Merge pull request #2088 from Giskard-AI/feature/gsk-4014-set-encoding-as-utf-8-for-all-open-statements

[GSK-4014] Set encoding as utf-8 for all open statements

Author: kevinmessiaen

giskard/core/model_validation.py

@@ -147,7 +147,7 @@ def validate_model_loading_and_saving(model: BaseModel):
         with tempfile.TemporaryDirectory(prefix="giskard-model-") as f:
             model.save(f)
 
-            with open(f + "/giskard-model-meta.yaml") as yaml_f:
+            with open(f + "/giskard-model-meta.yaml", encoding="utf-8") as yaml_f:
                 saved_meta = yaml.load(yaml_f, Loader=yaml.Loader)
 
             meta = ModelMeta(

giskard/core/savable.py

@@ -54,7 +54,7 @@ def _get_meta_endpoint(cls, uuid: str, project_key: str) -> str:
         return posixpath.join("project", project_key, cls._get_name(), uuid)
 
     def _save_meta_locally(self, local_dir):
-        with open(Path(local_dir) / "meta.yaml", "w") as f:
+        with open(Path(local_dir) / "meta.yaml", "w", encoding="utf-8") as f:
             yaml.dump(self.meta, f)
 
 
@@ -70,7 +70,7 @@ def _load_meta_locally(cls, local_dir, uuid: str) -> Optional[SMT]:
         if meta is not None:
             return meta
 
-        with open(local_dir / "meta.yaml", "r") as f:
+        with open(local_dir / "meta.yaml", "r", encoding="utf-8") as f:
             return yaml.load(f, Loader=yaml.Loader)
 
     @classmethod

giskard/core/suite.py

@@ -162,8 +162,8 @@ def to_json(self, filename=None):
                 "metric_value": suite_result.result.metric,
             }
         if filename is not None:
-            with open(filename, "w") as json_file:
-                json.dump(results, json_file, indent=4)
+            with open(filename, "w", encoding="utf-8") as json_file:
+                json.dump(results, json_file, indent=4, ensure_ascii=False)
         else:
             return json.dumps(results, indent=4)
 
@@ -628,8 +628,8 @@ def save(self, folder: str):
 
         json_content = self._to_json(folder_path, saved_uuid_status)
 
-        with open(folder_path / "suite.json", "w") as f:
-            json.dump(json_content, f)
+        with open(folder_path / "suite.json", "w", encoding="utf-8") as f:
+            json.dump(json_content, f, ensure_ascii=False)
 
         analytics.track("lib:test_suite:saved")
 
@@ -843,7 +843,7 @@ def _contains_test(self, test: TestFunctionMeta):
     def load(cls, folder: str) -> "Suite":
         folder_path = Path(folder)
 
-        with open(folder_path / "suite.json", "r") as f:
+        with open(folder_path / "suite.json", "r", encoding="utf-8") as f:
             suite_json = json.load(f)
 
         suite = Suite(name=suite_json.get("name", "Unnamed test suite"))

giskard/datasets/base/__init__.py

@@ -525,7 +525,7 @@ def cast_column_to_dtypes(df, column_dtypes):
     @classmethod
     def load(cls, local_path: str):
         # load metadata
-        with open(Path(local_path) / "giskard-dataset-meta.yaml", "r") as meta_f:
+        with open(Path(local_path) / "giskard-dataset-meta.yaml", "r", encoding="utf-8") as meta_f:
             meta = yaml.safe_load(meta_f)
 
         # load data
@@ -560,7 +560,7 @@ def save(self, local_path: str):
             f.write(compressed_bytes)
             original_size_bytes, compressed_size_bytes = len(uncompressed_bytes), len(compressed_bytes)
 
-            with open(Path(local_path) / "giskard-dataset-meta.yaml", "w") as meta_f:
+            with open(Path(local_path) / "giskard-dataset-meta.yaml", "w", encoding="utf-8") as meta_f:
                 yaml.dump(
                     {
                         "id": str(self.id),

giskard/models/base/wrapper.py

@@ -230,7 +230,7 @@ def save_model_postprocessing_function(self, local_path: Union[str, Path], *_arg
             cloudpickle.dump(self.model_postprocessing_function, f, protocol=pickle.DEFAULT_PROTOCOL)
 
     def save_wrapper_meta(self, local_path, *_args, **_kwargs):
-        with open(Path(local_path) / "giskard-model-wrapper-meta.yaml", "w") as f:
+        with open(Path(local_path) / "giskard-model-wrapper-meta.yaml", "w", encoding="utf-8") as f:
             yaml.dump(
                 {
                     "batch_size": self.batch_size,
@@ -313,7 +313,7 @@ def load_model_postprocessing_function(cls, local_path: Union[str, Path], *_args
     def load_wrapper_meta(cls, local_dir, *args, **kwargs):
         wrapper_meta_file = Path(local_dir) / "giskard-model-wrapper-meta.yaml"
         if wrapper_meta_file.exists():
-            with open(wrapper_meta_file) as f:
+            with open(wrapper_meta_file, encoding="utf-8") as f:
                 wrapper_meta = yaml.load(f, Loader=yaml.Loader)
                 wrapper_meta["batch_size"] = int(wrapper_meta["batch_size"]) if wrapper_meta["batch_size"] else None
                 return wrapper_meta

