[WIP] Fix #13: fix type checking

.github/workflows/static.yml

@@ -44,7 +44,7 @@ jobs:
           # treat Usage error, Warning, Error, Fatal as failure.
           # see bit-encoded exit codes of pylint for detail:
           # https://pylint.pycqa.org/en/latest/user_guide/run.html#exit-codes
-  mypy:
+  pyre:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v2
@@ -56,5 +56,40 @@ jobs:
           curl -sSL https://raw.githubusercontent.com/python-poetry/poetry/master/install-poetry.py | python
           echo "$HOME/.local/bin" >> $GITHUB_PATH
           poetry install --no-interaction
-      - name: mypy
-        run: poetry run mypy src tests
+      - name: install pyre
+        run: poetry add -D pyre-check
+      - name: pyre
+        run: poetry run pyre --source-directory src check
+  pyright:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v2
+      - uses: actions/setup-python@v2
+        with:
+          python-version: "3.8"
+      - name: install dependencies
+        run: |
+          curl -sSL https://raw.githubusercontent.com/python-poetry/poetry/master/install-poetry.py | python
+          echo "$HOME/.local/bin" >> $GITHUB_PATH
+          poetry install --no-interaction
+      - uses: actions/setup-node@v2
+      - name: install pyright
+        run: npm install -g pyright
+      - name: pyright
+        run: poetry run pyright src tests
+  pytype:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v2
+      - uses: actions/setup-python@v2
+        with:
+          python-version: "3.8"
+      - name: install dependencies
+        run: |
+          curl -sSL https://raw.githubusercontent.com/python-poetry/poetry/master/install-poetry.py | python
+          echo "$HOME/.local/bin" >> $GITHUB_PATH
+          poetry install --no-interaction
+      - name: install pytype
+        run: poetry add -D pytype
+      - name: pytype
+        run: poetry run pytype src tests

src/psykoda/cli/internal.py

@@ -9,7 +9,7 @@
 import warnings
 from dataclasses import dataclass
 from datetime import datetime, timedelta
-from typing import Dict, List, Optional, Tuple
+from typing import Dict, List, Optional, Tuple, cast
 
 import numpy as np
 import pandas as pd
@@ -55,7 +55,7 @@ def configure_logging(debug: bool):
     stderr_handler.addFilter(stderr_filter)
     stderr_handler.setLevel(logging.INFO)
     stderr_handler.setFormatter(logging.Formatter("%(message)s"))
-    handlers = [stderr_handler]
+    handlers: list[logging.Handler] = [stderr_handler]
 
     logfile_handler = logging.FileHandler(PATH_LOG)
     logfile_handler.setLevel(logging.DEBUG)
@@ -405,7 +405,7 @@ def main_detection_skip_or_detect(
     logger.info("outputting detection reports")
     anomaly_score = detector.compute_anomaly_score(x_test, scale=True)
     num_anomaly = min(
-        sum(anomaly_score > anomaly_detection_config.threshold.min_score),
+        np.count_nonzero(anomaly_score > anomaly_detection_config.threshold.min_score),
         anomaly_detection_config.threshold.num_anomaly,
     )
 
@@ -523,6 +523,7 @@ def report_all(path_list_stats: List[str], path_save: str):
         [], columns=["datetime_rounded", "src_ip", "subnet", "service"]
     )
     idx = 0
+    results_shaps = pd.DataFrame()
     for path in path_list_stats:
         # Load stats
         stats = utils.load_json(path)
@@ -545,7 +546,7 @@ def report_all(path_list_stats: List[str], path_save: str):
             results_pd.loc[idx] = [dt, src_ip, subnet, service]
 
             if idx == 0:
-                results_shaps = pd.DataFrame([], columns=report.columns)
+                results_shaps.columns = report.columns
             results_shaps.loc[idx] = report.loc[(dt, src_ip)]
 
             idx += 1
@@ -564,13 +565,14 @@ def report_all(path_list_stats: List[str], path_save: str):
             ret = pd.concat([ret, results_pd_group.get_group(key)])
 
         ret.round(4).to_csv(path_save, index=False)
+        num_anomaly_ipaddr = len(keys)
     else:
         # Anomaly not found
         pd.DataFrame([["no anomaly found"]]).to_csv(path_save, index=False)
+        num_anomaly_ipaddr = 0
 
     logger.info("[RESULT]", extra=to_stderr)
     logger.info("Detection summary file: %s", path_save, extra=to_stderr)
-    num_anomaly_ipaddr = len(keys) if anomaly_found else 0
     logger.info(
         "Number of unique anomaly IP addresses: %s", num_anomaly_ipaddr, extra=to_stderr
     )
@@ -719,7 +721,9 @@ def detect_per_unit(
         label_value=1,
     )
     log_labeled = labeled.factory(config.io.previous.log)[0].load_previous_log(
-        entries=known_normal.index,
+        entries=cast(pd.MultiIndex, known_normal.index),
+        # we can safely assume that known_normal.Index is MultiIndex
+        # since it is empty otherwise.
     )
     log_labeled = apply_exclude_lists(log_labeled, config.preprocess.exclude_lists)
     log_labeled = preprocess.extract_log(
@@ -784,12 +788,12 @@ def _load_log_catch(load, r):
 
 
 def load_previous(
-    config: LoadPreviousConfigItem, date_to: datetime, label_value: float
+    config: Optional[LoadPreviousConfigItem], date_to: datetime, label_value: float
 ) -> pd.Series:
     from psykoda.preprocess import round_datetime
     from psykoda.utils import DateRange
 
-    if config.list is None:
+    if config is None or config.list is None:
         return pd.Series()
 
     def date_filter(row):

src/psykoda/detection.py

@@ -454,6 +454,9 @@ def compute_anomaly_score(
         """
         # Without type annotation ": ndarray" after score, sphinx treats "score" as type.
         # some text and a blank line is needed before :shape: too.
+
+        if self.detector is None:
+            raise AttributeError("detector is not set")
         score = self.detector.predict(X)
         if not scale:
             return score
@@ -483,12 +486,15 @@ def compute_embeddings(
             :shape: (n_samples, dim_embedding)
         """
 
+        detector = self.detector
+        if detector is None:
+            raise AttributeError("detector is not set")
         if X.shape[0] == 0:
             return None
 
         encoder = tf.keras.Model(
-            inputs=self.detector.input,
-            outputs=self.detector.get_layer(LAYERNAME_ENCODER_OUTPUT).output,
+            inputs=detector.input,
+            outputs=detector.get_layer(LAYERNAME_ENCODER_OUTPUT).output,
         )
         return encoder.predict(X)
 
@@ -612,7 +618,9 @@ def detection_report(
 
     for i, sample in enumerate(shap_value_idx_sorted.index):
         shap_values = shap_value_idx_sorted.loc[sample].sort_values(ascending=False)
-        fe = ["__".join(l) for l in list(shap_values.index[:shap_top_k])]
+        fe: List[Union[int, str]] = [
+            "__".join(l) for l in list(shap_values.index[:shap_top_k])
+        ]
         value = list(shap_values.iloc[:shap_top_k])
         for k in range(shap_top_k):
             if value[k] == 0: