFpLml0LyJd"),c("YVtocmVmXj0iaHR0cHM6Ly9hZHNlcnZlci5odG1sLml0LyJd"),c("YVtocmVmXj0iaHR0cHM6Ly9hZmZpbGlhemlvbmlhZHMuc25haS5pdC8iXQ==")],easyListLithuania:[c("LnJla2xhbW9zX3RhcnBhcw=="),c("LnJla2xhbW9zX251b3JvZG9z"),c("aW1nW2FsdD0iUmVrbGFtaW5pcyBza3lkZWxpcyJd"),c("aW1nW2FsdD0iRGVkaWt1b3RpLmx0IHNlcnZlcmlhaSJd"),c("aW1nW2FsdD0iSG9zdGluZ2FzIFNlcnZlcmlhaS5sdCJd")],estonian:[c("QVtocmVmKj0iaHR0cDovL3BheTRyZXN1bHRzMjQuZXUiXQ==")],fanboyAnnoyances:["#ac-lre-player",".navigate-to-top","#subscribe_popup",".newsletter_holder","#back-top"],fanboyAntiFacebook:[".util-bar-module-firefly-visible"],fanboyEnhancedTrackers:[".open.pushModal","#issuem-leaky-paywall-articles-zero-remaining-nag","#sovrn_container",'div[class$="-hide"][zoompage-fontsize][style="display: block;"]',".BlockNag__Card"],fanboySocial:["#FollowUs","#meteored_share","#social_follow",".article-sharer",".community__social-desc"],frellwitSwedish:[c("YVtocmVmKj0iY2FzaW5vcHJvLnNlIl1bdGFyZ2V0PSJfYmxhbmsiXQ=="),c("YVtocmVmKj0iZG9rdG9yLXNlLm9uZWxpbmsubWUiXQ=="),"article.category-samarbete",c("ZGl2LmhvbGlkQWRz"),"ul.adsmodern"],greekAdBlock:[c("QVtocmVmKj0iYWRtYW4ub3RlbmV0LmdyL2NsaWNrPyJd"),c("QVtocmVmKj0iaHR0cDovL2F4aWFiYW5uZXJzLmV4b2R1cy5nci8iXQ=="),c("QVtocmVmKj0iaHR0cDovL2ludGVyYWN0aXZlLmZvcnRobmV0LmdyL2NsaWNrPyJd"),"DIV.agores300","TABLE.advright"],hungarian:["#cemp_doboz",".optimonk-iframe-container",c("LmFkX19tYWlu"),c("W2NsYXNzKj0iR29vZ2xlQWRzIl0="),"#hirdetesek_box"],iDontCareAboutCookies:['.alert-info[data-block-track*="CookieNotice"]',".ModuleTemplateCookieIndicator",".o--cookies--container","#cookies-policy-sticky","#stickyCookieBar"],icelandicAbp:[c("QVtocmVmXj0iL2ZyYW1ld29yay9yZXNvdXJjZXMvZm9ybXMvYWRzLmFzcHgiXQ==")],latvian:[c("YVtocmVmPSJodHRwOi8vd3d3LnNhbGlkemluaS5sdi8iXVtzdHlsZT0iZGlzcGxheTogYmxvY2s7IHdpZHRoOiAxMjBweDsgaGVpZ2h0OiA0MHB4OyBvdmVyZmxvdzogaGlkZGVuOyBwb3NpdGlvbjogcmVsYXRpdmU7Il0="),c("YVtocmVmPSJodHRwOi8vd3d3LnNhbGlkemluaS5sdi8iXVtzdHlsZT0iZGlzcGxheTogYmxvY2s7IHdpZHRoOiA4OHB4OyBoZWlnaHQ6IDMxcHg7IG92ZXJmbG93OiBoaWRkZW47IHBvc2l0aW9uOiByZWxhdGl2ZTsiXQ==")],listKr:[c("YVtocmVmKj0iLy9hZC5wbGFuYnBsdXMuY28ua3IvIl0="),c("I2xpdmVyZUFkV3JhcHBlcg=="),c("YVtocmVmKj0iLy9hZHYuaW1hZHJlcC5jby5rci8iXQ=="),c("aW5zLmZhc3R2aWV3LWFk"),".revenue_unit_item.dable"],listeAr:[c("LmdlbWluaUxCMUFk"),".right-and-left-sponsers",c("YVtocmVmKj0iLmFmbGFtLmluZm8iXQ=="),c("YVtocmVmKj0iYm9vcmFxLm9yZyJd"),c("YVtocmVmKj0iZHViaXp6bGUuY29tL2FyLz91dG1fc291cmNlPSJd")],listeFr:[c("YVtocmVmXj0iaHR0cDovL3Byb21vLnZhZG9yLmNvbS8iXQ=="),c("I2FkY29udGFpbmVyX3JlY2hlcmNoZQ=="),c("YVtocmVmKj0id2Vib3JhbWEuZnIvZmNnaS1iaW4vIl0="),".site-pub-interstitiel",'div[id^="crt-"][data-criteo-id]'],officialPolish:["#ceneo-placeholder-ceneo-12",c("W2hyZWZePSJodHRwczovL2FmZi5zZW5kaHViLnBsLyJd"),c("YVtocmVmXj0iaHR0cDovL2Fkdm1hbmFnZXIudGVjaGZ1bi5wbC9yZWRpcmVjdC8iXQ=="),c("YVtocmVmXj0iaHR0cDovL3d3dy50cml6ZXIucGwvP3V0bV9zb3VyY2UiXQ=="),c("ZGl2I3NrYXBpZWNfYWQ=")],ro:[c("YVtocmVmXj0iLy9hZmZ0cmsuYWx0ZXgucm8vQ291bnRlci9DbGljayJd"),c("YVtocmVmXj0iaHR0cHM6Ly9ibGFja2ZyaWRheXNhbGVzLnJvL3Ryay9zaG9wLyJd"),c("YVtocmVmXj0iaHR0cHM6Ly9ldmVudC4ycGVyZm9ybWFudC5jb20vZXZlbnRzL2NsaWNrIl0="),c("YVtocmVmXj0iaHR0cHM6Ly9sLnByb2ZpdHNoYXJlLnJvLyJd"),'a[href^="/url/"]'],ruAd:[c("YVtocmVmKj0iLy9mZWJyYXJlLnJ1LyJd"),c("YVtocmVmKj0iLy91dGltZy5ydS8iXQ=="),c("YVtocmVmKj0iOi8vY2hpa2lkaWtpLnJ1Il0="),"#pgeldiz",".yandex-rtb-block"],thaiAds:["a[href*=macau-uta-popup]",c("I2Fkcy1nb29nbGUtbWlkZGxlX3JlY3RhbmdsZS1ncm91cA=="),c("LmFkczMwMHM="),".bumq",".img-kosana"],webAnnoyancesUltralist:["#mod-social-share-2","#social-tools",c("LmN0cGwtZnVsbGJhbm5lcg=="),".zergnet-recommend",".yt.btn-link.btn-md.btn"]},n=Object.keys(e),[4,T((i=[]).concat.apply(i,n.map((function(n){return e[n]}))))]):[2,void 0];case 1:return o=t.sent(),r&&function(e,n){for(var t="DOM blockers debug:\n```",r=0,o=Object.keys(e);r<o.length;r++){var a=o[r];t+="\n".concat(a,":");for(var i=0,c=e[a];i<c.length;i++){var u=c[i];t+="\n  ".concat(n[u]?"🚫":"➡️"," ").concat(u)}}console.log("".concat(t,"\n```"))}(e,o),(a=n.filter((function(n){var t=e[n];return g(t.map((function(e){return o[e]})))>.6*t.length}))).sort(),[2,a]}var c}))}))},fontPreferences:function(){return function(e,n){void 0===n&&(n=4e3);return X((function(t,o){var a=o.document,i=a.body,c=i.style;c.width="".concat(n,"px"),c.webkitTextSizeAdjust=c.textSizeAdjust="none",S()?i.style.zoom="".concat(1/o.devicePixelRatio):x()&&(i.style.zoom="reset");var u=a.createElement("div");return u.textContent=r([],Array(n/20<<0),!0).map((function(){return"word"})).join(" "),i.appendChild(u),e(a,i)}),'<!doctype html><html><head><meta name="viewport" content="width=device-width, initial-scale=1">')}((function(e,n){for(var t={},r={},o=0,a=Object.keys(ee);o<a.length;o++){var i=a[o],c=ee[i],u=c[0],l=void 0===u?{}:u,s=c[1],d=void 0===s?"mmMwWLliI0fiflO&1":s,m=e.createElement("span");m.textContent=d,m.style.whiteSpace="nowrap";for(var f=0,v=Object.keys(l);f<v.length;f++){var h=v[f],p=l[h];void 0!==p&&(m.style[h]=p)}t[i]=m,n.appendChild(e.createElement("br")),n.appendChild(m)}for(var b=0,y=Object.keys(ee);b<y.length;b++){r[i=y[b]]=t[i].getBoundingClientRect().width}return r}))},audio:function(){var e=window,n=e.OfflineAudioContext||e.webkitOfflineAudioContext;if(!n)return-2;if(x()&&!F()&&!function(){var e=window;return g(["DOMRectList"in e,"RTCPeerConnectionIceEvent"in e,"SVGGeometryElement"in e,"ontransitioncancel"in e])>=3}())return-1;var t=new n(1,5e3,44100),r=t.createOscillator();r.type="triangle",r.frequency.value=1e4;var o=t.createDynamicsCompressor();o.threshold.value=-50,o.knee.value=40,o.ratio.value=12,o.attack.value=0,o.release.value=.25,r.connect(o),o.connect(t.destination),r.start(0);var i=function(e){var n=3,t=500,r=500,o=5e3,i=function(){};return[new Promise((function(c,l){var s=!1,d=0,m=0;e.oncomplete=function(e){return c(e.renderedBuffer)};var f=function(){setTimeout((function(){return l(R("timeout"))}),Math.min(r,m+o-Date.now()))},v=function(){try{var