giskard/models/huggingface.py

@@ -92,6 +92,7 @@ class explicitly using :class:`giskard.models.huggingface.HuggingFaceModel`.
 the `model_postprocessing_function` argument. This function should take the
 raw output of your model and return a numpy array of probabilities.
 """
+
 from typing import Any, Callable, Iterable, Optional, Tuple, Union
 
 import logging
@@ -199,7 +200,7 @@ def __init__(
     def load_model(cls, local_path, model_py_ver: Optional[Tuple[str, str, str]] = None, *args, **kwargs):
         huggingface_meta_file = Path(local_path) / "giskard-model-huggingface-meta.yaml"
         if huggingface_meta_file.exists():
-            with open(huggingface_meta_file) as f:
+            with open(huggingface_meta_file, encoding="utf-8") as f:
                 huggingface_meta = yaml.load(f, Loader=yaml.Loader)
 
         if huggingface_meta["pipeline_task"]:
@@ -208,7 +209,7 @@ def load_model(cls, local_path, model_py_ver: Optional[Tuple[str, str, str]] = N
         return huggingface_meta["huggingface_module"].from_pretrained(local_path)
 
     def save_huggingface_meta(self, local_path, *args, **kwargs):
-        with open(Path(local_path) / "giskard-model-huggingface-meta.yaml", "w") as f:
+        with open(Path(local_path) / "giskard-model-huggingface-meta.yaml", "w", encoding="utf-8") as f:
             yaml.dump(
                 {
                     "huggingface_module": self.huggingface_module,

giskard/models/pytorch.py

@@ -200,7 +200,7 @@ def _convert_to_numpy(self, raw_predictions):
         return super()._convert_to_numpy(raw_predictions)
 
     def save_pytorch_meta(self, local_path, *_args, **_kwargs):
-        with open(Path(local_path) / "giskard-model-pytorch-meta.yaml", "w") as f:
+        with open(Path(local_path) / "giskard-model-pytorch-meta.yaml", "w", encoding="utf-8") as f:
             yaml.dump(
                 {
                     "device": self.device,
@@ -224,7 +224,7 @@ def load(cls, local_dir, model_py_ver: Optional[Tuple[str, str, str]] = None, *a
     def load_pytorch_meta(cls, local_dir):
         pytorch_meta_file = Path(local_dir) / "giskard-model-pytorch-meta.yaml"
         if pytorch_meta_file.exists():
-            with open(pytorch_meta_file) as f:
+            with open(pytorch_meta_file, encoding="utf-8") as f:
                 pytorch_meta = yaml.load(f, Loader=yaml.Loader)
                 pytorch_meta["device"] = pytorch_meta.get("device")
                 pytorch_meta["torch_dtype"] = pytorch_meta.get("torch_dtype")

giskard/rag/report.py

@@ -193,11 +193,11 @@ def load(
             The embedding model to use inside the knowledge base. If not provided, the default model will be used.
         """
         path = Path(folder_path)
-        knowledge_base_meta = json.load(open(path / "knowledge_base_meta.json", "r"))
+        knowledge_base_meta = json.load(open(path / "knowledge_base_meta.json", "r", encoding="utf-8"))
         knowledge_base_data = pd.read_json(path / "knowledge_base.jsonl", orient="records", lines=True)
         testset = QATestset.load(path / "testset.jsonl")
 
-        answers = json.load(open(path / "agent_answer.json", "r"))
+        answers = json.load(open(path / "agent_answer.json", "r", encoding="utf-8"))
         model_outputs = [AgentAnswer(**answer) for answer in answers]
 
         topics = {int(k): topic for k, topic in knowledge_base_meta.pop("topics", None).items()}
@@ -219,9 +219,9 @@ def load(
 
         metrics_results = {}
         if (path / "metrics_results.json").exists():
-            metrics_results = json.load(open(path / "metrics_results.json", "r"))
+            metrics_results = json.load(open(path / "metrics_results.json", "r", encoding="utf-8"))
 
-        report_details = json.load(open(path / "report_details.json", "r"))
+        report_details = json.load(open(path / "report_details.json", "r", encoding="utf-8"))
         testset._dataframe.index = testset._dataframe.index.astype(str)
 
         report = cls(testset, model_outputs, metrics_results, knowledge_base)

giskard/registry/giskard_test.py

@@ -73,7 +73,7 @@ def _load_meta_locally(cls, local_dir, uuid: str) -> Optional[TestFunctionMeta]:
         if meta is not None:
             return meta
 
