upgini 1.2.142a2__tar.gz → 1.2.144__tar.gz

{upgini-1.2.142a2 → upgini-1.2.144}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: upgini
-Version: 1.2.142a2
+Version: 1.2.144
 Summary: Intelligent data search & enrichment for Machine Learning
 Project-URL: Bug Reports, https://github.com/upgini/upgini/issues
 Project-URL: Homepage, https://upgini.com/

upgini-1.2.144/src/upgini/__about__.py

@@ -0,0 +1 @@
+__version__ = "1.2.144"

{upgini-1.2.142a2 → upgini-1.2.144}/src/upgini/features_enricher.py

@@ -274,7 +274,7 @@ class FeaturesEnricher(TransformerMixin):
         self.X: pd.DataFrame | None = None
         self.y: pd.Series | None = None
         self.eval_set: list[tuple] | None = None
-        self.autodetected_search_keys: dict[str, SearchKey] = dict()
+        self.autodetected_search_keys: dict[str, SearchKey] | None = None
         self.imbalanced = False
         self.fit_select_features = True
         self.__cached_sampled_datasets: dict[str, tuple[pd.DataFrame, pd.DataFrame, pd.Series, dict, dict, dict]] = (
@@ -1311,10 +1311,17 @@ class FeaturesEnricher(TransformerMixin):
     def _get_autodetected_search_keys(self):
         if self.autodetected_search_keys is None and self._search_task is not None:
             meta = self._search_task.get_file_metadata(self._get_trace_id())
-            self.autodetected_search_keys = {k: SearchKey[v] for k, v in meta.autodetectedSearchKeys.items()}
+            autodetected_search_keys = meta.autodetectedSearchKeys or {}
+            self.autodetected_search_keys = {k: SearchKey[v] for k, v in autodetected_search_keys.items()}
 
         return self.autodetected_search_keys
 
+    def _add_autodetected_search_keys(self, adding_search_keys: dict[str, SearchKey]):
+        if self.autodetected_search_keys is None:
+            self.autodetected_search_keys = dict()
+        self.autodetected_search_keys.update(adding_search_keys)
+        return self.autodetected_search_keys
+
     def _get_fit_search_keys_with_original_names(self):
         if self.fit_search_keys is None and self._search_task is not None:
             fit_search_keys = dict()
@@ -2553,9 +2560,7 @@ if response.status_code == 200:
 
         self.__validate_search_keys(search_keys, self.search_id)
 
-        validated_X, validated_y, validated_eval_set = self._validate_train_eval(
-            X, y, eval_set=None, is_transform=True
-        )
+        validated_X, validated_y, validated_eval_set = self._validate_train_eval(X, y, eval_set=None, is_transform=True)
         df = self.__combine_train_and_eval_sets(validated_X, validated_y, validated_eval_set)
 
         validated_Xy = df.copy()
@@ -2956,10 +2961,6 @@ if response.status_code == 200:
         if add_fit_system_record_id:
             result = result.rename(columns={SORT_ID: SYSTEM_RECORD_ID})
 
-        for c in result.columns:
-            if result[c].dtype == "category":
-                result.loc[:, c] = np.where(~result[c].isin(result[c].dtype.categories), np.nan, result[c])
-
         return result, columns_renaming, generated_features, search_keys
 
     def _selecting_input_and_generated_columns(
@@ -3649,7 +3650,8 @@ if response.status_code == 200:
             keys.append("EMAIL")
         if "DATE" in keys:
             keys.append("DATETIME")
-        search_keys_with_autodetection = {**self.search_keys, **self.autodetected_search_keys}
+        autodetected_search_keys = self.autodetected_search_keys or {}
+        search_keys_with_autodetection = {**self.search_keys, **autodetected_search_keys}
         return [c for c, v in search_keys_with_autodetection.items() if v.value.value in keys]
 
     def _validate_train_eval(
@@ -4784,7 +4786,7 @@ if response.status_code == 200:
         maybe_date = [k for k, v in valid_search_keys.items() if v in [SearchKey.DATE, SearchKey.DATETIME]]
         if (self.cv is None or self.cv == CVType.k_fold) and len(maybe_date) > 0 and not silent_mode:
             date_column = next(iter(maybe_date))
-            if x[date_column].nunique() > 0.9 * _num_samples(x):
+            if x[date_column].nunique() > 0.9 * _num_samples(x) and not is_transform:
                 msg = self.bundle.get("date_search_without_time_series")
                 self.__log_warning(msg)
 
@@ -4884,8 +4886,9 @@ if response.status_code == 200:
             maybe_keys = DateSearchKeyDetector().get_search_key_columns(sample, search_keys)
             if len(maybe_keys) > 0:
                 datetime_key = maybe_keys[0]
-                search_keys[datetime_key] = SearchKey.DATETIME
-                self.autodetected_search_keys[datetime_key] = SearchKey.DATETIME
+                new_keys = {datetime_key: SearchKey.DATETIME}
+                search_keys.update(new_keys)
+                self._add_autodetected_search_keys(new_keys)
                 self.logger.info(f"Autodetected search key DATETIME in column {datetime_key}")
                 print(self.bundle.get("datetime_detected").format(datetime_key))
 