r=e.startRendering();switch(a(r)&&u(r),e.state){case"running":m=Date.now(),s&&f();break;case"suspended":document.hidden||d++,s&&d>=n?l(R("suspended")):setTimeout(v,t)}}catch(o){l(o)}};v(),i=function(){s||(s=!0,m>0&&f())}})),i]}(t),c=i[0],l=i[1],s=c.then((function(e){return function(e){for(var n=0,t=0;t<e.length;++t)n+=Math.abs(e[t]);return n}(e.getChannelData(0).subarray(4500))}),(function(e){if("timeout"===e.name||"suspended"===e.name)return-3;throw e}));return u(s),function(){return l(),s}},screenFrame:function(){var e=this,r=z();return function(){return n(e,void 0,void 0,(function(){var e,n;return t(this,(function(t){switch(t.label){case 0:return[4,r()];case 1:return e=t.sent(),[2,[(n=function(e){return null===e?null:w(e,10)})(e[0]),n(e[1]),n(e[2]),n(e[3])]]}}))}))}},osCpu:function(){return navigator.oscpu},languages:function(){var e,n=navigator,t=[],r=n.language||n.userLanguage||n.browserLanguage||n.systemLanguage;if(void 0!==r&&t.push([r]),Array.isArray(n.languages))S()&&g([!("MediaSettingsRange"in(e=window)),"RTCEncodedAudioFrame"in e,""+e.Intl=="[object Intl]",""+e.Reflect=="[object Reflect]"])>=3||t.push(n.languages);else if("string"==typeof n.languages){var o=n.languages;o&&t.push(o.split(","))}return t},colorDepth:function(){return window.screen.colorDepth},deviceMemory:function(){return y(b(navigator.deviceMemory),void 0)},screenResolution:function(){var e=screen,n=function(e){return y(p(e),null)},t=[n(e.width),n(e.height)];return t.sort().reverse(),t},hardwareConcurrency:function(){return y(p(navigator.hardwareConcurrency),void 0)},timezone:function(){var e,n=null===(e=window.Intl)||void 0===e?void 0:e.DateTimeFormat;if(n){var t=(new n).resolvedOptions().timeZone;if(t)return t}var r,o=(r=(new Date).getFullYear(),-Math.max(b(new Date(r,0,1).getTimezoneOffset()),b(new Date(r,6,1).getTimezoneOffset())));return"UTC".concat(o>=0?"+":"").concat(Math.abs(o))},sessionStorage:function(){try{return!!window.sessionStorage}catch(e){return!0}},localStorage:function(){try{return!!window.localStorage}catch(e){return!0}},indexedDB:function(){if(!W()&&!C())try{return!!window.indexedDB}catch(e){return!0}},openDatabase:function(){return!!window.openDatabase},cpuClass:function(){return navigator.cpuClass},platform:function(){var e=navigator.platform;return"MacIntel"===e&&x()&&!F()?function(){if("iPad"===navigator.platform)return!0;var e=screen,n=e.width/e.height;return g(["MediaSource"in window,!!Element.prototype.webkitRequestFullscreen,n>.65&&n<1.53])>=2}()?"iPad":"iPhone":e},plugins:function(){var e=navigator.plugins;if(e){for(var n=[],t=0;t<e.length;++t){var r=e[t];if(r){for(var o=[],a=0;a<r.length;++a){var i=r[a];o.push({type:i.type,suffixes:i.suffixes})}n.push({name:r.name,description:r.description,mimeTypes:o})}}return n}},canvas:function(){var e,n,t=!1,r=function(){var e=document.createElement("canvas");return e.width=1,e.height=1,[e,e.getContext("2d")]}(),o=r[0],a=r[1];if(function(e,n){return!