Add author affiliation chart.

Also add a model for registering an alias for an affiliation so that
we can group affiliations that are considered the same for statistical
purposes, and a model for registering unimportant endings like Inc.
and GmbH.

Affiliation grouping is done through three means: stripping
uninteresting endings, merging entries that only differ in case and
aliases that map from case-insensitive alias to name.

Stripping endings and merging based on case seem to reduce the number
of needed manually maintained aliases greatly.
 - Legacy-Id: 12785

Author: OleLaursen

ietf/doc/admin.py

@@ -174,7 +174,7 @@ class BallotPositionDocEventAdmin(DocEventAdmin):
 admin.site.register(BallotPositionDocEvent, BallotPositionDocEventAdmin)
 
 class DocumentAuthorAdmin(admin.ModelAdmin):
-    list_display = ['id', 'document', 'person', 'email', 'order']
-    search_fields = [ 'document__name', 'person__name', 'email__address', ]
+    list_display = ['id', 'document', 'person', 'email', 'affiliation', 'order']
+    search_fields = [ 'document__name', 'person__name', 'email__address', 'affiliation']
 admin.site.register(DocumentAuthor, DocumentAuthorAdmin)
 

ietf/person/admin.py

@@ -1,7 +1,7 @@
 from django.contrib import admin
 
 
-from ietf.person.models import Email, Alias, Person
+from ietf.person.models import Email, Alias, Person, AffiliationAlias, AffiliationIgnoredEnding
 from ietf.person.name import name_parts
 
 class EmailAdmin(admin.ModelAdmin):
@@ -33,3 +33,13 @@ def plain_name(self, obj):
 #    actions = None
 admin.site.register(Person, PersonAdmin)
 
+class AffiliationAliasAdmin(admin.ModelAdmin):
+    list_filter = ["name"]
+    list_display = ["alias", "name"]
+    search_fields = ["alias", "name"]
+admin.site.register(AffiliationAlias, AffiliationAliasAdmin)
+
+class AffiliationIgnoredEndingAdmin(admin.ModelAdmin):
+    list_display = ["ending"]
+    search_fields = ["ending"]
+admin.site.register(AffiliationIgnoredEnding, AffiliationIgnoredEndingAdmin)

ietf/person/migrations/0015_affiliationalias_affiliationignoredending.py

@@ -0,0 +1,29 @@
+# -*- coding: utf-8 -*-
+from __future__ import unicode_literals
+
+from django.db import migrations, models
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('person', '0014_auto_20160613_0751'),
+    ]
+
+    operations = [
+        migrations.CreateModel(
+            name='AffiliationAlias',
+            fields=[
+                ('id', models.AutoField(verbose_name='ID', serialize=False, auto_created=True, primary_key=True)),
+                ('alias', models.CharField(help_text=b'Note that aliases are matched without regarding case.', max_length=255)),
+                ('name', models.CharField(max_length=255)),
+            ],
+        ),
+        migrations.CreateModel(
+            name='AffiliationIgnoredEnding',
+            fields=[
+                ('id', models.AutoField(verbose_name='ID', serialize=False, auto_created=True, primary_key=True)),
+                ('ending', models.CharField(max_length=255)),
+            ],
+        ),
+    ]

ietf/person/migrations/0016_auto_20170203_1030.py

@@ -0,0 +1,29 @@
+# -*- coding: utf-8 -*-
+from __future__ import unicode_literals
+
+from django.db import migrations
+
+def add_affiliation_info(apps, schema_editor):
+    AffiliationAlias = apps.get_model("person", "AffiliationAlias")
+
+    AffiliationAlias.objects.get_or_create(alias="cisco", name="Cisco Systems")
+    AffiliationAlias.objects.get_or_create(alias="cisco system", name="Cisco Systems")
+    AffiliationAlias.objects.get_or_create(alias="cisco systems (india) private limited", name="Cisco Systems")
+    AffiliationAlias.objects.get_or_create(alias="cisco systems india pvt", name="Cisco Systems")
+
+    AffiliationIgnoredEnding = apps.get_model("person", "AffiliationIgnoredEnding")
+    AffiliationIgnoredEnding.objects.get_or_create(ending="LLC\.?")
+    AffiliationIgnoredEnding.objects.get_or_create(ending="Ltd\.?")
+    AffiliationIgnoredEnding.objects.get_or_create(ending="Inc\.?")
+    AffiliationIgnoredEnding.objects.get_or_create(ending="GmbH\.?")
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('person', '0015_affiliationalias_affiliationignoredending'),
+    ]
+
+    operations = [
+        migrations.RunPython(add_affiliation_info, migrations.RunPython.noop)
+    ]

ietf/person/models.py

@@ -241,3 +241,26 @@ def email_address(self):
             return
         return self.address
 
