wip

analyze_batch.py

@@ -1,13 +1,15 @@
+import operator
 import os
 import sys
 from collections import defaultdict
 from datetime import timedelta
+from math import ceil
 
 import matplotlib.pyplot as plt
 import numpy as np
 from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer
 
-from common import calc_intervals
+from common import calc_intervals, imprt, FigSaver
 from loader import load, dmt, cms
 
 printnoln = lambda text: print(text, end='', flush=True)
@@ -15,8 +17,10 @@ rprint = lambda text: print('\r' + text)
 
 DAYS_NEW_USER = 7
 OLD_USER_YEAR = 3
+OLD_USER_PERCENTILE = 0.95
 
 analyser = SentimentIntensityAnalyzer()
+figsaver = FigSaver()
 colors = ['red', 'green', 'blue', 'orange', 'deeppink']
 
 
@@ -24,7 +28,10 @@ def main(folder):
     users, posts, firstcontrib, sumcontrib = load(folder)
 
     intervals = calc_intervals(posts)
-    cachedsentiments = {}
+    cachedsentiments = imprt(folder + "/output/sentiments.py").answers
+
+    outfolder = folder + "/output/batch/"
+    os.system("mkdir -p " + outfolder)
 
     postcounts = range(1, 5 + 1)
     for (option_date_from, option_date_to) in intervals:
@@ -44,8 +51,6 @@ def main(folder):
         gpos = []
         gcom = []
 
-        outfolder = folder + "/output/batch/"
-        os.system("mkdir -p " + outfolder)
         goutfilenamenewusers = outfolder + "batch_newusers_" + folder.split("/")[-1] + "_" + option_date_from.strftime("%d-%m-%Y") + "_" + option_date_to.strftime("%d-%m-%Y")
         goutfilenameoldusers = outfolder + "batch_oldusers_" + folder.split("/")[-1] + "_" + option_date_from.strftime("%d-%m-%Y") + "_" + option_date_to.strftime("%d-%m-%Y")
 
@@ -55,7 +60,7 @@ def main(folder):
             # computer toxic levels
             start = cms()
             printnoln("computing toxic levels: filtering")
-            toxlevels = defaultdict(list)
+            toxlevels = []
             searchedposts = defaultdict(int)
             filteredposts = []
             for (i, post) in enumerate(newposts):
@@ -73,26 +78,22 @@ def main(folder):
                 filteredposts.append(post)
 
             for (i, post) in enumerate(filteredposts):
-                if (i + 1) % 100 == 0:
-                    printnoln("\rcomputing toxic levels: post #" + str(i + 1) + "/" + str(len(filteredposts)))
-                if (i + 1) == len(newposts):
-                    printnoln("\rcomputing toxic levels: post #" + str(i + 1) + "/" + str(len(filteredposts)))
+                printnoln("\rcomputing toxic levels: post " + str(i + 1) + "/" + str(len(filteredposts)))
                 for a in post['Answers']:
                     if a['Id'] in cachedsentiments.keys():
                         toxlevel = cachedsentiments[a['Id']]
                     else:
-                        toxlevel = computeToxLevel(a['Body'])
-                        cachedsentiments[a['Id']] = toxlevel
-                    toxlevels[post['Id']].append(toxlevel)
-            rprint("computing toxic levels: post #" + str(len(filteredposts)) + "/" + str(len(filteredposts)) + " ... took " + str(cms() - start) + "ms")
+                        print("Sentiment not found for " + a['Id'])
+                    toxlevels.append(toxlevel)
+            printnoln("\rcomputing toxic levels: post " + str(len(filteredposts)) + "/" + str(len(filteredposts)) + " ... plotting ...")
 
             outfilename = goutfilenamenewusers + "_" + str(option_posts)
             dumptoxlevels(toxlevels, outfilename + ".py")
 
-            neglevelsflat = [item['neg'] for item in flatmap(toxlevels.values())]
-            neulevelsflat = [item['neu'] for item in flatmap(toxlevels.values())]
-            poslevelsflat = [item['pos'] for item in flatmap(toxlevels.values())]
-            comlevelsflat = [item['compound'] for item in flatmap(toxlevels.values())]
+            neglevelsflat = [item['neg'] for item in toxlevels]
+            neulevelsflat = [item['neu'] for item in toxlevels]
+            poslevelsflat = [item['pos'] for item in toxlevels]
+            comlevelsflat = [item['compound'] for item in toxlevels]
 
             gneg.append(neglevelsflat)
             gneu.append(neulevelsflat)
@@ -116,10 +117,15 @@ def main(folder):
             # plt.show()
             fig.suptitle("Sentiment of answers to the first " + str(option_posts) + " (max) posts within 1 week of 1st contribution\nPosts created between "
                          + option_date_from.strftime("%d-%m-%Y") + " to " + option_date_to.strftime("%d-%m-%Y"))
+            # figsaver.save(fig, outfilename + ".png", bbox_inches='tight')
+            printnoln("\rcomputing toxic levels: post " + str(len(filteredposts)) + "/" + str(len(filteredposts)) + " ... plotting ... saving ...")
             fig.savefig(outfilename + ".png", bbox_inches='tight')
             plt.close(fig)
+            rprint("computing toxic levels: post " + str(len(filteredposts)) + "/" + str(len(filteredposts)) + " ... plotting ... saving ... took " + str(cms() - start) + "ms")
 