-        with open(local_dir / "meta.yaml", "r") as f:
+        with open(local_dir / "meta.yaml", "r", encoding="utf-8") as f:
             return yaml.load(f, Loader=yaml.Loader)
 
     @classmethod

giskard/scanner/report.py

@@ -90,8 +90,8 @@ def to_json(self, filename=None):
                     results[issue.detector_name][issue.level] = []
                 results[issue.detector_name][issue.level].append(issue.description)
         if filename is not None:
-            with open(filename, "w") as json_file:
-                json.dump(results, json_file, indent=4)
+            with open(filename, "w", encoding="utf-8") as json_file:
+                json.dump(results, json_file, indent=4, ensure_ascii=False)
         else:
             return json.dumps(results, indent=4)
 
@@ -115,7 +115,7 @@ def to_html(self, filename=None, embed=False):
             html = widget.render_html(embed=embed)
 
         if filename is not None:
-            with open(filename, "w") as f:
+            with open(filename, "w", encoding="utf-8") as f:
                 f.write(html)
             return
 
@@ -139,7 +139,7 @@ def to_markdown(self, filename=None, template="summary"):
         markdown = widget.render_markdown(template=template)
 
         if filename is not None:
-            with open(filename, "w") as f:
+            with open(filename, "w", encoding="utf-8") as f:
                 f.write(markdown)
             return
 
@@ -349,7 +349,7 @@ def to_avid(self, filename=None):
         ]
 
         if filename is not None:
-            with open(filename, "w") as f, warnings.catch_warnings():
+            with open(filename, "w", encoding="utf-8") as f, warnings.catch_warnings():
                 warnings.filterwarnings("ignore", category=DeprecationWarning)  # we need to support both pydantic 1 & 2
                 f.writelines(r.json() + "\n" for r in reports)
             return
@@ -373,7 +373,7 @@ def generate_rails(self, filename=None, colang_version="1.0"):
         _rails = generate_rails_from_scan_report(self, colang_version=colang_version)
 
         if filename:
-            with open(filename, "a") as f:
+            with open(filename, "a", encoding="utf-8") as f:
                 f.write(_rails)
             return
 

giskard/scanner/robustness/text_transformations.py

@@ -341,7 +341,7 @@ class TextNationalityTransformation(TextLanguageBasedTransformation):
     name = "Switch countries from high- to low-income and vice versa"
 
     def _load_dictionaries(self):
-        with Path(__file__).parent.joinpath("nationalities.json").open("r") as f:
+        with Path(__file__).parent.joinpath("nationalities.json").open("r", encoding="utf-8") as f:
             nationalities_dict = json.load(f)
         self._lang_dictionary = {"en": nationalities_dict["en"], "fr": nationalities_dict["fr"]}
 

giskard/visualization/widget.py

@@ -98,9 +98,11 @@ def render_html(self, template="full", embed=False) -> str:
             escaped = escape(html)
             uid = id(self)
 
-            with Path(__file__).parent.joinpath("templates", "scan_report", "html", "static", "external.js").open(
-                "r"
-            ) as f:
+            with (
+                Path(__file__)
+                .parent.joinpath("templates", "scan_report", "html", "static", "external.js")
+                .open("r", encoding="utf-8") as f
+            ):
                 js_lib = f.read()
 
             html = f"""<iframe id="scan-{uid}" srcdoc="{escaped}" style="width: 100%; border: none;" class="gsk-scan"></iframe>

tests/fixtures/enron_multilabel_classification.py

@@ -56,7 +56,7 @@
 
 # get_labels returns a dictionary representation of these labels.
 def get_labels(filename):
-    with open(filename + ".cats") as f:
+    with open(filename + ".cats", encoding="utf-8") as f:
         labels = defaultdict(dict)
         line = f.readline()
         while line:
@@ -99,7 +99,7 @@ def enron_raw_data_full() -> pd.DataFrame:
 
         # Features are metadata from the email object
         filename = email_file + ".txt"
-        with open(filename) as f:
+        with open(filename, encoding="utf-8") as f:
             message = email.message_from_string(f.read())
 
             values_to_add["Subject"] = str(message["Subject"])

tests/integrations/test_avid.py

@@ -76,7 +76,7 @@ def test_scan_report_can_be_exported_to_avid():
         dest_path = Path(tmpdir).joinpath("test_report.avid")
         report.to_avid(dest_path)
 