(!n||!e.toDataURL)}(o,a)){t=function(e){return e.rect(0,0,10,10),e.rect(2,2,6,6),!e.isPointInPath(5,5,"evenodd")}(a),function(e,n){e.width=240,e.height=60,n.textBaseline="alphabetic",n.fillStyle="#f60",n.fillRect(100,1,62,20),n.fillStyle="#069",n.font='11pt "Times New Roman"';var t="Cwm fjordbank gly ".concat(String.fromCharCode(55357,56835));n.fillText(t,2,15),n.fillStyle="rgba(102, 204, 0, 0.2)",n.font="18pt Arial",n.fillText(t,4,45)}(o,a);var i=H(o);i!==H(o)?e=n="unstable":(n=i,function(e,n){e.width=122,e.height=110,n.globalCompositeOperation="multiply";for(var t=0,r=[["#f2f",40,40],["#2ff",80,40],["#ff2",60,80]];t<r.length;t++){var o=r[t],a=o[0],i=o[1],c=o[2];n.fillStyle=a,n.beginPath(),n.arc(i,c,40,0,2*Math.PI,!0),n.closePath(),n.fill()}n.fillStyle="#f9c",n.arc(60,60,60,0,2*Math.PI,!0),n.arc(60,60,20,0,2*Math.PI,!0),n.fill("evenodd")}(o,a),e=H(o))}else e=n="";return{winding:t,geometry:e,text:n}},touchSupport:function(){var e,n=navigator,t=0;void 0!==n.maxTouchPoints?t=p(n.maxTouchPoints):void 0!==n.msMaxTouchPoints&&(t=n.msMaxTouchPoints);try{document.createEvent("TouchEvent"),e=!0}catch(r){e=!1}return{maxTouchPoints:t,touchEvent:e,touchStart:"ontouchstart"in window}},vendor:function(){return navigator.vendor||""},vendorFlavors:function(){for(var e=[],n=0,t=["chrome","safari","__crWeb","__gCrWeb","yandex","__yb","__ybro","__firefox__","__edgeTrackingPreventionStatistics","webkit","oprt","samsungAr","ucweb","UCShellJava","puffinDevice"];n<t.length;n++){var r=t[n],o=window[r];o&&"object"==typeof o&&e.push(r)}return e.sort()},cookiesEnabled:function(){var e=document;try{e.cookie="cookietest=1; SameSite=Strict;";var n=-1!==e.cookie.indexOf("cookietest=");return e.cookie="cookietest=1; SameSite=Strict; expires=Thu, 01-Jan-1970 00:00:01 GMT",n}catch(t){return!1}},colorGamut:function(){for(var e=0,n=["rec2020","p3","srgb"];e<n.length;e++){var t=n[e];if(matchMedia("(color-gamut: ".concat(t,")")).matches)return t}},invertedColors:function(){return!!_("inverted")||!_("none")&&void 0},forcedColors:function(){return!!O("active")||!O("none")&&void 0},monochrome:function(){if(matchMedia("(min-monochrome: 0)").matches){for(var e=0;e<=100;++e)if(matchMedia("(max-monochrome: ".concat(e,")")).matches)return e;throw new Error("Too high value")}},contrast:function(){return U("no-preference")?0:U("high")||U("more")?1:U("low")||U("less")?-1:U("forced")?10:void 0},reducedMotion:function(){return!!Q("reduce")||!Q("no-preference")&&void 0},hdr:function(){return!!K("high")||!K("standard")&&void 0},math:function(){var e,n=q.acos||$,t=q.acosh||$,r=q.asin||$,o=q.asinh||$,a=q.atanh||$,i=q.atan||$,c=q.sin||$,u=q.sinh||$,l=q.cos||$,s=q.cosh||$,d=q.tan||$,m=q.tanh||$,f=q.exp||$,v=q.expm1||$,h=q.log1p||$;return{acos:n(.12312423423423424),acosh:t(1e308),acoshPf:(e=1e154,q.log(e+q.sqrt(e*e-1))),asin:r(.12312423423423424),asinh:o(1),asinhPf:function(e){return q.log(e+q.sqrt(e*e+1))}(1),atanh:a(.5),atanhPf:function(e){return q.log((1+e)/(1-e))/2}(.5),atan:i(.5),sin:c(-1e300),sinh:u(1),sinhPf:function(e){return q.exp(e)-1/q.exp(e)/2}(1),cos:l(10.000000000123),cosh:s(1),coshPf:function(e){return(q.exp(e)+1/q.exp(e))/2}(1),tan:d(-1e300),tanh:m(1),tanhPf:function(e){return(q.exp(2*e)-1)/(q.exp(2*e)+1)}(1),exp:f(1),expm1:v(1),expm1Pf:function(e){return q.exp(e)-1}(1),log1p:h(10),log1pPf:function(e){return q.log(1+e)}(10),powPI:function(e){return q.pow(q.PI,e)}(-100)}},videoCard:function(){var e,n=document.createElement("canvas"),t=null!