Process corpus files in task, not in database model

app/corpora/tasks.py

@@ -1,9 +1,52 @@
+from .. import db
 from ..decorators import background
 from ..models import Corpus, CorpusFile
+import xml.etree.ElementTree as ET
 import os
 import shutil
 
 
+@background
+def build_corpus(app, corpus_id):
+    with app.app_context():
+        corpus = Corpus.query.get(corpus_id)
+        if corpus is None:
+            return
+        corpus.status = 'File processing'
+        db.session.commit()
+        corpus_dir = os.path.join(app.config['NOPAQUE_STORAGE'],
+                                  str(corpus.user_id), 'corpora',
+                                  str(corpus.id))
+        output_dir = os.path.join(corpus_dir, 'merged')
+        shutil.rmtree(output_dir, ignore_errors=True)
+        os.mkdir(output_dir)
+        master_element_tree = ET.ElementTree(
+            ET.fromstring('<corpus>\n</corpus>'))
+        for corpus_file in corpus.files:
+            file = os.path.join(corpus_dir, corpus_file.filename)
+            element_tree = ET.parse(file)
+            text_node = element_tree.find('text')
+            text_node.set('address', corpus_file.address or "NULL")
+            text_node.set('author', corpus_file.author)
+            text_node.set('booktitle', corpus_file.booktitle or "NULL")
+            text_node.set('chapter', corpus_file.chapter or "NULL")
+            text_node.set('editor', corpus_file.editor or "NULL")
+            text_node.set('institution', corpus_file.institution or "NULL")
+            text_node.set('journal', corpus_file.journal or "NULL")
+            text_node.set('pages', corpus_file.pages or "NULL")
+            text_node.set('publisher', corpus_file.publisher or "NULL")
+            text_node.set('publishing_year', str(corpus_file.publishing_year))
+            text_node.set('school', corpus_file.school or "NULL")
+            text_node.set('title', corpus_file.title)
+            element_tree.write(file)
+            master_element_tree.getroot().insert(1, text_node)
+        output_file = os.path.join(output_dir, 'corpus.vrt')
+        master_element_tree.write(output_file, xml_declaration=True,
+                                  encoding='utf-8')
+        corpus.status = 'submitted'
+        db.session.commit()
+
+
 @background
 def delete_corpus(app, corpus_id):
     with app.app_context():
@@ -30,12 +73,3 @@ def delete_corpus_file(app, corpus_file_id):
             pass
         else:
             corpus_file.delete()
-
-
-@background
-def edit_corpus_file(app, corpus_file_id):
-    with app.app_context():
-        corpus_file = CorpusFile.query.get(corpus_file_id)
-        if corpus_file is None:
-            raise Exception('Corpus file {} not found!'.format(corpus_file_id))
-        corpus_file.insert_metadata()

app/corpora/views.py

@@ -60,14 +60,16 @@ def analyse_corpus(corpus_id):
     query_form = QueryForm(prefix='query-form',
                            query=request.args.get('query'))
     query_download_form = QueryDownloadForm(prefix='query-download-form')
-    inspect_display_options_form = InspectDisplayOptionsForm(prefix='inspect-display-options-form')
-    return render_template('corpora/analyse_corpus.html.j2',
-                           corpus_id=corpus_id,
-                           display_options_form=display_options_form,
-                           query_form=query_form,
-                           query_download_form=query_download_form,
-                           inspect_display_options_form=inspect_display_options_form,
-                           title='Corpus analysis')
+    inspect_display_options_form = InspectDisplayOptionsForm(
+        prefix='inspect-display-options-form')
+    return render_template(
+        'corpora/analyse_corpus.html.j2',
+        corpus_id=corpus_id,
+        display_options_form=display_options_form,
+        query_form=query_form,
+        query_download_form=query_download_form,
+        inspect_display_options_form=inspect_display_options_form,
+        title='Corpus analysis')
 