         # global
+        start = cms()
+        printnoln("\rglobal plot post ... plotting ...")
         gaxs[0, 0].hist(gneg, np.linspace(0, 1, 1 * 100), color=colors[:len(postcounts)], label=[str(option_posts) + " posts" for option_posts in postcounts])
         gaxs[1, 0].hist(gneu, np.linspace(0, 1, 1 * 100), color=colors[:len(postcounts)], label=[str(option_posts) + " posts" for option_posts in postcounts])
         gaxs[0, 1].hist(gpos, np.linspace(0, 1, 1 * 100), color=colors[:len(postcounts)], label=[str(option_posts) + " posts" for option_posts in postcounts])
@@ -132,19 +138,74 @@ def main(folder):
         gaxs[1, 0].set_yscale('log')
         gaxs[0, 1].set_yscale('log')
         gaxs[1, 1].set_yscale('log')
-        gfig.suptitle("Sentiment of answers to the first X (max) posts within 1 week of 1st contribution\nPosts created between " + option_date_from.strftime("%d-%m-%Y") + " to " + option_date_to.strftime("%d-%m-%Y"))
+        gfig.suptitle(
+            "Sentiment of answers to the first X (max) posts within 1 week of 1st contribution\nPosts created between " + option_date_from.strftime("%d-%m-%Y") + " to " + option_date_to.strftime(
+                "%d-%m-%Y"))
+        # figsaver.save(gfig, goutfilenamenewusers + ".png", bbox_inches='tight')
+        printnoln("\rglobal plot post ... plotting ... saving ...")
         gfig.savefig(goutfilenamenewusers + ".png", bbox_inches='tight')
         plt.close(gfig)
+        rprint("global plot post ... plotting ... saving ... took " + str(cms() - start) + "ms")
+
+        # for old users ---------------------------------------------------------------------------------
+        start = cms()
+        newuserids = set(dmt(newposts).map(lambda p: p['OwnerUserId']).getresults())
+        userposts = {u: 0 for u in newuserids}
+        for p in newposts:
+            userposts[p['OwnerUserId']] += 1
+        userposts = sorted(userposts.items(), key=operator.itemgetter(1))
+        oldusers = [k for k, v in userposts]
+        oldusers = set(oldusers[ceil(len(oldusers) * OLD_USER_PERCENTILE):])
+        filteredposts = dmt(newposts).filter(lambda p: p['OwnerUserId'] in oldusers).getresults()
+
+        toxlevels = []
+        for (i, post) in enumerate(filteredposts):
+            printnoln("\rcomputing toxic levels: post " + str(i + 1) + "/" + str(len(filteredposts)))
+            for a in post['Answers']:
+                if a['Id'] in cachedsentiments.keys():
+                    toxlevel = cachedsentiments[a['Id']]
+                else:
+                    print("Sentiment not found for " + a['Id'])
+                toxlevels.append(toxlevel)
+        printnoln("\rcomputing toxic levels: post " + str(len(filteredposts)) + "/" + str(len(filteredposts)) + " ... plotting ...")
+
+        dumptoxlevels(toxlevels, goutfilenameoldusers + ".py")
+
+        neglevelsflat = [item['neg'] for item in toxlevels]
+        neulevelsflat = [item['neu'] for item in toxlevels]
+        poslevelsflat = [item['pos'] for item in toxlevels]
+        comlevelsflat = [item['compound'] for item in toxlevels]
+
+        fig, axs = plt.subplots(2, 2, figsize=(16, 12))
+        axs[0, 0].set_title('Neg')
+        axs[1, 0].set_title('Neu')
+        axs[0, 1].set_title('Pos')
+        axs[1, 1].set_title('Compound')
+        axs[0, 0].hist(neglevelsflat, np.linspace(0, 1, 1 * 100))
+        axs[1, 0].hist(neulevelsflat, np.linspace(0, 1, 1 * 100))
+        axs[0, 1].hist(poslevelsflat, np.linspace(0, 1, 1 * 100))
+        axs[1, 1].hist(comlevelsflat, np.linspace(-1, 1, 2 * 100))
+        axs[0, 0].set_yscale('log')
+        axs[1, 0].set_yscale('log')
+        axs[0, 1].set_yscale('log')
+        axs[1, 1].set_yscale('log')
+
+        # plt.show()
+        fig.suptitle("Sentiment of answers to posts by most posting users (95%tile)\nPosts created between " + option_date_from.strftime("%d-%m-%Y") + " to " + option_date_to.strftime("%d-%m-%Y"))
+        # figsaver.save(fig, goutfilenameoldusers + ".png", bbox_inches='tight')
+        printnoln("\rcomputing toxic levels: post " + str(len(filteredposts)) + "/" + str(len(filteredposts)) + " ... plotting ... saving ...")
+        fig.savefig(goutfilenameoldusers + ".png", bbox_inches='tight')
+        plt.close(fig)
+        rprint("computing toxic levels: post " + str(len(filteredposts)) + "/" + str(len(filteredposts)) + " ... plotting ... saving ... took " + str(cms() - start) + "ms")
+
+        figsaver.join()
+    figsaver.join()
 