+
+class AffiliationAlias(models.Model):
+    """Records that alias should be treated as name for statistical
+    purposes."""
+
+    alias = models.CharField(max_length=255, help_text="Note that aliases are matched without regarding case.")
+    name = models.CharField(max_length=255)
+
+    def __unicode__(self):
+        return u"{} -> {}".format(self.alias, self.name)
+
+    def save(self, *args, **kwargs):
+        self.alias = self.alias.lower()
+        super(AffiliationAlias, self).save(*args, **kwargs)
+
+class AffiliationIgnoredEnding(models.Model):
+    """Records that ending should be stripped from the affiliation for statistical purposes."""
+
+    ending = models.CharField(max_length=255, help_text="Regexp with ending, e.g. 'Inc\\.?' - remember to escape .!")
+
+    def __unicode__(self):
+        return self.ending
+    

ietf/person/utils.py

@@ -1,8 +1,10 @@
-import pprint 
+import pprint
+import re
+from collections import defaultdict
 
 from django.contrib import admin
 from django.contrib.auth.models import User
-from ietf.person.models import Person
+from ietf.person.models import Person, AffiliationAlias, AffiliationIgnoredEnding
 
 def merge_persons(source,target,stream):
 
@@ -86,3 +88,88 @@ def merge_persons(source,target,stream):
     else:
         print >>stream, "Deleting Person: {}({})".format(source.ascii,source.pk)
         source.delete()
+
+
+def compile_affiliation_ending_stripping_regexp():
+    parts = []
+    for ending_re in AffiliationIgnoredEnding.objects.values_list("ending", flat=True):
+        try:
+            re.compile(ending_re)
+        except re.error:
+            pass
+
+        parts.append(ending_re)
+
+    re_str = ",? *({}) *$".format("|".join(parts))
+
+    return re.compile(re_str, re.IGNORECASE)
+
+
+def get_aliased_affiliations(affiliations):
+    """Given non-unique sequence of affiliations, returns dictionary with
+    aliases needed.
+
+    We employ the following strategies, interleaved:
+
+    - Stripping company endings like Inc., GmbH etc. from database
+
+    - Looking up aliases stored directly in the database, like
+      "Examplar International" -> "Examplar"
+
+    - Case-folding so Examplar and EXAMPLAR is merged with the
+      winner being the one with most occurrences (so input should not
+      be made unique) or most upper case letters in case of ties.
+      Case folding can be overridden by the aliases in the database."""
+
+    res = {}
+
+    ending_re = compile_affiliation_ending_stripping_regexp()
+
+    known_aliases = { alias.lower(): name for alias, name in AffiliationAlias.objects.values_list("alias", "name") }
+
+    affiliations_with_case_spellings = defaultdict(set)
+    case_spelling_count = defaultdict(int)
+    for affiliation in affiliations:
+        original_affiliation = affiliation
+
+        # check aliases from DB
+        alias = known_aliases.get(affiliation.lower())
+        if alias is not None:
+            affiliation = alias
+            res[original_affiliation] = affiliation
+
+        # strip ending
+        alias = ending_re.sub("", affiliation)
+        if alias != affiliation:
+            affiliation = alias
+            res[original_affiliation] = affiliation
+
+        # check aliases from DB
+        alias = known_aliases.get(affiliation.lower())
+        if alias is not None:
+            affiliation = alias
+            res[original_affiliation] = affiliation
+
+        affiliations_with_case_spellings[affiliation.lower()].add(original_affiliation)
+        case_spelling_count[affiliation] += 1
+
+    def affiliation_sort_key(affiliation):
+        count = case_spelling_count[affiliation]
+        uppercase_letters = sum(1 for c in affiliation if c.isupper())
+        return (count, uppercase_letters)
+
+    # now we just need to pick the most popular uppercase/lowercase
+    # spelling for each affiliation with more than one
+    for similar_affiliations in affiliations_with_case_spellings.itervalues():
+        if len(similar_affiliations) > 1:
+            most_popular = sorted(similar_affiliations, key=affiliation_sort_key, reverse=True)[0]
+            print similar_affiliations, most_popular
+            for affiliation in similar_affiliations:
+                if affiliation != most_popular:
+                    res[affiliation] = most_popular
+                    print affiliation, "->", most_popular
+
+    return res
+
+
+