 
 @corpora.route('/<int:corpus_id>/delete')
@@ -114,8 +116,8 @@ def add_corpus_file(corpus_id):
             school=add_corpus_file_form.school.data,
             title=add_corpus_file_form.title.data)
         db.session.add(corpus_file)
+        corpus.status = 'unprepared'
         db.session.commit()
-        tasks.edit_corpus_file(corpus_file.id)
         flash('Corpus file added!')
         return make_response(
             {'redirect_url': url_for('corpora.corpus', corpus_id=corpus.id)},
@@ -181,8 +183,8 @@ def edit_corpus_file(corpus_id, corpus_file_id):
             edit_corpus_file_form.publishing_year.data
         corpus_file.school = edit_corpus_file_form.school.data
         corpus_file.title = edit_corpus_file_form.title.data
+        corpus.status = 'unprepared'
         db.session.commit()
-        tasks.edit_corpus_file(corpus_file_id)
         flash('Corpus file edited!')
         return redirect(url_for('corpora.corpus', corpus_id=corpus_id))
     # If no form is submitted or valid, fill out fields with current values
@@ -211,9 +213,8 @@ def prepare_corpus(corpus_id):
     if not (corpus.creator == current_user or current_user.is_administrator()):
         abort(403)
     if corpus.files.all():
-        corpus.status = 'submitted'
-        db.session.commit()
-        flash('Corpus marked for preparation!')
+        tasks.build_corpus(corpus_id)
+        flash('Corpus gets build now.')
     else:
-        flash('Can not prepare corpus, please add corpus file(s).')
+        flash('Can not build corpus, please add corpus file(s).')
     return redirect(url_for('corpora.corpus', corpus_id=corpus_id))

app/models.py

@@ -7,7 +7,6 @@ from werkzeug.utils import secure_filename
 from . import db, logger, login_manager
 import os
 import shutil
-import xml.etree.ElementTree as ET
 
 
 class Permission:
@@ -380,28 +379,6 @@ class CorpusFile(db.Model):
         db.session.delete(self)
         db.session.commit()
 
-    def insert_metadata(self):
-        file = os.path.join(current_app.config['NOPAQUE_STORAGE'],
-                            self.dir, self.filename)
-        element_tree = ET.parse(file)
-        text_node = element_tree.find('text')
-        # TODO: USE OR
-        text_node.set('address', self.address if self.address else "NULL")
-        text_node.set('author', self.author)
-        text_node.set('booktitle', self.booktitle if self.booktitle else "NULL")
-        text_node.set('chapter', self.chapter if self.chapter else "NULL")
-        text_node.set('editor', self.editor if self.editor else "NULL")
-        text_node.set('institution', self.institution if self.institution else "NULL")
-        text_node.set('journal', self.journal if self.journal else "NULL")
-        text_node.set('pages', self.pages if self.pages else "NULL")
-        text_node.set('publisher', self.publisher if self.publisher else "NULL")
-        text_node.set('publishing_year', str(self.publishing_year))
-        text_node.set('school', self.school if self.school else "NULL")
-        text_node.set('title', self.title)
-        element_tree.write(file)
-        self.corpus.status = 'unprepared'
-        db.session.commit()
-
     def to_dict(self):
         return {'id': self.id,
                 'address': self.address,
@@ -447,9 +424,6 @@ class Corpus(db.Model):
                 'title': self.title,
                 'user_id': self.user_id}
 
-    def build(self):
-        pass
-
     def delete(self):
         for corpus_file in self.files:
             db.session.delete(corpus_file)

docker-compose.yml

@@ -5,6 +5,9 @@ networks:
     external:
       name: reverse-proxy
 
+volumes:
+  redis-trash1:
+
 services:
   web:
     depends_on:
@@ -52,3 +55,5 @@ services:
       - "/srv/nopaque/db:/var/lib/postgresql/data"
   redis:
     image: redis:5
+    volumes:
+      - "redis-trash1:/data"