 
 def computeToxLevel(text):
     return analyser.polarity_scores(text)
 
 
-def flatmap(arr):
-    return [item for sublist in arr for item in sublist]
-
-
 def dumptoxlevels(lvls, filename):
     with open(filename, "w") as file:
         file.write("from collections import defaultdict\n\n")

calctoxdiff.py

@@ -1,4 +1,3 @@
-import importlib
 import os
 import sys
 from collections import defaultdict
@@ -10,27 +9,36 @@ import matplotlib.pyplot as plt
 import numpy as np
 from scipy.stats import ks_2samp
 
+from common import imprt
+
 colors = {'neg': 'red', 'neu': 'green', 'pos': 'blue', 'com': 'orange'}
+IMAGE_MAGICK = "magick"
 
 
 def main(folder):
     outputdir = folder + "/output/ksbatch/"
     os.system("mkdir -p " + outputdir)
-    folder = folder + "/output/batch/"
+    srcfolder = folder + "/output/batch/"
 
-    onlyfiles = [folder + f for f in listdir(folder) if isfile(join(folder, f)) and f.endswith(".py")]
+    onlyfiles = [srcfolder + f for f in listdir(srcfolder) if isfile(join(srcfolder, f)) and f.endswith(".py") and "newusers" in f]
     onlyfiles = sorted(onlyfiles)
 
-    plotbypost(onlyfiles, outputdir)
-    plotbydate(onlyfiles, outputdir)
+    # plotbypost(onlyfiles, outputdir)
+    # plotbydate(onlyfiles, outputdir)
+
+    oldfiles = [srcfolder + f for f in listdir(srcfolder) if isfile(join(srcfolder, f)) and f.endswith(".py") and "oldusers" in f]
+    oldfiles = sorted(oldfiles)
+
+    plotbydateold(onlyfiles, oldfiles, outputdir)
 
 
 def plotbypost(onlyfiles, outputdir):
+    print("plotbypost")
     files = defaultdict(list)
     for f in onlyfiles:
         s = f[:-3].split("_")
-        files[int(s[4])].append(f)
-    files = {p: sorted(l, key=lambda e: datetime.strptime(e.split("_")[2], "%d-%m-%Y")) for (p, l) in files.items()}
+        files[int(s[5])].append(f)
+    files = {p: sorted(l, key=lambda e: datetime.strptime(e.split("_")[3], "%d-%m-%Y")) for (p, l) in files.items()}
 
     changes_neg = defaultdict(list)
     changes_neu = defaultdict(list)
@@ -45,15 +53,15 @@ def plotbypost(onlyfiles, outputdir):
             tox1 = imprt(l[i]).toxlevels
             tox2 = imprt(l[i + 1]).toxlevels
 
-            neglevelsflat1 = [item['neg'] for item in flatmap(tox1.values())]
-            neulevelsflat1 = [item['neu'] for item in flatmap(tox1.values())]
-            poslevelsflat1 = [item['pos'] for item in flatmap(tox1.values())]
-            comlevelsflat1 = [item['compound'] for item in flatmap(tox1.values())]
+            neglevelsflat1 = [item['neg'] for item in tox1]
+            neulevelsflat1 = [item['neu'] for item in tox1]
+            poslevelsflat1 = [item['pos'] for item in tox1]
+            comlevelsflat1 = [item['compound'] for item in tox1]
 
