Update to latest ES versions and fix unit tests (#512)

Update the test matrix to the latest Elasticsearch versions and fix the broken unit tests on the CI.

.ci/run-tests

@@ -12,7 +12,7 @@ set -euxo pipefail
 
 TEST_SUITE=${TEST_SUITE-xpack}
 NODE_NAME=localhost
-PANDAS_VERSION=${PANDAS_VERSION-1.3.0}
+PANDAS_VERSION=${PANDAS_VERSION-1.5.0}
 
 
 elasticsearch_image=elasticsearch

.ci/test-matrix.yml

@@ -1,18 +1,16 @@
 ---
 
 ELASTICSEARCH_VERSION:
-  - '8.1.0-SNAPSHOT'
-  - '8.0.0-SNAPSHOT'
+  - '8.6.0-SNAPSHOT'
+  - '8.7.0-SNAPSHOT'
 
 PANDAS_VERSION:
-  - '1.2.0'
-  - '1.3.0'
+  - '1.5.0'
 
 PYTHON_VERSION:
   - '3.10'
   - '3.9'
   - '3.8'
-  - '3.7'
 
 TEST_SUITE:
   - xpack

README.md

@@ -49,7 +49,7 @@ $ conda install -c conda-forge eland
 
 ### Compatibility
 
-- Supports Python 3.7+ and Pandas 1.3
+- Supports Python 3.8+ and Pandas 1.5
 - Supports Elasticsearch clusters that are 7.11+, recommended 8.3 or later for all features to work.
   If you are using the NLP with PyTorch feature make sure your Eland minor version matches the minor 
   version of your Elasticsearch cluster. For all other features it is sufficient for the major versions

docs/requirements-docs.txt

@@ -1,10 +1,11 @@
 elasticsearch>=7.7
-pandas>=1.2.0
-matplotlib
+pandas>=1.5
+matplotlib>=3.6
 nbval
 scikit-learn>=0.22.1
 xgboost>=1
 lightgbm
+sphinx==5.3.0
 nbsphinx
 git+https://github.com/pandas-dev/pydata-sphinx-theme.git
 

eland/ndframe.py

@@ -172,7 +172,7 @@ class NDFrame(ABC):
         head = self.head(head_rows).to_pandas()
         tail = self.tail(tail_rows).to_pandas()
 
-        return head.append(tail)
+        return pd.concat([head, tail])
 
     def __sizeof__(self) -> int:
         # Don't default to pandas, just return approximation TODO - make this more accurate

eland/series.py

@@ -714,7 +714,7 @@ class Series(NDFrame):
         >>> ed_ecommerce = ed.DataFrame('http://localhost:9200', 'ecommerce')
         >>> ed_ecommerce["day_of_week"].mode()
         0    Thursday
-        dtype: object
+        Name: day_of_week, dtype: object
 
         >>> ed_ecommerce["order_date"].mode()
         0   2016-12-02 20:36:58
@@ -727,16 +727,18 @@ class Series(NDFrame):
         7   2016-12-15 11:38:24
         8   2016-12-22 19:39:22
         9   2016-12-24 06:21:36
-        dtype: datetime64[ns]
+        Name: order_date, dtype: datetime64[ns]
 
         >>> ed_ecommerce["order_date"].mode(es_size=3)
         0   2016-12-02 20:36:58
         1   2016-12-04 23:44:10
         2   2016-12-08 06:21:36
-        dtype: datetime64[ns]
+        Name: order_date, dtype: datetime64[ns]
 
         """
-        return self._query_compiler.mode(is_dataframe=False, es_size=es_size)
+        result = self._query_compiler.mode(is_dataframe=False, es_size=es_size)
+        result.name = self.name
+        return result
 
     def es_match(
         self,

noxfile.py

@@ -100,8 +100,8 @@ def lint(session):
             session.error("\n" + "\n".join(sorted(set(errors))))
 
 
-@nox.session(python=["3.7", "3.8", "3.9", "3.10"])
-@nox.parametrize("pandas_version", ["1.2.0", "1.3.0"])
+@nox.session(python=["3.8", "3.9", "3.10"])
+@nox.parametrize("pandas_version", ["1.5.0"])
 def test(session, pandas_version: str):
     session.install("-r", "requirements-dev.txt")
     session.install(".")

requirements-dev.txt

@@ -2,8 +2,8 @@
 # Basic requirements
 #
 elasticsearch>=8.3,<9
-pandas>=1.2,<2
-matplotlib<4
+pandas>=1.5
+matplotlib>=3.6
 numpy<2
 tqdm<5
 
@@ -12,7 +12,6 @@ tqdm<5
 #
 scikit-learn>=0.22.1,<2
 xgboost>=0.90,<2
-scikit-learn>=0.22.1,<2
 lightgbm>=2,<4
 
 # PyTorch doesn't support Python 3.10 yet (pytorch/pytorch#66424)

requirements.txt

@@ -2,6 +2,6 @@
 # Basic requirements
 #
 elasticsearch>=8.3,<9
-pandas>=1.2,<2
-matplotlib<4
+pandas>=1.5
+matplotlib>=3.6
 numpy<2

setup.py

@@ -83,12 +83,12 @@ setup(
     packages=find_packages(include=["eland", "eland.*"]),
     install_requires=[
         "elasticsearch>=8.3,<9",
-        "pandas>=1.2,<2",
-        "matplotlib<4",
+        "pandas>=1.5",
+        "matplotlib>=3.6",
         "numpy<2",
     ],
     scripts=["bin/eland_import_hub_model"],
-    python_requires=">=3.7",
+    python_requires=">=3.8",
     package_data={"eland": ["py.typed"]},
     include_package_data=True,
     zip_safe=False,

tests/common.py

@@ -15,6 +15,8 @@
 #  specific language governing permissions and limitations
 #  under the License.
 
+import gzip
+import json
 import os
 from datetime import timedelta
 
@@ -30,15 +32,21 @@ from tests import (
     ECOMMERCE_DF_FILE_NAME,
     ECOMMERCE_INDEX_NAME,
     ES_TEST_CLIENT,
-    FLIGHTS_DF_FILE_NAME,
+    FLIGHTS_FILE_NAME,
     FLIGHTS_INDEX_NAME,
     FLIGHTS_SMALL_INDEX_NAME,
 )
 
-_pd_flights = pd.read_json(FLIGHTS_DF_FILE_NAME).sort_index()
+_ed_flights = ed.DataFrame(ES_TEST_CLIENT, FLIGHTS_INDEX_NAME)
+flight_records = []
+with gzip.open(FLIGHTS_FILE_NAME) as f:
+    for json_obj in f:
+        flight_records.append(json.loads(json_obj))
+_pd_flights = pd.DataFrame.from_records(flight_records).reindex(
+    _ed_flights.columns, axis=1
+)
 _pd_flights["timestamp"] = pd.to_datetime(_pd_flights["timestamp"])
 _pd_flights.index = _pd_flights.index.map(str)  # make index 'object' not int
-_ed_flights = ed.DataFrame(ES_TEST_CLIENT, FLIGHTS_INDEX_NAME)
 
 _pd_flights_small = _pd_flights.head(48)
 _ed_flights_small = ed.DataFrame(ES_TEST_CLIENT, FLIGHTS_SMALL_INDEX_NAME)