add testing script

bash/init

@@ -0,0 +1,19 @@
+#!/bin/bash
+DIR=$(dirname "${BASH_SOURCE[0]}")
+
+datasets="SN_gender citeseer cora dblp mit nyu pubmed stanford uIllinois"
+methods="deepwalk node2vec line grarep abrw attrpure attrcomb tadw aane sagemean sagegcn asne"
+
+for dataset in $datasets; do
+    echo cd \`dirname \"\${BASH_SOURCE[0]}\"\`/.. > $DIR/$dataset.sh
+    for method in $methods; do
+        echo python -u src/main.py --method $method --graph-file data/$dataset/${dataset}_adjlist.txt --attribute-file data/$dataset/${dataset}_attr.txt --label-file data/$dataset/${dataset}_label.txt --emb-file emb/${dataset}_${method}_emb --save-emb 2\>\&1 \| tee log/$dataset-$method-\`date +%Y%m%d-%H%M%S\`.log >> $DIR/$dataset.sh
+    done
+done
+
+echo DIR=\$\(dirname \"\${BASH_SOURCE[0]}\"\) > $DIR/runall.sh
+for dataset in $datasets; do
+    echo bash \$DIR/${dataset}.sh >> $DIR/runall.sh
+done
+
+chmod a+x $DIR/*.sh

log/parse.py

@@ -0,0 +1,37 @@
+#!/usr/bin/env python
+# coding: utf-8
+
+import os
+import re
+
+import numpy as np
+import pandas as pd
+
+DIR = os.path.dirname(os.path.realpath(__file__))
+
+data = pd.DataFrame(columns=["Samples", "AUC", "Micro-F1", "Macro-F1", "Time"], dtype="float")
+data.set_index(pd.MultiIndex.from_tuples((), names=("dataset", "method")), inplace=True)
+for fname in os.listdir(DIR):
+    if fname.endswith(".log"):
+        try:
+            name, method = fname.split(".")[0].split('-')[0:2]
+            with open(os.path.join(DIR, fname)) as f:
+                l = np.array(
+                    re.findall(
+                        r"STEP3: end learning embeddings; time cost: (\d+\.\d+)s.*roc= (\d+\.\d+).*{'micro': (\d+\.\d+), 'macro': (\d+\.\d+), 'samples': \d+\.\d+, 'weighted': \d+\.\d+}",
+                        f.read(),
+                        re.DOTALL
+                    )[-1],
+                    dtype="float"
+                )[[1,2,3,0]]
+                if (name, method) not in data.index:
+                    data.loc[(name, method), :] = 0
+                n = data.loc[(name, method)][0]
+                l = (n * np.array(data.loc[(name, method)][1:5]) + l)/(n + 1)
+                data.loc[(name, method)][0] += 1
+                data.loc[(name, method), 1:5] = l
+        except Exception as e:
+            print(f"failed with {fname}: {e}")
+data.sort_index(inplace=True)
+
+data.to_csv(os.path.join(DIR, "result.csv"))