-            neglevelsflat2 = [item['neg'] for item in flatmap(tox2.values())]
-            neulevelsflat2 = [item['neu'] for item in flatmap(tox2.values())]
-            poslevelsflat2 = [item['pos'] for item in flatmap(tox2.values())]
-            comlevelsflat2 = [item['compound'] for item in flatmap(tox2.values())]
+            neglevelsflat2 = [item['neg'] for item in tox2]
+            neulevelsflat2 = [item['neu'] for item in tox2]
+            poslevelsflat2 = [item['pos'] for item in tox2]
+            comlevelsflat2 = [item['compound'] for item in tox2]
 
             ksneg = ks_2samp(neglevelsflat1, neglevelsflat2)
             ksneu = ks_2samp(neulevelsflat1, neulevelsflat2)
@@ -74,7 +82,7 @@ def plotbypost(onlyfiles, outputdir):
                         + "; ks pos = " + str(changes_pos[p][i]) + "; ks com = " + str(changes_com[p][i]) + "\n")
     # pval
     for (p, l) in files.items():
-        x = [l[i].split("_")[2] + " -\n" + l[i + 1].split("_")[2] for i in range(len(l) - 1)]
+        x = [l[i].split("_")[3] + " -\n" + l[i + 1].split("_")[3] for i in range(len(l) - 1)]
         fig = plt.figure(figsize=(16, 12))
         for type, changes in {"neg": changes_neg[p], "neu": changes_neu[p], "pos": changes_pos[p], "com": changes_com[p]}.items():
             pval = [x.pvalue for x in changes]
@@ -92,7 +100,7 @@ def plotbypost(onlyfiles, outputdir):
 
     # stat
     for (p, l) in files.items():
-        x = [l[i].split("_")[2] + " -\n" + l[i + 1].split("_")[2] for i in range(len(l) - 1)]
+        x = [l[i].split("_")[3] + " -\n" + l[i + 1].split("_")[3] for i in range(len(l) - 1)]
         fig = plt.figure(figsize=(16, 12))
         for type, changes in {"neg": changes_neg[p], "neu": changes_neu[p], "pos": changes_pos[p], "com": changes_com[p]}.items():
             stat = [x.statistic for x in changes]
@@ -110,11 +118,12 @@ def plotbypost(onlyfiles, outputdir):
 
 
 def plotbydate(onlyfiles, outputdir):
+    print("plotbydate")
     files = defaultdict(list)
     for f in onlyfiles:
         s = f[:-3].split("_")
-        files[(s[2], s[3])].append(f)
-    files = {d: sorted(l, key=lambda e: e.split("_")[4]) for (d, l) in files.items()}
+        files[(s[3], s[4])].append(f)
+    files = {d: sorted(l, key=lambda e: e.split("_")[5]) for (d, l) in files.items()}
 
     changes_neg = defaultdict(list)
     changes_neu = defaultdict(list)
@@ -129,15 +138,15 @@ def plotbydate(onlyfiles, outputdir):
             tox1 = imprt(l[i]).toxlevels
             tox2 = imprt(l[i + 1]).toxlevels
 
-            neglevelsflat1 = [item['neg'] for item in flatmap(tox1.values())]
-            neulevelsflat1 = [item['neu'] for item in flatmap(tox1.values())]
-            poslevelsflat1 = [item['pos'] for item in flatmap(tox1.values())]
-            comlevelsflat1 = [item['compound'] for item in flatmap(tox1.values())]
+            neglevelsflat1 = [item['neg'] for item in tox1]
+            neulevelsflat1 = [item['neu'] for item in tox1]
+            poslevelsflat1 = [item['pos'] for item in tox1]
+            comlevelsflat1 = [item['compound'] for item in tox1]
 
-            neglevelsflat2 = [item['neg'] for item in flatmap(tox2.values())]
-            neulevelsflat2 = [item['neu'] for item in flatmap(tox2.values())]
-            poslevelsflat2 = [item['pos'] for item in flatmap(tox2.values())]
-            comlevelsflat2 = [item['compound'] for item in flatmap(tox2.values())]
+            neglevelsflat2 = [item['neg'] for item in tox2]
+            neulevelsflat2 = [item['neu'] for item in tox2]
+            poslevelsflat2 = [item['pos'] for item in tox2]
+            comlevelsflat2 = [item['compound'] for item in tox2]
 
             ksneg = ks_2samp(neglevelsflat1, neglevelsflat2)
             ksneu = ks_2samp(neulevelsflat1, neulevelsflat2)
@@ -159,7 +168,7 @@ def plotbydate(onlyfiles, outputdir):
 
     # pval
     for (d, l) in files.items():
-        x = [l[i].split("_")[4][:-3] + "-" + l[i + 1].split("_")[4][:-3] for i in range(len(l) - 1)]
+        x = [l[i].split("_")[5][:-3] + "-" + l[i + 1].split("_")[5][:-3] for i in range(len(l) - 1)]
         fig = plt.figure(figsize=(16, 12))
         for type, changes in {"neg": changes_neg[d], "neu": changes_neu[d], "pos": changes_pos[d], "com": changes_com[d]}.items():
             pval = [x.pvalue for x in changes]
@@ -177,7 +186,7 @@ def plotbydate(onlyfiles, outputdir):
 