==(e=n.getContext("webgl"))&&void 0!==e?e:n.getContext("experimental-webgl");if(t&&"getExtension"in t){var r=t.getExtension("WEBGL_debug_renderer_info");if(r)return{vendor:(t.getParameter(r.UNMASKED_VENDOR_WEBGL)||"").toString(),renderer:(t.getParameter(r.UNMASKED_RENDERER_WEBGL)||"").toString()}}},pdfViewerEnabled:function(){return navigator.pdfViewerEnabled},architecture:function(){var e=new Float32Array(1),n=new Uint8Array(e.buffer);return e[0]=1/0,e[0]=e[0]-e[0],n[3]}};function te(e){var n=function(e){if(G())return.4;if(x())return F()?.5:.3;var n=e.platform.value||"";if(/^Win/.test(n))return.6;if(/^Mac/.test(n))return.5;return.7}(e),t=function(e){return w(.99+.01*e,1e-4)}(n);return{score:n,comment:"$ if upgrade to Pro: https://fpjs.dev/pro".replace(/\$/g,"".concat(t))}}function re(n){return JSON.stringify(n,(function(n,t){return t instanceof Error?e({name:(r=t).name,message:r.message,stack:null===(o=r.stack)||void 0===o?void 0:o.split("\n")},r):t;var r,o}),2)}function oe(e){return h(function(e){for(var n="",t=0,r=Object.keys(e).sort();t<r.length;t++){var o=r[t],a=e[o],i=a.error?"error":JSON.stringify(a.value);n+="".concat(n?"|":"").concat(o.replace(/([:|\\])/g,"\\$1"),":").concat(i)}return n}(e))}function ae(e){return void 0===e&&(e=50),function(e,n){void 0===n&&(n=1/0);var t=window.requestIdleCallback;return t?new Promise((function(e){return t.call(window,(function(){return e()}),{timeout:n})})):o(Math.min(e,n))}(e,2*e)}function ie(e,r){var o=Date.now();return{get:function(a){return n(this,void 0,void 0,(function(){var n,i,c;return t(this,(function(t){switch(t.label){case 0:return n=Date.now(),[4,e()];case 1:return i=t.sent(),c=function(e){var n;return{get visitorId(){return void 0===n&&(n=oe(this.components)),n},set visitorId(e){n=e},confidence:te(e),components:e,version:"3.4.2"}}(i),(r||(null==a?void 0:a.debug))&&console.log("Copy the text below to get the debug data:\n\n```\nversion: ".concat(c.version,"\nuserAgent: ").concat(navigator.userAgent,"\ntimeBetweenLoadAndGet: ").concat(n-o,"\nvisitorId: ").concat(c.visitorId,"\ncomponents: ").concat(re(i),"\n```")),[2,c]}}))}))}}}function ce(e){var r=void 0===e?{}:e,o=r.delayFallback,a=r.debug;return r.monitoring,n(this,void 0,void 0,(function(){return t(this,(function(e){switch(e.label){case 0:return[4,ae(o)];case 1:return e.sent(),[2,ie(V(ne,{debug:a},[]),a)]}}))}))}var ue={load:ce,hashComponents:oe,componentsToDebugString:re},le=h;export{re as componentsToDebugString,ue as default,M as getFullscreenElement,z as getScreenFrame,oe as hashComponents,G as isAndroid,S as isChromium,F as isDesktopSafari,C as isEdgeHTML,Y as isGecko,W as isTrident,x as isWebKit,ce as load,V as loadSources,le as murmurX64Hash128,ae as prepareForSources,ne as sources,Z as transformSource,X as withIframe};

