From 9c494bac60061fbd8528d7a7518edcb4ae7464a0 Mon Sep 17 00:00:00 2001
From: "Jip J. Dekker" <jip@dekker.one>
Date: Fri, 9 Jul 2021 14:07:26 +1000
Subject: [PATCH] Update analysis scripts

---
 analyse_chuffed.py | 162 ++++++++++++++++++++++++++++++---------------
 analyse_gecode.py  |   1 -
 2 files changed, 107 insertions(+), 56 deletions(-)

diff --git a/analyse_chuffed.py b/analyse_chuffed.py
index cda80c8..551a3b2 100755
--- a/analyse_chuffed.py
+++ b/analyse_chuffed.py
@@ -13,6 +13,8 @@ import os
 import re
 import sys
 
+from statistics import stdev
+
 
 def compute_area(file, time):
     area = -1
@@ -30,74 +32,124 @@ def compute_area(file, time):
         if match:
             objectives.append(int(match.group(1)))
             continue
-        match = re.match(r"%\stime elapsed:\s(\d+)\sms", line)
+        match = re.match(r"%\stime elapsed:\s(\d+\.\d+)\ss", line)
         if match:
-            times.append(int(match.group(1)))
+            times.append(float(match.group(1)))
             continue
-    times.append(time)
+    # Not proven optimal
+    if len(times) == len(objectives):
+        times.append(time)
 
     assert len(objectives) > 0
     assert len(objectives) + 1 == len(times)
     area = 0
     for i in range(len(objectives)):
-        area += ((times[i + 1] - times[i]) / 1000) * objectives[i]
+        area += (times[i + 1] - times[i]) * objectives[i]
     return int(area)
 
 
 folder = sys.argv[1]
-stats = {}
-for root, dirs, files in os.walk(folder):
-    for name in files:
-        if name.endswith(".sol"):
-            seed = 1
-            match = re.search(r"\.(\d+)\.sol", name)
-            if match:
-                seed = int(match.group(1))
-            with open(os.path.join(root, name)) as f:
-                contents = f.readlines()
+statistics = dict()
+instances = set()
+for config in ["original", "restart"]:
+    for root, dirs, files in os.walk(folder + "/" + config):
+        for name in files:
+            if name.endswith(".sol"):
+                components = name[:-(4)].split(".")
+                data = components[0]
+                instances.add(data)
+                seed = 1
+                if len(components) > 1:
+                    assert len(components) == 2
+                    seed = components[1]
 
-            statistics = {}
-            print(contents)
-            for line in contents:
-                # Nodes
-                match = re.search(r"nodes=(\d+)", line)
-                if match:
-                    statistics["nodes"] = int(match.group(1))
-                    continue
-                # Solve time
-                match = re.search(r"solveTime=(\d+\.\d+)", line)
-                if match:
-                    statistics["search_time"] = int(float(match.group(1)) * 1000)
-                    continue
-                # Restarts
-                match = re.search(r"restart count:\s+(\d+)", line)
-                if match:
-                    statistics["restarts"] = int(match.group(1))
-                    continue
+                if data not in statistics:
+                    statistics[data] = dict()
+                if config not in statistics[data]:
+                    statistics[data][config] = []
 
-            for line in contents[::-1]:
-                # Best objective
-                match = re.match(r"objective\s=\s(\d+)", line)
-                if match:
-                    statistics["objective"] = int(match.group(1))
-                    break
-            # Area
-            area = compute_area(contents, statistics["search_time"])
+                with open(os.path.join(root, name)) as f:
+                    contents = f.readlines()
 
-            stats[name[:-(4)].replace(".", ",")] = (
-                area,
-                statistics["objective"],
-                statistics["search_time"],
-                statistics["restarts"],
-                statistics["nodes"],
+                nodes = None
+                solvetime = None
+                restarts = None
+                objective = None
+                for line in contents:
+                    # Nodes
+                    match = re.search(r"%%%mzn-stat: nodes=(\d+)", line)
+                    if match:
+                        nodes = int(match.group(1))
+                        continue
+                    # Solve time
+                    match = re.search(r"%%%mzn-stat: solveTime=(\d+\.\d+)", line)
+                    if match:
+                        solvetime = float(match.group(1))
+                        continue
+                    # Restarts
+                    match = re.search(r"%%%mzn-stat: restarts=(\d+)", line)
+                    if match:
+                        restarts = int(match.group(1))
+                        continue
+
+                for line in contents[::-1]:
+                    # Best objective
+                    match = re.match(r"%%%mzn-stat: objective=(-?\d+)", line)
+                    if match:
+                        objective = int(match.group(1))
+                        break
+                # Area
+                area = compute_area(contents, solvetime)
+
+                statistics[data][config].append(
+                    (
+                        area,
+                        objective,
+                        solvetime,
+                        restarts,
+                        nodes,
+                    )
+                )
+
+for data in instances:
+    for config in ["original", "restart"]:
+        stats = statistics[data][config]
+        cumulative = stats[0]
+        for i in range(1, len(stats)):
+            cumulative = tuple(map(sum, zip(cumulative, stats[i])))
+        avg = tuple([x / len(stats) for x in cumulative])
+        dev = stdev([x[1] for x in stats]) if len(stats) > 1 else 0
+        # (avg area, avg objective, stdev objective)
+        statistics[data][config] = (avg[0], avg[1], dev)
+
+# Print header
+print(
+    """
+\\begin{tabular}{l|rr|rr|rr}
+\\toprule
+& \multicolumn{2}{c|}{Chuffed} & \multicolumn{2}{c|}{Chuffed Restart} \\\\
+Instance & $\intobj$ & $\minobj$ & $\intobj$ & $\minobj$ \\\\
+\midrule
+"""
+)
+
+sorted_instances = sorted(instances)
+for data in sorted_instances:
+    print(f"{data}", end="")
+    for config in ["original", "restart"]:
+        print(
+            f" & {int(statistics[data][config][0] / 1000) }k & {int(statistics[data][config][1])}",
+            end="",
+        )
+        if statistics[data][config][2] != 0:
+            print("^{", end="")
+            print(
+                int(statistics[data][config][2] / statistics[data][config][1] * 100),
+                # int(statistics[data][config][2]),
+                end="",
             )
+            print("}", end="")
+    print(" \\\\")
 
-sorted_stats = sorted(stats.items())
-a = sorted_stats[0][0][: sorted_stats[0][0].find(",")]
-for key, val in sorted_stats:
-    if key[: key.find(",")] != a:
-        print("\n\n")
-        a = key[: key.find(",")]
-    print("%s,%s" % (key, ",".join([v.__str__() for v in val])))
-
-exit(1)
+# Print footer
+print("\n\\bottomrule\n\end{tabular}")
diff --git a/analyse_gecode.py b/analyse_gecode.py
index 90fb8ce..99c6c53 100755
--- a/analyse_gecode.py
+++ b/analyse_gecode.py
@@ -8,7 +8,6 @@ area which can be found by adding
 'constraint trace("% init_area = \(ub(objective));\n", true);'
 to the model in question.
 """
-import csv
 import os
 import re
 import sys