     # stat
     for (d, l) in files.items():
-        x = [l[i].split("_")[4][:-3] + "-" + l[i + 1].split("_")[4][:-3] for i in range(len(l) - 1)]
+        x = [l[i].split("_")[5][:-3] + "-" + l[i + 1].split("_")[5][:-3] for i in range(len(l) - 1)]
         fig = plt.figure(figsize=(16, 12))
         for type, changes in {"neg": changes_neg[d], "neu": changes_neu[d], "pos": changes_pos[d], "com": changes_com[d]}.items():
             stat = [x.statistic for x in changes]
@@ -194,15 +203,122 @@ def plotbydate(onlyfiles, outputdir):
         plt.close(fig)
 
 
-def imprt(file):
-    spec = importlib.util.spec_from_file_location("module.name", file)
-    foo = importlib.util.module_from_spec(spec)
-    spec.loader.exec_module(foo)
-    return foo
+def plotbydateold(onlyfiles, oldfiles, outputdir):
+    print("plotbydateold")
+    files = defaultdict(list)
+    for f in onlyfiles:
+        s = f[:-3].split("_")
+        files[(s[3], s[4])].append(f)
+    dates = sorted(files.keys(), key=lambda e: "-".join(reversed(e[0].split("-"))))
+    files = {d: files[d] for d in dates}
+    files = {d: sorted(l, key=lambda e: e.split("_")[5]) for (d, l) in files.items()}
+    oldfiles = {(f[:-3].split("_")[3], f[:-3].split("_")[4]): f for f in oldfiles}
 
+    changes_neg = defaultdict(list)
+    changes_neu = defaultdict(list)
+    changes_pos = defaultdict(list)
+    changes_com = defaultdict(list)
 