-        with dest_path.open("r") as f:
+        with dest_path.open("r", encoding="utf-8") as f:
             avid_reports_read = [json.loads(line) for line in f.readlines()]
 
     assert len(avid_reports_read) == len(avid_reports)

tests/integrations/test_nemoguardrails.py

@@ -1,22 +1,39 @@
 import json
+import tempfile
+from pathlib import Path
 from unittest.mock import Mock, patch
 
 import pandas as pd
+import pytest
 from nemoguardrails.colang import parse_colang_file
 
 from giskard.llm.client.base import ChatMessage
 from giskard.scanner.issues import Issue, Robustness
 from giskard.scanner.report import ScanReport
 
 
+def _generate_rails(report: ScanReport, filename=None, colang_version="1.0"):
+    if filename:
+        with tempfile.TemporaryDirectory() as tmpdir:
+            dest = Path(tmpdir).joinpath("rails.co")
+            report.generate_rails(filename=dest, colang_version=colang_version)
+            assert dest.exists()
+            assert dest.is_file()
+            rails = dest.read_text(encoding="utf-8")
+    else:
+        rails = report.generate_rails(colang_version=colang_version)
+    return rails
+
+
+@pytest.mark.parametrize("filename", [(None), ("rails.co")])
 @patch("giskard.integrations.nemoguardrails.get_default_client")
-def test_generate_colang_v1_rails_from_scan(get_default_client_mock):
+def test_generate_colang_v1_rails_from_scan(get_default_client_mock, filename):
     report = make_test_report()
 
     llm_client = get_default_client_mock()
     llm_client.complete.side_effect = make_llm_answers()
 
-    rails = report.generate_rails()
+    rails = _generate_rails(report, filename=filename, colang_version="1.0")
 
     # Check that the file is correctly formatted
     parsed = parse_colang_file("rails.co", rails, version="1.0")
@@ -27,14 +44,15 @@ def test_generate_colang_v1_rails_from_scan(get_default_client_mock):
     assert parsed["flows"][1]["id"] == "ask help on illegal activities"
 
 
+@pytest.mark.parametrize("filename", [(None), ("rails.co")])
 @patch("giskard.integrations.nemoguardrails.get_default_client")
-def test_generate_colang_v2_rails_from_scan(get_default_client_mock):
+def test_generate_colang_v2_rails_from_scan(get_default_client_mock, filename):
     report = make_test_report()
 
     llm_client = get_default_client_mock()
     llm_client.complete.side_effect = make_llm_answers()
 
-    rails = report.generate_rails(colang_version="2.x")
+    rails = _generate_rails(report, filename=filename, colang_version="2.x")
 
     # Check that the file is correctly formatted
     parsed = parse_colang_file("rails.co", rails, version="2.x")

tests/registry/module_utils.py

@@ -23,7 +23,7 @@ class PythonModule:
 
 def _write_file(dir: Path, file: Union[str, Path], content: str):
     os.makedirs(os.path.dirname(dir / file), exist_ok=True)
-    with open(dir / file, "w") as f:
+    with open(dir / file, "w", encoding="utf-8") as f:
         f.write(content)
 
 

tests/scan/test_scan_report.py

@@ -50,7 +50,7 @@ def test_scan_report_exports_to_markdown():
 
         assert dest.exists()
         assert dest.is_file()
-        assert dest.read_text() == markdown
+        assert dest.read_text(encoding="utf-8") == markdown
 
 
 def test_scan_report_to_json():

tests/scan/test_scanner.py

@@ -1,6 +1,8 @@
 import re
 import sys
+import tempfile
 import warnings
+from pathlib import Path
 from unittest import mock
 
 import numpy as np
@@ -279,3 +281,35 @@ def test_min_slice_size(titanic_model, titanic_dataset):
     detector = SpuriousCorrelationDetector(min_slice_size=2000)
     issues = detector.run(titanic_model, titanic_dataset, features=titanic_model.feature_names)
     assert len(issues) == 0
+
+
+@pytest.mark.parametrize(
+    "filename",
+    [(None), ("scan_test_suite_results.json")],
+)
+@pytest.mark.slow
+def test_export_scan_test_suite_results_to_json(filename, request):
+    DATASET_NAME = "diabetes_dataset_with_target"
+    MODEL_NAME = "linear_regression_diabetes"
+
+    dataset = request.getfixturevalue(DATASET_NAME)
+    model = request.getfixturevalue(MODEL_NAME)
+
+    scanner = Scanner()
+    scan_results = scanner.analyze(model, dataset)
+    test_suite_results = scan_results.generate_test_suite().run()
+
+    if filename:
+        with tempfile.TemporaryDirectory() as tmpdir:
+            dest = Path(tmpdir).joinpath(filename)
+            test_suite_results.to_json(dest)
+            assert dest.exists()
+            assert dest.is_file()
+            test_results_json = dest.read_text(encoding="utf-8")
+
+    else:
+        test_results_json = test_suite_results.to_json()
+        assert test_results_json is not None
+
+    assert test_results_json.startswith("{")
+    assert test_results_json.strip().endswith("}")