upgini/metrics.py

@@ -3,15 +3,16 @@ import re
 from copy import deepcopy
 from typing import Any, Callable, Dict, List, Optional, Tuple, Union
 
+import catboost
 import numpy as np
 import pandas as pd
 from catboost import CatBoostClassifier, CatBoostRegressor
-import catboost
 from lightgbm import LGBMClassifier, LGBMRegressor
 from numpy import log1p
 from pandas.api.types import is_numeric_dtype
 from sklearn.metrics import check_scoring, get_scorer, make_scorer, roc_auc_score
 
+from upgini.utils.features_validator import FeaturesValidator
 from upgini.utils.sklearn_ext import cross_validate
 
 try:
@@ -352,6 +353,7 @@ class EstimatorWrapper:
             "target_type": target_type,
             "groups": groups,
             "text_features": text_features,
+            "logger": logger,
         }
         if estimator is None:
             params = dict()
@@ -414,12 +416,22 @@ class CatBoostWrapper(EstimatorWrapper):
         target_type: ModelTaskType,
         groups: Optional[List[str]] = None,
         text_features: Optional[List[str]] = None,
+        logger: Optional[logging.Logger] = None,
     ):
         super(CatBoostWrapper, self).__init__(
-            estimator, scorer, metric_name, multiplier, cv, target_type, groups=groups, text_features=text_features
+            estimator,
+            scorer,
+            metric_name,
+            multiplier,
+            cv,
+            target_type,
+            groups=groups,
+            text_features=text_features,
+            logger=logger,
         )
         self.cat_features = None
         self.emb_features = None
+        self.exclude_features = []
 
     def _prepare_to_fit(self, X: pd.DataFrame, y: pd.Series) -> Tuple[pd.DataFrame, np.ndarray, np.ndarray, dict]:
         X, y, groups, params = super()._prepare_to_fit(X, y)
@@ -437,9 +449,7 @@ class CatBoostWrapper(EstimatorWrapper):
                 X, embedding_features = self.group_embeddings(X)
                 params["embedding_features"] = embedding_features
             else:
-                self.logger.info(
-                    f"Embedding features count less than 3, so use them separately: {self.emb_features}"
-                )
+                self.logger.info(f"Embedding features count less than 3, so use them separately: {self.emb_features}")
                 self.emb_features = []
         else:
             self.logger.warning(f"Embedding features are not supported by Catboost version {catboost.__version__}")
@@ -498,6 +508,8 @@ class CatBoostWrapper(EstimatorWrapper):
         return df, [emb_name]
 