-def flatmap(arr):
-    return [item for sublist in arr for item in sublist]
+    for (d, l) in files.items():
+        print(d)
+        toxold = imprt(oldfiles[d]).toxlevels
+
+        neglevelsold = [item['neg'] for item in toxold]
+        neulevelsold = [item['neu'] for item in toxold]
+        poslevelsold = [item['pos'] for item in toxold]
+        comlevelsold = [item['compound'] for item in toxold]
+
+        for i in range(len(l)):
+            tox1 = imprt(l[i]).toxlevels
+            if len(tox1) == 0 or len(toxold) == 0:
+                changes_neg[d].append(None)
+                changes_neu[d].append(None)
+                changes_pos[d].append(None)
+                changes_com[d].append(None)
+                continue
+
+            neglevelsflat1 = [item['neg'] for item in tox1]
+            neulevelsflat1 = [item['neu'] for item in tox1]
+            poslevelsflat1 = [item['pos'] for item in tox1]
+            comlevelsflat1 = [item['compound'] for item in tox1]
+
+            ksneg = ks_2samp(neglevelsflat1, neglevelsold)
+            ksneu = ks_2samp(neulevelsflat1, neulevelsold)
+            kspos = ks_2samp(poslevelsflat1, poslevelsold)
+            kscom = ks_2samp(comlevelsflat1, comlevelsold)
+
+            changes_neg[d].append(ksneg)
+            changes_neu[d].append(ksneu)
+            changes_pos[d].append(kspos)
+            changes_com[d].append(kscom)
+
+    print("logs")
+    for (d, l) in files.items():
+        # print(d)
+        # print("neg is: " + str(len(changes_neg[d])) + " should: " + str(len(l)))
+        # print("neu is: " + str(len(changes_neu[d])) + " should: " + str(len(l)))
+        # print("pos is: " + str(len(changes_pos[d])) + " should: " + str(len(l)))
+        # print("com is: " + str(len(changes_com[d])) + " should: " + str(len(l)))
+        f1 = oldfiles[d]
+        with open(outputdir + "/ks_olddate_" + d[0] + "_" + d[1] + ".log", "w") as f:
+            for i in range(len(l)):
+                if changes_neg[d][i] is None:
+                    continue
+                f2 = l[i]
+                f.write(f1 + " -> " + f2 + ": ks neg = " + str(changes_neg[d][i]) + "; ks neu = " + str(changes_neu[d][i])
+                        + "; ks pos = " + str(changes_pos[d][i]) + "; ks com = " + str(changes_com[d][i]) + "\n")
+
+    # pval
+    print("pval")
+    imgmagickcmd = IMAGE_MAGICK
+    for (d, l) in files.items():
+        print(d)
+        x = [l[i][:-3].split("_")[5] for i in range(len(l))]
+        fig = plt.figure(figsize=(16, 12))
+        for type, changes in {"neg": changes_neg[d], "neu": changes_neu[d], "pos": changes_pos[d], "com": changes_com[d]}.items():
+            pval = [(xx, c.pvalue) for xx, c in zip(x, changes) if c is not None]
+            plt.plot([p[0] for p in pval], [p[1] for p in pval], label=type + ".pval", color=colors[type])
+            if len(pval) == 0:
+                continue
+            mean = np.mean([p[1] for p in pval])
+            std = np.std([p[1] for p in pval])
+            dev = [(xx, s) for (xx, s) in pval if s <= mean - std or s >= mean + std]
+            plt.plot([p[0] for p in pval], [mean] * len(pval), color=colors[type], ls='dashed')
+            plt.plot([dx[0] for dx in dev], [dx[1] for dx in dev], color=colors[type], ls='None', marker='o')
+        plt.title("KS 2-sided test with new and old users between " + d[0] + " and " + d[1])
+        plt.xticks(rotation=90)
+        plt.legend(loc="upper right")
+        outfile = outputdir + "/ks_olddate_pval_" + d[0] + "_" + d[1] + ".png"
+        plt.savefig(outfile, bbox_inches='tight')
+        plt.close(fig)
+        imgmagickcmd += " " + outfile
+    os.system(imgmagickcmd + " " + outputdir + "/ks_olddate_pval.pdf")
+
+    # stat
+    print("stat")
+    imgmagickcmd = IMAGE_MAGICK
+    for (d, l) in files.items():
+        print(d)
+        x = [l[i][:-3].split("_")[5] for i in range(len(l))]
+        fig = plt.figure(figsize=(16, 12))
+        for type, changes in {"neg": changes_neg[d], "neu": changes_neu[d], "pos": changes_pos[d], "com": changes_com[d]}.items():
+            stat = [(xx, c.statistic) for xx, c in zip(x, changes) if c is not None]
+            plt.plot([p[0] for p in stat], [p[1] for p in stat], label=type + ".stat", color=colors[type])
+            if len(stat) == 0:
+                continue
+            mean = np.mean([p[1] for p in stat])
+            std = np.std([p[1] for p in stat])
+            dev = [(xx, s) for (xx, s) in stat if s <= mean - std or s >= mean + std]
+            plt.plot([p[0] for p in stat], [mean] * len(stat), color=colors[type], ls='dashed')
+            plt.plot([dx[0] for dx in dev], [dx[1] for dx in dev], color=colors[type], ls='None', marker='o')
+        plt.title("KS 2-sided test with new and old users between " + d[0] + " and " + d[1])
+        plt.xticks(rotation=90)
+        plt.legend(loc="upper right")
+        outfile = outputdir + "/ks_olddate_stat_" + d[0] + "_" + d[1] + ".png"
+        plt.savefig(outfile, bbox_inches='tight')
+        plt.close(fig)
+        imgmagickcmd += " " + outfile
+    os.system(imgmagickcmd + " " + outputdir + "/ks_olddate_stat.pdf")
 
 
 def filecmp(file1, file2):

common.py

@@ -1,8 +1,14 @@
+import importlib
+from threading import Thread, Lock
+
+import matplotlib.pyplot as plt
+
 from loader import dmt
 
 
 def calc_intervals(posts):
-    firstpost = dmt(posts).reduce(lambda acc, e: acc if acc < e['CreationDate'] else e['CreationDate'], lambda acc, e: acc if acc < e else e, lambda: posts[0]['CreationDate'], "firstpost").getresults()
+    firstpost = dmt(posts).reduce(lambda acc, e: acc if acc < e['CreationDate'] else e['CreationDate'], lambda acc, e: acc if acc < e else e, lambda: posts[0]['CreationDate'],
+                                  "firstpost").getresults()
     lastpost = dmt(posts).reduce(lambda acc, e: acc if acc > e['CreationDate'] else e['CreationDate'], lambda acc, e: acc if acc > e else e, lambda: posts[0]['CreationDate'], "lastpost").getresults()
 