ietf/static/ietf/css/ietf.css

@@ -568,7 +568,7 @@ table.simple-table td:last-child {
   width: 7em;
 }
 
-.popover .docname {
+.document-stats .popover .element {
   padding-left: 1em;
   text-indent: -1em;
 }

ietf/static/ietf/js/document-stats.js

@@ -30,10 +30,10 @@ $(document).ready(function () {
                 if (stdNameRegExp.test(element))
                     displayName = element.slice(0, 3).toUpperCase() + " " + element.slice(3);
 
-                html.push('<div class="docname"><a href="/doc/' + element + '/">' + displayName + '</a></div>');
+                html.push('<div class="element"><a href="/doc/' + element + '/">' + displayName + '</a></div>');
             }
             else {
-                html.push('<div>' + element + '</div>');
+                html.push('<div class="element">' + element + '</div>');
             }
         });
 
@@ -44,6 +44,7 @@ $(document).ready(function () {
             trigger: "focus",
             template: '<div class="popover" role="tooltip"><div class="arrow"></div><h3 class="popover-title"></h3><div class="popover-content"></div></div>',
             content: html.join(""),
+            placement: "top",
             html: true
         }).on("click", function (e) {
             e.preventDefault();

ietf/stats/tests.py

@@ -25,7 +25,7 @@ def test_document_stats(self):
         self.assertTrue(authors_url in r["Location"])
 
         # check various stats types
-        for stats_type in ["authors", "pages", "words", "format", "formlang", "author/documents"]:
+        for stats_type in ["authors", "pages", "words", "format", "formlang", "author/documents", "author/affiliation"]:
             for document_type in ["", "rfc", "draft"]:
                 for time_choice in ["", "5y"]:
                     url = urlreverse(ietf.stats.views.document_stats, kwargs={ "stats_type": stats_type })

ietf/stats/views.py

@@ -25,6 +25,7 @@
 from ietf.person.models import Person
 from ietf.name.models import ReviewRequestStateName, ReviewResultName
 from ietf.doc.models import DocAlias, Document
+from ietf.person.utils import get_aliased_affiliations
 from ietf.ietfauth.utils import has_role
 
 def stats_index(request):
@@ -351,7 +352,7 @@ def generate_canonical_names(docalias_qs):
         total_persons = person_qs.count()
 
         if stats_type == "author/documents":
-            stats_title = "Number of {}s for each author".format(doc_label)
+            stats_title = "Number of {}s per author".format(doc_label)
 
             bins = defaultdict(list)
 
@@ -369,6 +370,38 @@ def generate_canonical_names(docalias_qs):
                 "animation": False,
             })
 
+        elif stats_type == "author/affiliation":
+            stats_title = "Number of {} authors per affiliation".format(doc_label)
+
+            bins = defaultdict(list)
+
+            # Since people don't write the affiliation names in the
+            # same way, and we don't want to go back and edit them
+            # either, we transform them here.
+
+            name_affiliation_set = set((name, affiliation)
+                                       for name, affiliation in person_qs.values_list("name", "documentauthor__affiliation"))
+
+            aliases = get_aliased_affiliations(affiliation for _, affiliation in name_affiliation_set)
+
+            for name, affiliation in name_affiliation_set:
+                bins[aliases.get(affiliation, affiliation)].append(name)
+
+            series_data = []
+            for affiliation, names in sorted(bins.iteritems(), key=lambda t: t[0].lower()):
+                percentage = len(names) * 100.0 / total_persons
+                if affiliation:
+                    series_data.append((affiliation, len(names)))
+                table_data.append((affiliation, percentage, names))
+
+            series_data.sort(key=lambda t: t[1], reverse=True)
+            series_data = series_data[:30]
+
+            chart_data.append({
+                "data": series_data,
+                "animation": False,
+            })
+
 
     return render(request, "stats/document_stats.html", {
         "chart_data": mark_safe(json.dumps(chart_data)),