     def _prepare_to_calculate(self, X: pd.DataFrame, y: pd.Series) -> Tuple[pd.DataFrame, np.ndarray, dict]:
+        if self.exclude_features:
+            X = X.drop(columns=self.exclude_features)
         X, y, params = super()._prepare_to_calculate(X, y)
         if self.text_features:
             params["text_features"] = self.text_features
@@ -510,6 +522,26 @@ class CatBoostWrapper(EstimatorWrapper):
 
         return X, y, params
 
+    def cross_val_predict(
+        self, X: pd.DataFrame, y: np.ndarray, baseline_score_column: Optional[Any] = None
+    ) -> Optional[float]:
+        try:
+            return super().cross_val_predict(X, y, baseline_score_column)
+        except Exception as e:
+            if "Dictionary size is 0" in e.args[0] and self.text_features:
+                high_cardinality_features = FeaturesValidator.find_high_cardinality(X[self.text_features])
+                self.logger.warning(
+                    "Failed to calculate metrics. Try to remove high cardinality"
+                    f" text features {high_cardinality_features} and retry"
+                )
+                for f in high_cardinality_features:
+                    self.text_features.remove(f)
+                    self.exclude_features.append(f)
+                    X = X.drop(columns=f)
+                return super().cross_val_predict(X, y, baseline_score_column)
+            else:
+                raise e
+
 
 class LightGBMWrapper(EstimatorWrapper):
     def __init__(
@@ -522,9 +554,18 @@ class LightGBMWrapper(EstimatorWrapper):
         target_type: ModelTaskType,
         groups: Optional[List[str]] = None,
         text_features: Optional[List[str]] = None,
+        logger: Optional[logging.Logger] = None,
     ):
         super(LightGBMWrapper, self).__init__(
-            estimator, scorer, metric_name, multiplier, cv, target_type, groups=groups, text_features=text_features
+            estimator,
+            scorer,
+            metric_name,
+            multiplier,
+            cv,
+            target_type,
+            groups=groups,
+            text_features=text_features,
+            logger=logger,
         )
         self.cat_features = None
 
@@ -561,9 +602,18 @@ class OtherEstimatorWrapper(EstimatorWrapper):
         target_type: ModelTaskType,
         groups: Optional[List[str]] = None,
         text_features: Optional[List[str]] = None,
+        logger: Optional[logging.Logger] = None,
     ):
         super(OtherEstimatorWrapper, self).__init__(
-            estimator, scorer, metric_name, multiplier, cv, target_type, groups=groups, text_features=text_features
+            estimator,
+            scorer,
+            metric_name,
+            multiplier,
+            cv,
+            target_type,
+            groups=groups,
+            text_features=text_features,
+            logger=logger,
         )
         self.cat_features = None
 
@@ -595,6 +645,7 @@ class OtherEstimatorWrapper(EstimatorWrapper):
 
 
 def validate_scoring_argument(scoring: Union[Callable, str, None]):
+    # TODO validate that if it is Callable then it accepts 3 arguments
     if isinstance(scoring, str) and scoring is not None:
         _get_scorer_by_name(scoring)
 

upgini/normalizer/phone_normalizer.py

@@ -1,7 +1,7 @@
 from typing import Optional
 
 import pandas as pd
-from pandas.api.types import is_float_dtype, is_int64_dtype, is_string_dtype
+from pandas.api.types import is_float_dtype, is_int64_dtype, is_string_dtype, is_object_dtype
 
 from upgini.errors import ValidationError
 
@@ -44,7 +44,7 @@ class PhoneNormalizer:
         Method will remove all non numeric chars from string and convert it to int.
         None will be set for phone numbers that couldn"t be converted to int
         """
-        if is_string_dtype(self.df[self.phone_column_name]):
+        if is_string_dtype(self.df[self.phone_column_name]) or is_object_dtype(self.df[self.phone_column_name]):
             convert_func = self.phone_str_to_int_safe
         elif is_float_dtype(self.df[self.phone_column_name]):
             convert_func = self.phone_float_to_int_safe