     # calc quarter beginning
@@ -25,5 +31,32 @@ def calc_intervals(posts):
         print("adding interval: " + cdate.strftime("%d-%m-%Y") + " - " + nextquarter.strftime("%d-%m-%Y"))
         intervals.append((cdate, nextquarter))
         cdate = nextquarter
-        # sys.exit(0)
     return intervals
+
+
+def imprt(file):
+    spec = importlib.util.spec_from_file_location("module.name", file)
+    foo = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(foo)
+    return foo
+
+
+class FigSaver():
+    def __init__(self):
+        self.__lock = Lock()
+        self.__threads = []
+
+    def save(self, fig, path, **kwargs):
+        thread = Thread(target=self.__dosave, args=(fig, path, kwargs))
+        with self.__lock:
+            self.__threads.append(thread)
+        thread.start()
+
+    def __dosave(self, fig, path, kwargs):
+        fig.savefig(path, **kwargs)
+        plt.close(fig)
+
+    def join(self):
+        with self.__lock:
+            for thread in self.__threads:
+                thread.join()

loader.py

@@ -1,5 +1,7 @@
+import html
 import multiprocessing
 import operator
+import re
 import time
 import xml.etree.cElementTree as et
 from collections import defaultdict
@@ -7,11 +9,13 @@ from datetime import datetime
 
 from mt import mt
 
+TAG_RE = re.compile(r'<[^>]+>')
+
 printnoln = lambda text: print(text, end='', flush=True)
 rprint = lambda text: print('\r' + text)
 
 
-def dmt(data): return mt(multiprocessing.cpu_count(), data, False)
+def dmt(data, progressinterval=1000): return mt(multiprocessing.cpu_count(), data, False, progressinterval)
 
 
 def cms(): return int(round(time.time() * 1000))
@@ -75,6 +79,7 @@ def mapQuestion(item):
     question = {tag: getTag(item, tag) for tag in tags}
     for tag in datetags:
         question[tag] = datetime.fromisoformat(question[tag])
+    question['Body'] = removetags(html.unescape(question['Body']))
     return question
 
 
@@ -84,6 +89,7 @@ def mapAnswer(item):
     answer = {tag: getTag(item, tag) for tag in tags}
     for tag in datetags:
         answer[tag] = datetime.fromisoformat(answer[tag])
+    answer['Body'] = removetags(html.unescape(answer['Body']))
     return answer
 
 
@@ -93,6 +99,7 @@ def mapComment(item):
     comment = {tag: getTag(item, tag) for tag in tags}
     for tag in datetags:
         comment[tag] = datetime.fromisoformat(comment[tag])
+    comment['Body'] = removetags(html.unescape(comment['Body']))
     return comment
 
 
@@ -201,3 +208,7 @@ def tagExists(item, tag):
 def setprop(dic, key, value):
     dic[key] = value
     return dic
+
+
+def removetags(text):
+    return TAG_RE.sub('', text)

mt.py

@@ -1,10 +1,10 @@
-from threading import Thread, Lock
 import time
 from math import ceil
+from threading import Thread, Lock
 
 
 class mt():
-    def __init__(self, threads, data, verbose=False):
+    def __init__(self, threads, data, verbose=False, progressinterval=1000):
         self.__running = False
         self.__closed = False
         self.__data = data
@@ -21,6 +21,7 @@ class mt():
         self.__lock = Lock()
         self.__results = []
         self.__progress = 0
+        self.__progressinterval = progressinterval
         for i in range(self.__threadcount):
             self.__results.append([])
             self.__threads.append(None)
@@ -35,9 +36,9 @@ class mt():
         self.__running = True
         self.__final = self.__getresultsmapfilter
         self.__type = "filter"
-        self.__comment = comment if comment is not None else ""
+        self.__comment = comment
         if comment is not None:
-            print(self.__comment + ": #" + str(len(self.__data)) + " ...", end='\n' if self.__verbose else '', flush=True)
+            print(self.__comment + ": 0/" + str(len(self.__data)) + " ...", end='\n' if self.__verbose else '', flush=True)
         self.__starttime = self.__cms()
         self.__endtime = None
         for i in range(self.__threadcount):
@@ -50,20 +51,22 @@ class mt():
         now = self.__cms()
 
         results = []
-        for j in range(ceil(len(list) / 1000)):
-            part = list[j * 1000: min((j + 1) * 1000, len(list))]
+        for j in range(ceil(len(list) / self.__progressinterval)):
+            part = list[j * self.__progressinterval: min((j + 1) * self.__progressinterval, len(list))]
             results += [l for l in part if cond(l)]
             with self.__lock:
                 self.__progress += len(part)
                 if self.__comment is not None:
                     print("\r" + self.__comment + ": " + str(self.__progress) + "/" + str(len(self.__data)) + " ...", end='', flush=True)
 
