[MERGE] maybe this works?

gargantext_web/api.py

@@ -82,7 +82,7 @@ class NodesChildrenNgrams(APIView):
     def get(self, request, node_id):
         # query ngrams
         ParentNode = aliased(Node)
-        ngrams_query = (Ngram
+        ngrams_query = (session
             .query(Ngram.terms, func.count().label('count'))
             # .query(Ngram.id, Ngram.terms, func.count().label('count'))
             .join(Node_Ngram, Node_Ngram.ngram_id == Ngram.id)
@@ -128,7 +128,7 @@ class NodesChildrenDuplicates(APIView):
             raise APIException('Missing GET parameter: "keys"', 400)
         keys = request.GET['keys'].split(',')
         # metadata retrieval
-        metadata_query = (Metadata
+        metadata_query = (session
             .query(Metadata)
             .filter(Metadata.name.in_(keys))
         )
@@ -213,7 +213,7 @@ class NodesChildrenMetatadata(APIView):
         
         # query metadata keys
         ParentNode = aliased(Node)
-        metadata_query = (Metadata
+        metadata_query = (session
             .query(Metadata)
             .join(Node_Metadata, Node_Metadata.metadata_id == Metadata.id)
             .join(Node, Node.id == Node_Metadata.node_id)
@@ -233,7 +233,7 @@ class NodesChildrenMetatadata(APIView):
             values_to = None
             if metadata.type != 'text':
                 value_column = getattr(Node_Metadata, 'value_' + metadata.type)
-                node_metadata_query = (Node_Metadata
+                node_metadata_query = (session
                     .query(value_column)
                     .join(Node, Node.id == Node_Metadata.node_id)
                     .filter(Node.parent_id == node_id)
@@ -381,9 +381,9 @@ class NodesChildrenQueries(APIView):
         for field_name in fields_names:
             split_field_name = field_name.split('.')
             if split_field_name[0] == 'metadata':
-                metadata = Metadata.query(Metadata).filter(Metadata.name == split_field_name[1]).first()
+                metadata = session.query(Metadata).filter(Metadata.name == split_field_name[1]).first()
                 if metadata is None:
-                    metadata_query = Metadata.query(Metadata.name).order_by(Metadata.name)
+                    metadata_query = session.query(Metadata.name).order_by(Metadata.name)
                     metadata_names = [metadata.name for metadata in metadata_query.all()]
                     raise APIException('Invalid key for "%s" in parameter "field", should be one of the following values: "%s". "%s" was found instead' % (field[0], '", "'.join(metadata_names), field[1]), 400)
                 # check or create Node_Metadata alias; join if necessary
@@ -422,7 +422,7 @@ class NodesChildrenQueries(APIView):
             )
 
         # starting the query!
-        document_type_id = NodeType.query(NodeType.id).filter(NodeType.name == 'Document').scalar()
+        document_type_id = session.query(NodeType.id).filter(NodeType.name == 'Document').scalar()
         query = (session
             .query(*fields_list)
             .select_from(Node)
@@ -451,9 +451,9 @@ class NodesChildrenQueries(APIView):
             # 
             if field[0] == 'metadata':
                 # which metadata?
-                metadata = Metadata.query(Metadata).filter(Metadata.name == field[1]).first()
+                metadata = session.query(Metadata).filter(Metadata.name == field[1]).first()
                 if metadata is None:
-                    metadata_query = Metadata.query(Metadata.name).order_by(Metadata.name)
+                    metadata_query = session.query(Metadata.name).order_by(Metadata.name)
                     metadata_names = [metadata.name for metadata in metadata_query.all()]
                     raise APIException('Invalid key for "%s" in parameter "field", should be one of the following values: "%s". "%s" was found instead' % (field[0], '", "'.join(metadata_names), field[1]), 400)                
                 # check or create Node_Metadata alias; join if necessary
@@ -475,7 +475,7 @@ class NodesChildrenQueries(APIView):
                 ))
             elif field[0] == 'ngrams': 
                 query = query.filter(
-                    Node.id.in_(Node_Metadata
+                    Node.id.in_(session
                         .query(Node_Ngram.node_id)
                         .filter(Node_Ngram.ngram_id == Ngram.id)
                         .filter(operator(
@@ -551,7 +551,7 @@ class NodesChildrenQueries(APIView):
 class NodesList(APIView):
 
     def get(self, request):
-        query = (Node
+        query = (session
             .query(Node.id, Node.name, NodeType.name.label('type'))
             .filter(Node.user_id == request.session._session_cache['_auth_user_id'])
             .join(NodeType)
@@ -626,7 +626,7 @@ class CorpusController:
 
         # build query
         ParentNode = aliased(Node)
-        query = (Ngram
+        query = (session
             .query(Ngram.terms, func.count('*'))
             .join(Node_Ngram, Node_Ngram.ngram_id == Ngram.id)
             .join(Node, Node.id == Node_Ngram.node_id)

gargantext_web/db.py

-from node import models
 from gargantext_web import settings
+from node import models
+
+
+__all__ = ['literalquery', 'session', 'cache', 'Session', 'bulk_insert', 'engine', 'get_cursor']
 
