Wrong time in threshold segments #403 (#405)

* Add convert_pd_timestamp_to_ms function to utils

* Use datapoint time in segment if it is available

analytics/analytics/detectors/pattern_detector.py

@@ -9,6 +9,7 @@ from typing import Optional
 from detectors import Detector
 from buckets import DataBucket
 from models import ModelCache
+from utils import convert_pd_timestamp_to_ms
 
 
 logger = logging.getLogger('PATTERN_DETECTOR')
@@ -56,8 +57,7 @@ class PatternDetector(Detector):
         newCache = detected['cache']
 
         last_dataframe_time = dataframe.iloc[-1]['timestamp']
-        # TODO: convert from nanoseconds to millisecond in a better way: not by dividing by 10^6
+        last_detection_time = convert_pd_timestamp_to_ms(last_dataframe_time)
-        last_detection_time = last_dataframe_time.value / 1000000
         return {
             'cache': newCache,
             'segments': segments,

analytics/analytics/detectors/threshold_detector.py

@@ -6,7 +6,7 @@ from typing import Optional
 from detectors import Detector
 from models import ModelCache
 from time import time
-from utils import convert_sec_to_ms
+from utils import convert_sec_to_ms, convert_pd_timestamp_to_ms
 
 
 logger = log.getLogger('THRESHOLD_DETECTOR')
@@ -25,44 +25,43 @@ class ThresholdDetector(Detector):
             }
         }
 
-    def detect(self, dataframe: pd.DataFrame, cache: Optional[ModelCache]) -> dict:
+    def detect(self, dataframe: pd.DataFrame, cache: ModelCache) -> dict:
+        if cache == None:
+            raise 'Threshold detector error: cannot detect before learning'
         value = cache['value']
         condition = cache['condition']
 
         now = convert_sec_to_ms(time())
-        segment = ({'from': now, 'to': now})
         segments = []
 
         dataframe_without_nans = dataframe.dropna()
         if len(dataframe_without_nans) == 0:
             if condition == 'NO_DATA':
-                segments.append(segment)
+                segments.append({ 'from': now, 'to': now })
-                return {
-                    'cache': cache,
-                    'segments': segments,
-                    'lastDetectionTime': now
-                }
             else:
                 return None
+        else:
+            last_entry = dataframe_without_nans.iloc[-1]
+            last_time = convert_pd_timestamp_to_ms(last_entry['timestamp'])
+            last_value = last_entry['value']
+            segment = { 'from': last_time, 'to': last_time }
 
-        last_entry = dataframe_without_nans.iloc[-1]
+            if condition == '>':
-        last_value = last_entry['value']
+                if last_value > value:
+                    segments.append(segment)
+            elif condition == '>=':
+                if last_value >= value:
+                    segments.append(segment)
+            elif condition == '=':
+                if last_value == value:
+                    segments.append(segment)
+            elif condition == '<=':
+                if last_value <= value:
+                    segments.append(segment)
+            elif condition == '<':
+                if last_value < value:
+                    segments.append(segment)
 
-        if condition == '>':
-            if last_value > value:
-                segments.append(segment)
-        elif condition == '>=':
-            if last_value >= value:
-                segments.append(segment)
-        elif condition == '=':
-            if last_value == value:
-                segments.append(segment)
-        elif condition == '<=':
-            if last_value <= value:
-                segments.append(segment)
-        elif condition == '<':
-            if last_value < value:
-                segments.append(segment)
         return {
             'cache': cache,
             'segments': segments,

analytics/analytics/models/model.py

@@ -86,10 +86,9 @@ class Model(ABC):
             self.state = cache
 
         result = self.do_detect(dataframe)
-        # TODO: convert from ns to ms more proper way (not dividing by 10^6)
         segments = [(
-            dataframe['timestamp'][x - 1].value / 1000000,
+            utils.convert_pd_timestamp_to_ms(dataframe['timestamp'][x - 1]),
-            dataframe['timestamp'][x + 1].value / 1000000
+            utils.convert_pd_timestamp_to_ms(dataframe['timestamp'][x + 1])
         ) for x in result]
 
         return {

analytics/analytics/utils/__init__.py

@@ -1,2 +1,3 @@
 from utils.common import *
 from utils.segments import *
+from utils.time import *

analytics/analytics/utils/common.py

@@ -326,5 +326,3 @@ def cut_dataframe(data: pd.DataFrame) -> pd.DataFrame:
 def get_min_max(array, default):
     return float(min(array, default=default)), float(max(array, default=default))
 
-def convert_sec_to_ms(sec):
-    return int(sec) * 1000

analytics/analytics/utils/time.py

@@ -0,0 +1,8 @@
+import pandas as pd
+
+def convert_sec_to_ms(sec) -> int:
+    return int(sec) * 1000
+
+def convert_pd_timestamp_to_ms(timestamp: pd.Timestamp) -> int:
+    # TODO: convert from nanoseconds to millisecond in a better way: not by dividing by 10^6
+    return int(timestamp.value) / 1000000