@@ -4894,15 +4897,16 @@ if response.status_code == 200:
         if maybe_keys:
             new_keys = {key: SearchKey.POSTAL_CODE for key in maybe_keys}
             search_keys.update(new_keys)
-            self.autodetected_search_keys.update(new_keys)
+            self._add_autodetected_search_keys(new_keys)
             self.logger.info(f"Autodetected search key POSTAL_CODE in column {maybe_keys}")
             print(self.bundle.get("postal_code_detected").format(maybe_keys))
 
         if SearchKey.COUNTRY not in search_keys.values() and self.country_code is None:
             maybe_key = CountrySearchKeyDetector().get_search_key_columns(sample, search_keys)
             if maybe_key:
-                search_keys[maybe_key[0]] = SearchKey.COUNTRY
-                self.autodetected_search_keys[maybe_key[0]] = SearchKey.COUNTRY
+                new_keys = {maybe_key[0]: SearchKey.COUNTRY}
+                search_keys.update(new_keys)
+                self._add_autodetected_search_keys(new_keys)
                 self.logger.info(f"Autodetected search key COUNTRY in column {maybe_key}")
                 print(self.bundle.get("country_detected").format(maybe_key))
 
@@ -4912,7 +4916,7 @@ if response.status_code == 200:
                 if self.__is_registered or is_demo_dataset:
                     new_keys = {key: SearchKey.EMAIL for key in maybe_keys}
                     search_keys.update(new_keys)
-                    self.autodetected_search_keys.update(new_keys)
+                    self._add_autodetected_search_keys(new_keys)
                     self.logger.info(f"Autodetected search key EMAIL in column {maybe_keys}")
                     print(self.bundle.get("email_detected").format(maybe_keys))
                 else:
@@ -4928,7 +4932,7 @@ if response.status_code == 200:
             if self.__is_registered or is_demo_dataset:
                 new_keys = {key: SearchKey.PHONE for key in maybe_keys}
                 search_keys.update(new_keys)
-                self.autodetected_search_keys.update(new_keys)
+                self._add_autodetected_search_keys(new_keys)
                 self.logger.info(f"Autodetected search key PHONE in column {maybe_keys}")
                 print(self.bundle.get("phone_detected").format(maybe_keys))
             else:
@@ -5025,7 +5029,9 @@ if response.status_code == 200:
                                 f"File x.parquet was already uploaded with digest {x_digest_sha256}, skipping"
                             )
                         else:
-                            self.rest_client.dump_input_file(f"{tmp_dir}/x.parquet", "x.parquet", x_digest_sha256)
+                            self.rest_client.dump_input_file(
+                                trace_id_, f"{tmp_dir}/x.parquet", "x.parquet", x_digest_sha256
+                            )
 
                         if y_ is not None:
                             if isinstance(y_, pd.Series):

{upgini-1.2.142a2 → upgini-1.2.144}/src/upgini/search_task.py

@@ -434,4 +434,5 @@ def _read_parquet(file_content: bytes, file_name: str = "features.parquet"):
         tmp_file_name = f"{tmp_dir}/{file_name}"
         with open(tmp_file_name, "wb") as gzip_file:
             gzip_file.write(file_content)
-        return pd.read_parquet(tmp_file_name, engine="fastparquet")
+        # Note: MLB writes files using pyarrow, so reading with fastparquet may cause errors.
+        return pd.read_parquet(tmp_file_name, engine="pyarrow")

{upgini-1.2.142a2 → upgini-1.2.144}/src/upgini/utils/postal_code_utils.py

@@ -4,16 +4,49 @@ from pandas.api.types import (
     is_object_dtype,
     is_string_dtype,
 )
+import re
 
 from upgini.utils.base_search_key_detector import BaseSearchKeyDetector
 
 
 class PostalCodeSearchKeyDetector(BaseSearchKeyDetector):
+    postal_pattern = re.compile(r'^[A-Za-z0-9][A-Za-z0-9\s\-]{1,9}$')
+
     def _is_search_key_by_name(self, column_name: str) -> bool:
-        return str(column_name).lower() in ["zip", "zipcode", "zip_code", "postal_code", "postalcode"]
+        return "zip" in str(column_name).lower() or "postal" in str(column_name).lower()
 
     def _is_search_key_by_values(self, column: pd.Series) -> bool:
-        return False
+        """
+        # Fast two-step check whether the column looks like a postal code.
+        # Returns True if, after removing missing values, values remain,
+        # and all of them match the common characteristics of a postal code.
+        """
+        # Check only columns that are candidates for postal code by column name
+        if not self._is_search_key_by_name(column.name):
+            return False
+
+        s = column.copy().dropna().astype(str).str.strip()
+        s = s[s != ""]  # remove empty strings
+        if s.empty:
+            return False
+
+        # remove suffix ".0" (often after float)
+        s = s.str.replace(r"\.0$", "", regex=True)
+
+        # --- Step 1: fast filtering ---
+        mask_len = s.str.len().between(2, 10)
+        mask_digit = s.str.contains(r'\d', regex=True)
+        mask_chars = ~s.str.contains(r'[^A-Za-z0-9\s\-]', regex=True)
+        fast_mask = mask_len & mask_digit & mask_chars
+
+        # if any of them failed the fast check, return False
+        if not fast_mask.all():
+            return False
+
+        # --- Step 2: regex check ---
+        # only if the first step passed
+        valid_mask = s.apply(lambda x: bool(self.postal_pattern.fullmatch(x)))
+        return valid_mask.all()
 
 
 class PostalCodeSearchKeyConverter:

upgini-1.2.142a2/src/upgini/__about__.py

@@ -1 +0,0 @@
-__version__ = "1.2.142a2"