 
-__all__ = ['literalquery', 'session', 'cache']
+# initialize sqlalchemy
 
+from sqlalchemy.ext.automap import automap_base
+from sqlalchemy.orm import Session
+from sqlalchemy import create_engine
+
+engine = create_engine('postgresql+psycopg2://{USER}:{PASSWORD}@{HOST}/{NAME}'.format(
+    **settings.DATABASES['default']
+))
+Base = automap_base()
+Base.prepare(engine, reflect=True)
+
+# model representation
+
+def model_repr(modelname):
+    def _repr(obj):
+        result = '<' + modelname
+        isfirst = True
+        for key, value in obj.__dict__.items():
+            if key[0] != '_':
+                value = repr(value)
+                if len(value) > 64:
+                    value = value[:30] + '....' + value[-30:]
+                if isfirst:
+                    isfirst = False
+                else:
+                    result += ','
+                result += ' ' + key + '=' + value
+        result += '>'
+        return result
+    return _repr
 
 # map the Django models found in node.models to SQLAlchemy models
 
 for model_name, model in models.__dict__.items():
-    if hasattr(model, 'sa'):
-        globals()[model_name] = model.sa
+    if hasattr(model, '_meta'):
+        table_name = model._meta.db_table
+        if hasattr(Base.classes, table_name):
+            sqla_model = getattr(Base.classes, table_name)
+            setattr(sqla_model, '__repr__', model_repr(model_name))
+            globals()[model_name] = sqla_model
             __all__.append(model_name)
 
 NodeNgram = Node_Ngram
@@ -61,16 +97,17 @@ def literalquery(statement, dialect=None):
 
 # SQLAlchemy session management
 
-def get_sessionmaker():
-    from django.db import connections
-    from sqlalchemy.orm import sessionmaker
+def get_engine():
     from sqlalchemy import create_engine
-    alias = 'default'
-    connection = connections[alias]
     url = 'postgresql+psycopg2://{USER}:{PASSWORD}@{HOST}/{NAME}'.format(
         **settings.DATABASES['default']
     )
-    engine = create_engine(url, use_native_hstore=True)
+    return create_engine(url, use_native_hstore=True)
+
+engine = get_engine()
+
+def get_sessionmaker():
+    from sqlalchemy.orm import sessionmaker
     return sessionmaker(bind=engine)
 
 Session = get_sessionmaker()
@@ -84,7 +121,7 @@ from sqlalchemy import or_
 class ModelCache(dict):
 
     def __init__(self, model, preload=False):
-        self._model = model.sa
+        self._model = globals()[model.__name__]
         self._columns_names = [column.name for column in model._meta.fields if column.unique]
         self._columns = [getattr(self._model, column_name) for column_name in self._columns_names]
         self._columns_validators = []
@@ -92,16 +129,13 @@ class ModelCache(dict):
             self.preload()
 
     def __missing__(self, key):
-        for column in self._columns:
-            conditions = []
-            try:
-                formatted_key = column.type.python_type(key)
-                conditions.append(column == key)
-            except ValueError:
-                pass
-        if formatted_key in self:
-            self[key] = self[formatted_key]
-        else:
+        conditions = [
+            (column == key)
+            for column in self._columns
+            if key.__class__ == column.type.python_type
+        ]
+        if len(conditions) == 0:
+            raise KeyError
         element = session.query(self._model).filter(or_(*conditions)).first()
         if element is None:
             raise KeyError
@@ -127,3 +161,48 @@ class Cache:
         return modelcache
 
 cache = Cache()
+
+
+# Insert many elements at once
+
+import psycopg2
+
+def get_cursor():
+    db_settings = settings.DATABASES['default']
+    db = psycopg2.connect(**{
+        'database': db_settings['NAME'],
+        'user':     db_settings['USER'],
+        'password': db_settings['PASSWORD'],
+        'host':     db_settings['HOST'],
+    })
+    return db, db.cursor()
+
+class bulk_insert:
+
+    def __init__(self, table, keys, data, cursor=None):
+        # prepare the iterator
+        self.iter = iter(data)
+        # template
+        self.template = '%s' + (len(keys) - 1) * '\t%s' + '\n'
+        # prepare the cursor
+        if cursor is None:
+            db, cursor = get_cursor()
+            mustcommit = True
+        else:
+            mustcommit = False
+        # insert data
+        if not isinstance(table, str):
+            table = table.__table__.name
+        cursor.copy_from(self, table, columns=keys)
+        # commit if necessary
+        if mustcommit:
+            db.commit()
+
+    def read(self, size=None):
+        try:
+            return self.template % next(self.iter)
+        except StopIteration:
+            return ''
+
+    readline = read
+

gargantext_web/urls.py

@@ -3,7 +3,7 @@ from django.conf.urls import patterns, include, url
 from django.contrib import admin
 from django.contrib.auth.views import login
 
-from gargantext_web import views
+from gargantext_web import views, views_optimized
 
 import gargantext_web.api
 import scrap_pubmed.views as pubmedscrapper
@@ -31,7 +31,7 @@ urlpatterns = patterns('',
     # Project Management
     url(r'^projects/$', views.projects),
     url(r'^project/(\d+)/delete/$', views.delete_project),
-    url(r'^project/(\d+)/$', views.project),
+    url(r'^project/(\d+)/$', views_optimized.project),
     
     # Corpus management
     url(r'^project/(\d+)/corpus/(\d+)/$', views.corpus),

gargantext_web/views_optimized.py

+from django.shortcuts import redirect
+from django.shortcuts import render
+from django.http import Http404, HttpResponse, HttpResponseRedirect
+
+from sqlalchemy import func
+from sqlalchemy.orm import aliased
+
+from collections import defaultdict
+from datetime import datetime
+
+from node.admin import CustomForm
+from gargantext_web.db import *
+from gargantext_web.settings import DEBUG, MEDIA_ROOT
+
+from parsing.corpustools import add_resource, parse_resources, extract_ngrams, compute_tfidf
+
+
+def project(request, project_id):
+
+    # SQLAlchemy session
+    session = Session()
+
+    # do we have a valid project id?
+    try:
+        project_id = int(project_id)
+    except ValueError:
+        raise Http404()
+    # do we have a valid project?
+    project = (session
+        .query(Node)
+        .filter(Node.id == project_id)
+        .filter(Node.type_id == cache.NodeType['Project'].id)
+    ).first()
+    if project is None:
+        raise Http404()
+
+    # do we have a valid user?
+    user = request.user
+    if not user.is_authenticated():
+        return redirect('/login/?next=%s' % request.path)
+    if project.user_id != user.id:
+        return HttpResponseForbidden()
+
+    # Let's find out about the children nodes of the project
+    ChildrenNode = aliased(Node)
+    corpus_query = (session
+        .query(Node, Resource, func.count(ChildrenNode.id))
+        .outerjoin(ChildrenNode, ChildrenNode.parent_id == Node.id)
+        .outerjoin(Node_Resource, Node_Resource.node_id == Node.id)
+        .outerjoin(Resource, Resource.id == Node_Resource.resource_id)
+        .filter(Node.parent_id == project.id)
+        .group_by(Node, Resource)
+        .order_by(Node.name)
+    )
+    corpora_by_resourcetype = defaultdict(list)
+    documents_count_by_resourcetype = defaultdict(int)
+    corpora_count = 0
+    for corpus, resource, document_count in corpus_query:
+        if resource is None:
+            resourcetype_name = '(no resource)'
+        else:
+            resourcetype = cache.ResourceType[resource.type_id]
+            resourcetype_name = resourcetype.name
+        corpora_by_resourcetype[resourcetype_name].append({
+            'id': corpus.id,
+            'name': corpus.name,
+            'count': document_count,
+        })
+        documents_count_by_resourcetype[resourcetype_name] += document_count
+        corpora_count += 1
+
+    # do the donut
+    total_documents_count = sum(documents_count_by_resourcetype.values())
+    donut = [
+        {   'source': key, 
+            'count': value,
+            'part' : round(value * 100 / total_documents_count) if total_documents_count else 0,
+        }
+        for key, value in documents_count_by_resourcetype.items()
+    ]
+
+    # deal with the form
+    if request.method == 'POST':
+        # fomr validation
+        form = CustomForm(request.POST, request.FILES)
+        if form.is_valid():
+            # extract information from the form
+            name = form.cleaned_data['name']
+            thefile = form.cleaned_data['file']
+            resourcetype = cache.ResourceType[form.cleaned_data['type']]
+            # which default language shall be used?
+            if resourcetype.name == "europress_french":
+                language_id = cache.Language['fr'].id
+            elif resourcetype.name == "europress_english":
+                language_id = cache.Language['en'].id
+            else:
+                language_id = None
+            # corpus node instanciation as a Django model
+            corpus = Node(
+                name = name,
+                user_id = request.user.id,
+                parent_id = project_id,
+                type_id = cache.NodeType['Corpus'].id,
+                language_id = language_id,
+            )
+            session.add(corpus)
+            session.commit()
+            # save the uploaded file
+            filepath = '%s/corpora/%s/%s' % (MEDIA_ROOT, request.user.username, thefile._name)
+            f = open(filepath, 'wb')
+            f.write(thefile.read())
+            f.close()
+            # add the uploaded resource to the corpus
+            add_resource(corpus,
+                user_id = request.user.id,
+                type_id = resourcetype.id,
+                file = filepath,
+            )
+            # let's start the workflow
+            try:
+                parse_resources(corpus)
+                extract_ngrams(corpus, ['title'])
+                compute_tfidf(corpus)
+            except Exception as error:
+                print('WORKFLOW ERROR')
+                print(error)
+            # redirect to the main project page
+            return HttpResponseRedirect('/project/' + str(project_id))
+        else:
+            print('ERROR: BAD FORM')
+    else:
+        form = CustomForm()
+
+    # HTML output
+    return render(request, 'project.html', {
+        'form'          : form,
+        'user'          : user,
+        'date'          : datetime.now(),
+        'project'       : project,
+        'donut'         : donut,
+        'list_corpora'  : dict(corpora_by_resourcetype),
+        'whitelists'    : '',
+        'blacklists'    : '',
+        'cooclists'     : '',
+        'number'        : corpora_count,
+    })
+

node/models.py

@@ -70,7 +70,7 @@ class Resource(models.Model):
         return self.file
 
 class NodeType(models.Model):
-    name        = models.CharField(max_length=200, unique=True)
+    name        = models.CharField(max_length=255, unique=True)
     def __str__(self):
         return self.name
 
@@ -111,7 +111,7 @@ class NodeManager(CTENodeManager):
         return getattr(self.get_queryset(), name, *args)
 
 class Metadata(models.Model):
-    name        = models.CharField(max_length=32, db_index=True)
+    name        = models.CharField(max_length=32, unique=True)
     type        = models.CharField(max_length=16, db_index=True)
         
 class Node(CTENode):
@@ -120,7 +120,7 @@ class Node(CTENode):
 
     user        = models.ForeignKey(User)
     type        = models.ForeignKey(NodeType)
-    name        = models.CharField(max_length=200)
+    name        = models.CharField(max_length=255)
     
     language    = models.ForeignKey(Language, blank=True, null=True, on_delete=models.SET_NULL)
     
@@ -189,7 +189,7 @@ class Node(CTENode):
         for i, metadata_values in enumerate(metadata_list):
             if verbose:
                 print(i, end='\r', flush=True)
-            name = metadata_values.get('title', '')[:200]
+            name = metadata_values.get('title', '')[:255]
             language = langages_cache[metadata_values['language_iso2']] if 'language_iso2' in metadata_values else None,
             if isinstance(language, tuple):
                 language = language[0]

parsing/FileParsers/EuropressFileParser.py

@@ -48,11 +48,9 @@ class EuropressFileParser(FileParser):
                 print(error)
             
 
-        except:
-            return []
+        except Exception as error:
+            print(error)
 
-        # initialize the list of metadata
-        metadata_list = []
         # parse all the articles, one by one
         try:
             for html_article in html_articles:
@@ -201,16 +199,9 @@ class EuropressFileParser(FileParser):
                   #metadata_str = {}
                     for key, value in metadata.items():
                         metadata[key] = value.decode() if isinstance(value, bytes) else value
-                    metadata_list.append(metadata)
+                    yield metadata
                     count += 1
         
         except Exception as error:
             print(error)
             pass
-
-#       from pprint import pprint
-#       pprint(metadata_list)
-#       return []
-        return metadata_list
-
-

parsing/FileParsers/FileParser.py

@@ -103,15 +103,21 @@ class FileParser:
             zipArchive = zipfile.ZipFile(file)
             for filename in zipArchive.namelist():
                 try:
-                    metadata_list += self.parse(zipArchive.open(filename, "r"))
+                    f = zipArchive.open(filename, 'r')
+                    metadata_list += self.parse(f)
+                    f.close()
                 except Exception as error:
                     print(error)
         # ...otherwise, let's parse it directly!
         else:
             try:
-                metadata_list += self._parse(file)
+                for metadata in self._parse(file):
+                    metadata_list.append(self.format_metadata(metadata))
+                if hasattr(file, 'close'):
+                    file.close()
             except Exception as error:
                 print(error)
         # return the list of formatted metadata
-        return map(self.format_metadata, metadata_list)
+        return metadata_list
+
 

parsing/FileParsers/PubmedFileParser.py

@@ -2,22 +2,14 @@ from django.db import transaction
 from lxml import etree
 from .FileParser import FileParser
 from ..NgramsExtractors import *
-from datetime import datetime
-from io import BytesIO
 
 class PubmedFileParser(FileParser):
     
     def _parse(self, file):
         # open the file as XML
         xml_parser = etree.XMLParser(resolve_entities=False, recover=True)
-        
-        xml = ""
-        if type(file)==bytes: xml = etree.parse( BytesIO(file) , parser=xml_parser)
-        else: xml = etree.parse(file, parser=xml_parser)
-
+        xml = etree.parse(file, parser=xml_parser)
         xml_articles = xml.findall('PubmedArticle')
-        # initialize the list of metadata
-        metadata_list = []
         # parse all the articles, one by one
         for xml_article in xml_articles:
             # extract data from the document
@@ -25,14 +17,9 @@ class PubmedFileParser(FileParser):
             metadata_path = {
                 "journal"           : 'MedlineCitation/Article/Journal/Title',
                 "title"             : 'MedlineCitation/Article/ArticleTitle',
-                # "abstract"          : 'MedlineCitation/Article/Abstract/AbstractText',
-                "title2"            : 'MedlineCitation/Article/VernacularTitle',
                 "language_iso3"     : 'MedlineCitation/Article/Language',
                 "doi"               : 'PubmedData/ArticleIdList/ArticleId[@type=doi]',
-                "realdate_full_"     : 'MedlineCitation/Article/Journal/JournalIssue/PubDate/MedlineDate',
-                "realdate_year_"     : 'MedlineCitation/Article/Journal/JournalIssue/PubDate/Year',
-                "realdate_month_"    : 'MedlineCitation/Article/Journal/JournalIssue/PubDate/Month',
-                "realdate_day_"      : 'MedlineCitation/Article/Journal/JournalIssue/PubDate/Day',
+                "abstract"          : 'MedlineCitation/Article/Abstract/AbstractText',
                 "publication_year"  : 'MedlineCitation/DateCreated/Year',
                 "publication_month" : 'MedlineCitation/DateCreated/Month',
                 "publication_day"   : 'MedlineCitation/DateCreated/Day',
@@ -41,7 +28,6 @@ class PubmedFileParser(FileParser):
             for key, path in metadata_path.items():
                 try:
                     xml_node = xml_article.find(path)
-                    # Authors tag
                     if key == 'authors':
                         metadata[key] = ', '.join([
                             xml_author.find('ForeName').text + ' ' + xml_author.find('LastName').text
@@ -49,61 +35,6 @@ class PubmedFileParser(FileParser):
                         ])
                     else:
                         metadata[key] = xml_node.text
-
                 except:
                     pass
-
-            #Title-Decision
-            Title=""
-            if not metadata["title"] or metadata["title"]=="":
-                if "title2" in metadata:
-                    metadata["title"] = metadata["title2"]
-                else: metadata["title"] = ""
-
-            # Date-Decision
-            # forge.iscpif.fr/issues/1418
-            RealDate = ""
-            if "realdate_full_" in metadata:
-                RealDate = metadata["realdate_full_"]
-            else:
-                if "realdate_year_" in metadata: RealDate+=metadata["realdate_year_"]
-                if "realdate_month_" in metadata: RealDate+=" "+metadata["realdate_month_"]
-                if "realdate_day_" in metadata: RealDate+=" "+metadata["realdate_day_"]
-            metadata["realdate_full_"] = RealDate
-            RealDate = RealDate.split("-")[0]
-
-            PubmedDate = ""
-            if "publication_year" in metadata: PubmedDate+=metadata["publication_year"]
-            if "publication_month" in metadata: PubmedDate+=" "+metadata["publication_month"]
-            if "publication_day" in metadata: PubmedDate+=" "+metadata["publication_day"]
-
-            Decision=""
-            if len(RealDate)>4:
-                if len(RealDate)>8:
-                    try: Decision = datetime.strptime(RealDate, '%Y %b %d').date()
-                    except: 
-                        try: Decision = datetime.strptime(PubmedDate, '%Y %m %d').date()
-                        except: Decision=False
-                else: 
-                    try: Decision = datetime.strptime(RealDate, '%Y %b').date()
-                    except: 
-                        try: Decision = datetime.strptime(PubmedDate, '%Y %m %d').date()
-                        except: Decision=False
-            else: 
-                try: Decision = datetime.strptime(PubmedDate, '%Y %m %d').date()
-                except: Decision=False
-
-            if Decision!=False:
-                if "publication_year" in metadata: metadata["publication_year"] = str(Decision.year)
-                if "publication_month" in metadata: metadata["publication_month"] = str(Decision.month)
-                if "publication_day" in metadata: metadata["publication_day"] = str(Decision.day)
-                if "realdate_year_" in metadata: metadata.pop("realdate_year_")
-                if "realdate_month_" in metadata: metadata.pop("realdate_month_")
-                if "realdate_day_" in metadata: metadata.pop("realdate_day_")
-                if "title2" in metadata: metadata.pop("title2")
-                
-                # print(metadata)
-                # print("* * * * ** * * * * ")
-                metadata_list.append(metadata)
-        # return the list of metadata
-        return metadata_list
+            yield metadata
\ No newline at end of file

parsing/FileParsers/RisFileParser.py

@@ -17,42 +17,34 @@ class RisFileParser(FileParser):
     }
 
     def _parse(self, file):
-        metadata_list = []
         metadata = {}
         last_key = None
         last_values = []
+        # browse every line of the file
         for line in file:
             if len(line) > 2:
+                # extract the parameter key
                 parameter_key = line[:2]
-#                print(parameter_key)
                 if parameter_key != b'  ' and parameter_key != last_key:
                     if last_key in self._parameters:
+                        # translate the parameter key
                         parameter = self._parameters[last_key]
                         if parameter["type"] == "metadata":
                             separator = parameter["separator"] if "separator" in parameter else ""
                             metadata[parameter["key"]] = separator.join(last_values)
                         elif parameter["type"] == "delimiter":
-                            #language = self._languages_fullname[metadata["language"].lower()]
-                            #print(metadata)
-                            try:
-                                #print("append")
                             if 'language_fullname' not in metadata.keys():
                                 if 'language_iso3' not in metadata.keys():
                                     if 'language_iso2' not in metadata.keys():
                                         metadata['language_iso2'] = 'en'
-                                metadata_list.append(metadata)
+                            yield metadata
                             metadata = {}
-                                #print("append succeeded")
-                            except:
-                                pass
                     last_key = parameter_key
                     last_values = []
                 try:
                     last_values.append(line[3:-1].decode())
                 except Exception as error:
                     print(error)
-                    pass
-        #print(len(metadata_list))
-        #print(metadata_list)
-
-        return metadata_list
+        # if a metadata object is left in memory, yield it as well
+        if metadata:
+            yield metadata

parsing/NgramsExtractors/NgramsExtractor.py

@@ -13,6 +13,7 @@ class NgramsExtractor:
         self.start()
         self._label = "NP"
         self._rule = self._label + ": " + rule
+        self._grammar = nltk.RegexpParser(self._rule)
         
     def __del__(self):
         self.stop()
@@ -29,19 +30,8 @@ class NgramsExtractor:
     """
     def extract_ngrams(self, contents):
         tagged_ngrams = self.tagger.tag_text(contents)
-        if len(tagged_ngrams)==0: return []
-        
-        grammar = nltk.RegexpParser(self._rule)
-        result = []
-        # try:
-        grammar_parsed = grammar.parse(tagged_ngrams)
+        if len(tagged_ngrams):
+            grammar_parsed = self._grammar.parse(tagged_ngrams)
             for subtree in grammar_parsed.subtrees():
                 if subtree.label() == self._label:
-                result.append(subtree.leaves())
-        # except Exception as e:
-        #     print("Problem while parsing rule '%s'" % (self._rule, ))
-        #     print(e)
-        return result
-        
-        
-    
+                    yield subtree.leaves()

parsing/NgramsExtractors/__init__.py

 from .FrenchNgramsExtractor import FrenchNgramsExtractor
 from .TurboNgramsExtractor import TurboNgramsExtractor as EnglishNgramsExtractor
-# from parsing.NgramsExtractors.EnglishNgramsExtractor import EnglishNgramsExtractor
+# from .EnglishNgramsExtractor import EnglishNgramsExtractor
 from .NgramsExtractor import NgramsExtractor
 

parsing/Taggers/Tagger.py

@@ -71,4 +71,3 @@ class Tagger:
             tokens_tags += self.tag_tokens(tokens, False)
         self.tagging_end()
         return tokens_tags
-

parsing/Taggers/nlpserver/client.py

@@ -9,15 +9,24 @@ from .settings import implemented_methods
 class NLPClient:
 
     def __init__(self):
-        self._socket = socket.socket(*server_type_client)
-        self._socket.connect((server_host, server_port))
+        self._socket = None
         for method_name in dir(self):
             if method_name[0] != '_':
                 if method_name.upper() not in implemented_methods:
                     setattr(self, method_name, self._notimplemented)
 
     def __del__(self):
+        self._disconnect()
+
+    def _connect(self):
+        self._disconnect()
+        self._socket = socket.socket(*server_type_client)
+        self._socket.connect((server_host, server_port))
+
+    def _disconnect(self):
+        if self._socket is not None:
             self._socket.close()
+            self._socket = None
 
     def _notimplemented(self, *args, **kwargs):
         raise NotImplementedError(
@@ -51,7 +60,7 @@ class NLPClient:
         data += language + '\n'
         data += re.sub(r'\n+', '\n', text)
         data += '\n\n'
-        self.__init__()
+        self._connect()
         self._socket.sendall(data.encode())
         sentence = []
         if keys is None:
@@ -73,7 +82,6 @@ class NLPClient:
                     continue
                 values = line.split('\t')
                 sentence.append(dict(zip(keys, line.split('\t'))))
-        self.__del__()
 
     def tokenize(self, text, language='english', asdict=False):
         keys = ('token', ) if asdict else None

parsing/Taggers/nlpserver/settings.py

@@ -4,7 +4,7 @@ import socketserver
 
 # Server parameters
 server_host = 'localhost'
-server_port = 1234
+server_port = 7777
 server_type_server = socketserver.TCPServer
 server_type_client = socket.AF_INET, socket.SOCK_STREAM
 server_timeout = 2.0

parsing/corpustools.py

+from collections import defaultdict
+from datetime import datetime
+from random import random
+from hashlib import md5
+from time import time
+from math import log
+
+from gargantext_web.db import *
+
+from .FileParsers import *
+
+
+
+class DebugTime:
+
+    def __init__(self, prefix):
+        self.prefix = prefix
+        self.message = None
+        self.time = None
+
+    def __del__(self):
+        if self.message is not None and self.time is not None:
+            print('%s - %s: %.4f' % (self.prefix, self.message, time() - self.time))
+
+    def show(self, message):
+        self.__del__()
+        self.message = message
+        self.time = time()
+
+
+# keep all the parsers in a cache
+class Parsers(defaultdict):
+
+    _parsers = {
+        'pubmed'            : PubmedFileParser,
+        'isi'               : IsiFileParser,
+        'ris'               : RisFileParser,
+        'europress'         : EuropressFileParser,
+        'europress_french'  : EuropressFileParser,
+        'europress_english' : EuropressFileParser,
+    }
+
+    def __missing__(self, key):
+        if key not in self._parsers:
+            raise NotImplementedError('No such parser: "%s"' % (key))
+        parser = self._parsers[key]()
+        self[key] = parser
+        return parser
+
+parsers = Parsers()
+
+
+
+# resources managment
+
+def add_resource(corpus, **kwargs):
+    # only for tests
+    session = Session()
+    resource = Resource(guid=str(random()), **kwargs )
+    # User
+    if 'user_id' not in kwargs:
+        resource.user_id = corpus.user_id
+    # Compute the digest
+    h = md5()
+    f = open(str(resource.file), 'rb')
+    h.update(f.read())
+    f.close()
+    resource.digest = h.hexdigest()
+    # check if a resource on this node already has this hash
+    tmp_resource = (session
+        .query(Resource)
+        .join(Node_Resource, Node_Resource.resource_id == Resource.id)
+        .filter(Resource.digest == resource.digest)
+        .filter(Node_Resource.node_id == corpus.id)
+    ).first()
+    if tmp_resource is not None:
+        return tmp_resource
+    else:
+        session.add(resource)
+        session.commit()
+    # link with the resource
+    node_resource = Node_Resource(
+        node_id = corpus.id,
+        resource_id = resource.id,
+        parsed = False,
+    )
+    session.add(node_resource)
+    session.commit()
+    # return result
+    return resource
+
+
+def parse_resources(corpus, user=None, user_id=None):
+    dbg = DebugTime('Corpus #%d - parsing' % corpus.id)
+    session = Session()
+    corpus_id = corpus.id
+    type_id = cache.NodeType['Document'].id
+    if user_id is None and user is not None:
+        user_id = user.id
+    else:
+        user_id = corpus.user_id
+    # find resource of the corpus
+    resources_query = (session
+        .query(Resource, ResourceType)
+        .join(ResourceType, ResourceType.id == Resource.type_id)
+        .join(Node_Resource, Node_Resource.resource_id == Resource.id)
+        .filter(Node_Resource.node_id == corpus.id)
+        .filter(Node_Resource.parsed == False)
+    )
+    # make a new node for every parsed document of the corpus
+    dbg.show('analyze documents')
+    nodes = list()
+    for resource, resourcetype in resources_query:
+        parser = parsers[resourcetype.name]
+        for metadata_dict in parser.parse(resource.file):
+            # retrieve language ID from metadata
+            if 'language_iso2' in metadata_dict:
+                try:
+                    language_id = cache.Language[metadata_dict['language_iso2']].id
+                except KeyError:
+                    language_id = None
+            else:
+                language_id = None
+            # create new node
+            node = Node(
+                name = metadata_dict.get('title', '')[:255],
+                parent_id = corpus_id,
+                user_id = user_id,
+                type_id = type_id,
+                language_id = language_id,
+                metadata = metadata_dict,
+                date = datetime.utcnow(),
+            )
+            nodes.append(node)
+            #
+            # TODO: mark node-resources associations as parsed
+            # 
+    dbg.show('insert %d documents' % len(nodes))
+    session.add_all(nodes)
+    session.commit()
+    # now, index the metadata
+    dbg.show('insert metadata')
+    node_metadata_lists = defaultdict(list)
+    metadata_types = {
+        metadata.name: metadata
+        for metadata in session.query(Metadata)
+    }
+    for node in nodes:
+        node_id = node.id
+        for metadata_key, metadata_value in node.metadata.items():
+            try:
+                metadata = metadata_types[metadata_key]
+            except KeyError:
+                continue
+            if metadata.type == 'string':
+                metadata_value = metadata_value[:255]
+            node_metadata_lists[metadata.type].append((
+                node_id,
+                metadata.id,
+                metadata_value,
+            ))
+    for key, values in node_metadata_lists.items():
+        bulk_insert(Node_Metadata, ['node_id', 'metadata_id', 'value_'+key], values)
+    # mark the corpus as parsed
+    corpus.parsed = True
+
+
+# ngrams extraction
+
+from .NgramsExtractors import EnglishNgramsExtractor, FrenchNgramsExtractor, NgramsExtractor
+
+class NgramsExtractors(defaultdict):
+
+    def __init__(self):
+        # English
+        self['en'] = EnglishNgramsExtractor()
+        for key in ('eng', 'english'):
+            self[key] = self['en']
+        # French
+        self['fr'] = FrenchNgramsExtractor()
+        for key in ('fre', 'french'):
+            self[key] = self['fr']
+        # default
+        self['default'] = NgramsExtractor()
+
+    def __missing__(self, key):
+        formatted_key = key.strip().lower()
+        if formatted_key in self:
+            self[key] = self[formatted_key]
+        else:
+            self[key] = self['default']
+            # raise NotImplementedError
+        return self[key]
+
+ngramsextractors = NgramsExtractors()
+
+def extract_ngrams(corpus, keys):
+    dbg = DebugTime('Corpus #%d - ngrams' % corpus.id)
+    default_language_iso2 = None if corpus.language_id is None else cache.Language[corpus.language_id].iso2
+    # query the metadata associated with the given keys
+    columns = [Node.id, Node.language_id] + [Node.metadata[key] for key in keys]
+    metadata_query = (session
+        .query(*columns)
+        .filter(Node.parent_id == corpus.id)
+        .filter(Node.type_id == cache.NodeType['Document'].id)
+    )
+    # prepare data to be inserted
+    dbg.show('find ngrams')
+    languages_by_id = {
+        language.id: language.iso2
+        for language in session.query(Language)
+    }
+    ngrams_data = set()
+    node_ngram_list = defaultdict(lambda: defaultdict(int))
+    for nodeinfo in metadata_query:
+        node_id = nodeinfo[0]
+        language_id = nodeinfo[1]
+        if language_id is None:
+            language_iso2 = default_language_iso2
+        else:
+            language_iso2 = languages_by_id.get(language_id, None)
+        if language_iso2 is None:
+            continue
+        ngramsextractor = ngramsextractors[language_iso2]
+        for text in nodeinfo[2:]:
+            if text is not None and len(text):
+                ngrams = ngramsextractor.extract_ngrams(text)
+                for ngram in ngrams:
+                    terms = ' '.join([token for token, tag in ngram]).lower()
+                    n = len(ngram)
+                    node_ngram_list[node_id][terms] += 1
+                    ngrams_data.add(
+                        (n, terms)
+                    )
+    # insert ngrams to temporary table
+    dbg.show('find ids for the %d ngrams' % len(ngrams_data))
+    db, cursor = get_cursor()
+    cursor.execute('''
+        CREATE TEMPORARY TABLE tmp__ngrams (
+            id INT,
+            n INT NOT NULL,
+            terms VARCHAR(255) NOT NULL
+        )
+    ''')
+    bulk_insert('tmp__ngrams', ['n', 'terms'], ngrams_data, cursor=cursor)
+    # retrieve ngram ids from already inserted stuff
+    cursor.execute('''
+        UPDATE
+            tmp__ngrams
+        SET
+            id = ngram.id
+        FROM
+            %s AS ngram
+        WHERE
+            ngram.terms = tmp__ngrams.terms
+    ''' % (Ngram.__table__.name, ))
+    # insert, then get the ids back
+    cursor.execute('''
+        INSERT INTO
+            %s (n, terms)
+        SELECT
+            n, terms
+        FROM
+            tmp__ngrams
+        WHERE
+            id IS NULL
+    ''' % (Ngram.__table__.name, ))
+    cursor.execute('''
+        UPDATE
+            tmp__ngrams
+        SET
+            id = ngram.id
+        FROM
+            %s AS ngram
+        WHERE
+            ngram.terms = tmp__ngrams.terms
+        AND
+            tmp__ngrams.id IS NULL
+    ''' % (Ngram.__table__.name, ))
+    # get all ids
+    ngram_ids = dict()
+    cursor.execute('SELECT id, terms FROM tmp__ngrams')
+    for row in cursor.fetchall():
+        ngram_ids[row[1]] = row[0]
+    # 
+    dbg.show('insert associations')
+    node_ngram_data = list()
+    for node_id, ngrams in node_ngram_list.items():
+        for terms, weight in ngrams.items():
+            ngram_id = ngram_ids[terms]
+            node_ngram_data.append((node_id, ngram_id, weight, ))
+    bulk_insert(Node_Ngram, ['node_id', 'ngram_id', 'weight'], node_ngram_data, cursor=cursor)
+    dbg.message = 'insert %d associations' % len(node_ngram_data)
+    # commit to database
+    db.commit()
+
+
+
+# tfidf calculation
+
+def compute_tfidf(corpus):
+    dbg = DebugTime('Corpus #%d - tfidf' % corpus.id)
+    # compute terms frequency sum
+    dbg.show('calculate terms frequencies sums')
+    db, cursor = get_cursor()
+    cursor.execute('''
+        CREATE TEMPORARY TABLE tmp__st (
+            node_id INT NOT NULL,
+            frequency DOUBLE PRECISION NOT NULL
+        )
+    ''')
+    cursor.execute('''
+        INSERT INTO
+            tmp__st (node_id, frequency)
+        SELECT
+            node_ngram.node_id,
+            SUM(node_ngram.weight) AS frequency
+        FROM
+            %s AS node
+        INNER JOIN
+            %s AS node_ngram ON node_ngram.node_id = node.id
+        WHERE
+            node.parent_id = %d
+        GROUP BY
+            node_ngram.node_id
+    ''' % (Node.__table__.name, Node_Ngram.__table__.name, corpus.id, ))
+    # compute normalized terms frequencies
+    dbg.show('normalize terms frequencies')
+    cursor.execute('''
+        CREATE TEMPORARY TABLE tmp__tf (
+            node_id INT NOT NULL,
+            ngram_id INT NOT NULL,
+            frequency DOUBLE PRECISION NOT NULL
+        )
+    ''')
+    cursor.execute('''
+        INSERT INTO
+            tmp__tf (node_id, ngram_id, frequency)
+        SELECT
+            node_ngram.node_id,
+            node_ngram.ngram_id,
+            (node_ngram.weight / node.frequency) AS frequency
+        FROM
+            %s AS node_ngram
+        INNER JOIN
+            tmp__st AS node ON node.node_id = node_ngram.node_id
+    ''' % (Node_Ngram.__table__.name, ))
+    # show off
+    dbg.show('compute idf')
+    cursor.execute('''
+        CREATE TEMPORARY TABLE tmp__idf (
+            ngram_id INT NOT NULL,
+            idf DOUBLE PRECISION NOT NULL
+        )
+    ''')
+    cursor.execute('''
+        INSERT INTO
+            tmp__idf(ngram_id, idf)
+        SELECT
+            node_ngram.ngram_id,
+            -ln(COUNT(*))
+        FROM
+            %s AS node
+        INNER JOIN
+            %s AS node_ngram ON node_ngram.node_id = node.id
+        WHERE
+            node.parent_id = %d
+        GROUP BY
+            node_ngram.ngram_id
+    ''' % (Node.__table__.name, Node_Ngram.__table__.name, corpus.id, ))
+    cursor.execute('SELECT COUNT(*) FROM tmp__st')
+    D = cursor.fetchone()[0]
+    lnD = log(D)
+    cursor.execute('UPDATE tmp__idf SET idf = idf + %f' % (lnD, ))
+    # show off
+    dbg.show('insert tfidf for %d documents' % D)
+    cursor.execute('''
+        INSERT INTO
+            %s (nodex_id, nodey_id, ngram_id, score)
+        SELECT
+            %d AS nodex_id,
+            tf.node_id AS nodey_id,
+            tf.ngram_id AS ngram_id,
+            (tf.frequency * idf.idf) AS score
+        FROM
+            tmp__idf AS idf
+        INNER JOIN
+            tmp__tf AS tf ON tf.ngram_id = idf.ngram_id
+    ''' % (NodeNodeNgram.__table__.name, corpus.id, ))
+    # # show off
+    # cursor.execute('''
+    #     SELECT
+    #         node.name,
+    #         ngram.terms,
+    #         node_node_ngram.score AS tfidf
+    #     FROM
+    #         %s AS node_node_ngram
+    #     INNER JOIN
+    #         %s AS node ON node.id = node_node_ngram.nodey_id
+    #     INNER JOIN
+    #         %s AS ngram ON ngram.id = node_node_ngram.ngram_id
+    #     WHERE
+    #         node_node_ngram.nodex_id = %d
+    #     ORDER BY
+    #         score DESC
+    # ''' % (NodeNodeNgram.__table__.name, Node.__table__.name, Ngram.__table__.name, corpus.id, ))
+    # for row in cursor.fetchall():
+    #     print(row)
+    # the end!
+    db.commit()

test_db.py

+# Without this, we couldn't use the Django environment
+import os
+os.environ.setdefault("DJANGO_SETTINGS_MODULE", "gargantext_web.settings")
+os.environ.setdefault("DJANGO_HSTORE_GLOBAL_REGISTER", "False")
+
+# database tools
+from gargantext_web.db import *
+from parsing.corpustools import *
+
+
+
+
+user = session.query(User).first()
+project = session.query(Node).filter(Node.name == 'A').first()
+corpus = Node(
+    parent_id = project.id,
+    name = 'Test 456',
+    type_id = cache.NodeType['Corpus'].id,
+    user_id = user.id,
+)
+
+session.add(corpus)
+session.commit()
+
+add_resource(corpus,
+    # file = './data_samples/pubmed_result.xml',
+    file = './data_samples/pubmed_2013-04-01_HoneyBeesBeeBees.xml',
+    type_id = cache.ResourceType['pubmed'].id,
+)
+parse_resources(corpus)
+extract_ngrams(corpus, ('title', ))
+
+
+
+# print(corpus)
+# corpus = session.query(Node).filter(Node.id == 72771).first()
+# corpus = session.query(Node).filter(Node.id == 73017).first()
+compute_tfidf(corpus)