wendelin_telecom_base: Fix KPI graphing memory usage and delay

Use ZBigArray viewing as ndarray instead of copying

bt5/wendelin_telecom_base/ExtensionTemplateItem/portal_components/extension.erp5.WendelinTelecomGetOrsKpi.py

@@ -17,21 +17,21 @@ NA_VALUES_REPLACEMENTS = {
   'ul_hi': 0.
 }
 
-def resample_data_array(
-  np_data_array, time_field,
+def resample_data_zarray(
+  np_data_zarray, time_field,
   time_start, time_end, resample_size
 ):
   '''
     Resample the provided numpy data array to the provided new size
-    by averaging values in regular time intervals, if it is long enough to be downsized.
+    by averaging values in regular time intervals, if it is large enough to be downsized.
     Return a dictionary with the data arrays keyed to their respective data fields.
   '''
-  data_frame = pd.DataFrame.from_records(np_data_array)
+  data_frame = pd.DataFrame.from_records(np_data_zarray)
   data_frame[time_field] = pd.to_datetime(data_frame[time_field], unit='s')
   data_frame = data_frame.sort_values(by=time_field)
 
-  # Resample data
-  if len(np_data_array) > resample_size:
+  # Resample data if array is too large
+  if len(np_data_zarray) > resample_size:
     resample_period = '%ss' % int((time_end - time_start) / resample_size)
     data_frame = data_frame.resample(resample_period, on=time_field).mean()
     data_frame = data_frame.fillna(value=NA_VALUES_REPLACEMENTS)
@@ -56,36 +56,40 @@ def getOrsEnbKpi(
 
   try:
     data_array = portal.restrictedTraverse(data_array_url)
-    data_array = np.array(data_array.getArray())
   except KeyError:
     # Data Array does not exist
     RESPONSE.setStatus(404)
     return
   # Data Array is empty
-  if data_array.shape in [(), (0,)]:
+  data_array_shape = data_array.getArrayShape()
+  data_array_dtype = data_array.getArrayDtypeNames()
+  if data_array_shape is None \
+    or data_array_dtype is None \
+    or data_array_shape in [(), (0,)]:
     RESPONSE.write(json.dumps(dict()))
     return
 
   RESPONSE.setHeader("Content-Type", "application/json")
 
-  time_field = data_array.dtype.names[0]
+  data_zarray = data_array.getArray()[:]
+  time_field = data_array_dtype[0]
 
   if time_start is None or time_start == 'NaN':
-    time_start = data_array[time_field][0]
+    time_start = data_zarray[time_field][0]
   else:
     time_start = float(time_start)
   if time_end is None or time_end == 'NaN':
-    time_end = data_array[time_field][-1]
+    time_end = data_zarray[time_field][-1]
   else:
     time_end = float(time_end)
-  time_mask = (data_array[time_field] >= time_start) \
-    & (data_array[time_field] <= time_end)
+  time_mask = (data_zarray[time_field] >= time_start) \
+    & (data_zarray[time_field] <= time_end)
   if isinstance(time_mask, np.ndarray):
-    data_array = data_array[time_mask]
+    data_zarray = data_zarray[time_mask]
 
   if kpi_type == 'e_rab_accessibility':
-    resampled_data_dict = resample_data_array(
-      data_array,
+    resampled_data_dict = resample_data_zarray(
+      data_zarray,
       time_field,
       time_start,
       time_end,
@@ -109,16 +113,16 @@ def getOrsEnbKpi(
     ul_hi = []
 
     for qci in range(QCI_COUNT):
-      qci_data_array = data_array[qci::QCI_COUNT]
-      qci_dl_hi = qci_data_array['dl_hi']
-      qci_ul_hi = qci_data_array['ul_hi']
+      qci_data_zarray = data_zarray[qci::QCI_COUNT]
+      qci_dl_hi = qci_data_zarray['dl_hi']
+      qci_ul_hi = qci_data_zarray['ul_hi']
 
       # QCI is silent: skip it
       if (qci_dl_hi == 0.).all() and (qci_ul_hi == 0.).all():
         continue
 
-      resampled_qci_data_dict = resample_data_array(
-        qci_data_array,
+      resampled_qci_data_dict = resample_data_zarray(
+        qci_data_zarray,
         time_field,
         time_start,
         time_end,