-        # results = [l for l in list if cond(l)]
         with self.__lock:
             self.__results[i] = results
         dur = self.__cms() - now
         if self.__verbose:
+            if self.__comment is not None:
                 print(self.__comment + ": Thread " + str(i) + ": filter took " + str(dur) + "ms")
+            else:
+                print("Thread " + str(i) + ": filter took " + str(dur) + "ms")
 
     def map(self, func, comment=None):
         if self.__closed:
@@ -75,7 +78,7 @@ class mt():
         self.__running = True
         self.__final = self.__getresultsmapfilter
         self.__type = "map"
-        self.__comment = comment if comment is not None else ""
+        self.__comment = comment
         if comment is not None:
             print(self.__comment + ": 0/" + str(len(self.__data)) + " ...", end='\n' if self.__verbose else '', flush=True)
         self.__starttime = self.__cms()
@@ -89,20 +92,22 @@ class mt():
     def __domap(self, i, list, func):
         now = self.__cms()
         results = []
-        for j in range(ceil(len(list) / 1000)):
-            part = list[j * 1000: min((j + 1) * 1000, len(list))]
+        for j in range(ceil(len(list) / self.__progressinterval)):
+            part = list[j * self.__progressinterval: min((j + 1) * self.__progressinterval, len(list))]
             results += [func(l) for l in part]
             with self.__lock:
                 self.__progress += len(part)
                 if self.__comment is not None:
                     print("\r" + self.__comment + ": " + str(self.__progress) + "/" + str(len(self.__data)) + " ...", end='', flush=True)
 
-        # results = [func(l) for l in list]
         with self.__lock:
             self.__results[i] = results
         dur = self.__cms() - now
         if self.__verbose:
-            print(self.__comment + ": Thread " + str(i) + ": map took " + str(dur) + "ms")
+            if self.__comment is not None:
+                print(self.__comment + ": Thread " + str(i) + ": filter took " + str(dur) + "ms")
+            else:
+                print("Thread " + str(i) + ": filter took " + str(dur) + "ms")
 
     def reduce(self, reducer, aggregator, initval, comment=None):
         if self.__closed:
@@ -114,9 +119,9 @@ class mt():
         self.__running = True
         self.__final = lambda: self.__getresultsreduce(aggregator, initval)
         self.__type = "reduce"
-        self.__comment = comment if comment is not None else ""
+        self.__comment = comment
         if comment is not None:
-            print(self.__comment + ": #" + str(len(self.__data)) + " ...", end='\n' if self.__verbose else '', flush=True)
+            print(self.__comment + ": 0/" + str(len(self.__data)) + " ...", end='\n' if self.__verbose else '', flush=True)
         self.__starttime = self.__cms()
         self.__endtime = None
         for i in range(self.__threadcount):
@@ -129,8 +134,8 @@ class mt():
         now = self.__cms()
         val = initval()
 
-        for j in range(ceil(len(list) / 1000)):
-            part = list[j * 1000: min((j + 1) * 1000, len(list))]
+        for j in range(ceil(len(list) / self.__progressinterval)):
+            part = list[j * self.__progressinterval: min((j + 1) * self.__progressinterval, len(list))]
             for k in range(len(part)):
                 val = reducer(val, part[k])
             with self.__lock:
@@ -138,13 +143,14 @@ class mt():
                 if self.__comment is not None:
                     print("\r" + self.__comment + ": " + str(self.__progress) + "/" + str(len(self.__data)) + " ...", end='', flush=True)
 
-        # for j in range(len(list)):
-        #     val = reducer(val, list[j])
         with self.__lock:
             self.__results[i] = val
         dur = self.__cms() - now
         if self.__verbose:
-            print(self.__comment + ": Thread " + str(i) + ": reduce took " + str(dur) + "ms")
+            if self.__comment is not None:
+                print(self.__comment + ": Thread " + str(i) + ": filter took " + str(dur) + "ms")
+            else:
+                print("Thread " + str(i) + ": filter took " + str(dur) + "ms")
 
     def getresults(self):
         self.join()

sentiments.py

@@ -36,17 +36,12 @@ def main(folder):
     #         toxlevel = computeToxLevel(a['Body'])
     #     toxlevels[post['Id']].append(toxlevel)
     # rprint("computing toxic levels: post #" + str(len(posts)) + "/" + str(len(posts)) + " ... took " + str(cms() - start) + "ms")
-    toxlevels = dmt(posts).map(lambda p: (p['Id'], computeSentimentForPost(p)), "calculating sentiments").getresults()
+    toxlevels = dmt(posts, 10).map(lambda p: (p['Id'], {a['Id']: computeToxLevel(a['Body']) for a in p['Answers']}), "calculating sentiments").getresults()
     toxlevels = {id: p for (id, p) in toxlevels}
 
     dumptoxlevels(toxlevels, outfilename + ".py")
 
 
-def computeSentimentForPost(post):
-    anwsers = {a['Id']: computeToxLevel(a['Body']) for a in post['Answers']}
-    return anwsers
-
-
 def computeToxLevel(text):
     return analyser.polarity_scores(text)