fix(pm4py): refactoring discover_dfg_clean method

pm4py/algo/discovery/dfg/variants/clean.py

@@ -2,6 +2,7 @@
 from enum import Enum
 from typing import Optional, Dict, Any
 
+import numpy as np
 import pandas as pd
 
 from pm4py.objects.dfg.obj import DFG
@@ -15,9 +16,10 @@ class Parameters(Enum):
     TIMESTAMP_KEY = constants.PARAMETER_CONSTANT_TIMESTAMP_KEY
 
 
-CONST_AUX_ACT = 'aux_act_'
-CONST_AUX_CASE = 'aux_case_'
-CONST_COUNT = 'count_'
+CONST_AUX_ACT_START = 'aux_act_start'
+CONST_PROCESS_START = '#!$#PROCESS_START#!$#'
+CONST_AUX_ACT_END = 'aux_act_end'
+CONST_PROCESS_END = '#!$#PROCESS_END#!$#'
 
 
 def apply(log: pd.DataFrame, parameters: Optional[Dict[str, Any]] = None) -> DFG:
@@ -28,30 +30,28 @@ def apply(log: pd.DataFrame, parameters: Optional[Dict[str, Any]] = None) -> DFG
         Parameters.CASE_ID_KEY, parameters, constants.CASE_ATTRIBUTE_GLUE)
     time_key = exec_utils.get_param_value(
         Parameters.TIMESTAMP_KEY, parameters, xes_util.DEFAULT_TIMESTAMP_KEY)
-    aux_act = CONST_AUX_ACT + str(time.time())
-    aux_case = CONST_AUX_CASE + str(time.time())
-    count = CONST_COUNT + str(time.time())
+
     df = log.sort_values([cid_key, time_key]).loc[:, [cid_key, act_key]].reset_index()
-    df[aux_act] = df[act_key].shift(-1)
-    df[aux_case] = df[cid_key].shift(-1)
-    dfg = DFG()
 
-    excl_starter = df.at[0, act_key]
-    borders = df[(df[cid_key] != df[aux_case])]
-    starters = borders.groupby([aux_act]).size().reset_index()
-    starters.columns = [aux_act, count]
-    starters.loc[starters[aux_act] == excl_starter, count] += 1
+    aux_act_start = CONST_AUX_ACT_START + str(time.time())
+    aux_act_end = CONST_AUX_ACT_END + str(time.time())
+    df[aux_act_start] = df.groupby(cid_key)[act_key].shift(1).replace(np.nan, CONST_PROCESS_START)
+    df[aux_act_end] = df.groupby(cid_key)[act_key].shift(-1).replace(np.nan, CONST_PROCESS_END)
+
+    starters = df[(df[aux_act_start] == CONST_PROCESS_START)]
+    borders = df[(df[aux_act_end] == CONST_PROCESS_END)]
+    connections = df[((df[aux_act_start] != CONST_PROCESS_START) & (df[aux_act_end] != CONST_PROCESS_END))]
+
+    dfg = DFG()
 
-    df = df[(df[cid_key] == df[aux_case])]
-    for (a, b, f) in list(df.groupby([act_key, aux_act]).size(
-    ).reset_index().itertuples(index=False, name=None)):
+    for (a, b, f) in list(
+            connections.groupby([act_key, aux_act_end]).size().reset_index().itertuples(index=False, name=None)):
         dfg.graph[(a, b)] += f
 
-    for (a, f) in list(starters.itertuples(index=False, name=None)):
+    for (a, f) in list(starters.groupby([act_key]).size().reset_index().itertuples(index=False, name=None)):
         dfg.start_activities[a] += f
 
-    for (a, f) in list(borders.groupby(
-            [act_key]).size().reset_index().itertuples(index=False, name=None)):
+    for (a, f) in list(borders.groupby([act_key]).size().reset_index().itertuples(index=False, name=None)):
         dfg.end_activities[a] += f
 
     return dfg