def join_corpora(list_name, phase, source_lang, target_lang):
corpus = Corpus(owner_id=user_id, visible=False)
source_lang_id = UserLanguage.query.filter_by(
user_id=user_id, code=source_lang).one().id
for train_corpus in params[list_name]:
corpus_data = json.loads(train_corpus)
corpus_id = corpus_data['id']
corpus_size = corpus_data['size']
if corpus_id not in used_corpora: used_corpora[corpus_id] = 0
try:
og_corpus = Corpus.query.filter_by(id=corpus_id).first()
# We relate the original corpus with this engine in the database,
# for informational purposes. This way the user will be able to know
# which corpora were used to train the engine
engine.engine_corpora.append(
Corpus_Engine(corpus=og_corpus,
engine=engine,
phase=phase,
is_info=True,
selected_size=corpus_size))
corpus.user_source_id = og_corpus.user_source_id
corpus.user_target_id = og_corpus.user_target_id
for file_entry in og_corpus.corpus_files:
with open(file_entry.file.path, 'rb') as file_d:
db_file = data_utils.upload_file(
FileStorage(stream=file_d,
filename=file_entry.file.name),
file_entry.file.user_language_id,
selected_size=corpus_size,
offset=used_corpora[corpus_id],
user_id=user_id)
corpus.corpus_files.append(
Corpus_File(
db_file,
role="source" if file_entry.file.user_language_id
== source_lang_id else "target"))
used_corpora[corpus_id] += corpus_size
except:
raise Exception
try:
db.session.add(corpus)
db.session.commit()
except:
db.session.rollback()
raise Exception
# We put the contents of the several files in a new single one, and we shuffle the sentences
try:
data_utils.join_corpus_files(corpus, shuffle=True, user_id=user_id)
except:
db.session.delete(corpus)
db.session.commit()
raise Exception
return corpus
def test_find(tmpdir):
corpus = Corpus(str(tmpdir))
p = tmpdir.join("Andy_Warhol.rst")
content = ("***********\n"
"Andy Warhol\n"
"***********\n"
"**Andy Warhol** was an artist.")
p.write(content)
doc = corpus.find("Andy Warhol")
assert doc.filename == "Andy_Warhol"
def test_find(tmpdir):
corpus = Corpus(str(tmpdir))
p = tmpdir.join("Andy_Warhol.rst")
content = ("***********\n"
"Andy Warhol\n"
"***********\n"
"**Andy Warhol** was an artist.")
p.write(content)
doc = corpus.find("Andy Warhol")
assert doc.filename == "Andy_Warhol"
def dashboard():
"""admin dashboard page."""
if current_user.is_admin():
corpora = db.session.query(Corpus).all()
control_lists = db.session.query(ControlLists).all()
else:
corpora = Corpus.for_user(current_user)
control_lists = ControlLists.for_user(current_user)
return render_template_with_nav_info('main/dashboard.html',
current_user=current_user,
dashboard_corpora=corpora,
dashboard_control_lists=control_lists)
def test_db_create(self):
""" Test that db is created """
result = self.invoke(["db-create"])
self.assertIn("Created the database", result.output)
with self.app.app_context():
cl = ControlLists(name="Corpus1")
db.session.add(cl)
db.session.flush()
db.session.add(Corpus(name="Corpus1", control_lists_id=cl.id))
db.session.commit()
self.assertEqual(len(Corpus.query.all()), 1,
"There should have been an insert")
def add_n_corpora(self, n_corpus: int, **kwargs):
if not self.AUTO_LOG_IN:
raise Exception("This function only works with autologin")
user = User.query.filter(
User.email == self.app.config['ADMIN_EMAIL']).first()
for n in range(n_corpus):
corpus = Corpus(name="a" * n,
control_lists_id=1,
columns=[
Column(heading="Lemma"),
Column(heading="POS"),
Column(heading="Morph"),
Column(heading="Similar"),
])
new_cu = CorpusUser(corpus=corpus, user=user, is_owner=True)
db.session.add(corpus)
db.session.add(new_cu)
db.session.flush()
db.session.commit()
def process_upload_request(self,
user_id,
bitext_path,
src_path,
trg_path,
src_lang,
trg_lang,
corpus_name,
corpus_desc="",
corpus_topic=None):
type = "bitext" if bitext_path else "bilingual" if trg_path else "monolingual"
def process_file(file, language, corpus, role):
db_file = data_utils.upload_file(file, language, user_id=user_id)
if role == "source":
corpus.user_source_id = language
else:
corpus.user_target_id = language
db.session.add(db_file)
corpus.corpus_files.append(Corpus_File(db_file, role=role))
return db_file
def process_bitext(file):
file_name, file_extension = os.path.splitext(file.filename)
norm_name = utils.normname(user_id=user_id, filename=file_name)
tmp_file_fd, tmp_path = utils.tmpfile()
file.save(tmp_path)
data_utils.convert_file_to_utf8(tmp_path)
data_utils.fix_file(tmp_path)
if file_extension == ".tmx":
with open(utils.filepath('FILES_FOLDER', norm_name + "-src"), 'w') as src_file, \
open(utils.filepath('FILES_FOLDER', norm_name + "-trg"), 'w') as trg_file, \
open(tmp_path, 'rb') as tmx_file:
inside_tuv = False
seg_text = []
tu = []
def se(name, _):
nonlocal inside_tuv
if name == "seg":
inside_tuv = True
def lp(line):
return re.sub(r'[\r\n\t\f\v]', " ", line.strip())
def ee(name):
nonlocal inside_tuv, seg_text, tu, src_file
if name == "seg":
inside_tuv = False
tu.append("".join(seg_text))
seg_text = []
if len(tu) == 2:
print(lp(tu[0]), file=src_file)
print(lp(tu[1]), file=trg_file)
tu = []
def cd(data):
nonlocal inside_tuv, seg_text
if inside_tuv:
seg_text.append(data)
parser = xml.parsers.expat.ParserCreate()
parser.StartElementHandler = se
parser.EndElementHandler = ee
parser.CharacterDataHandler = cd
parser.ParseFile(tmx_file)
else:
# We assume it is a TSV
with open(utils.filepath('FILES_FOLDER', norm_name + "-src"), 'wb') as src_file, \
open(utils.filepath('FILES_FOLDER', norm_name + "-trg"), 'wb') as trg_file, \
open(tmp_path, 'r') as tmp_file:
for line in tmp_file:
cols = line.strip().split('\t')
src_file.write((cols[0] + '\n').encode('utf-8'))
trg_file.write((cols[1] + '\n').encode('utf-8'))
src_file = open(utils.filepath('FILES_FOLDER', norm_name + "-src"),
'rb')
trg_file = open(utils.filepath('FILES_FOLDER', norm_name + "-trg"),
'rb')
return FileStorage(src_file, filename=file.filename + "-src"), \
FileStorage(trg_file, filename=file.filename + "-trg")
# We create the corpus, retrieve the files and attach them to that corpus
target_db_file = None
try:
corpus = Corpus(name=corpus_name,
type="bilingual" if type == "bitext" else type,
owner_id=user_id,
description=corpus_desc,
topic_id=corpus_topic)
if type == "bitext":
with open(bitext_path, 'rb') as fbitext:
bitext_file = FileStorage(fbitext,
filename=os.path.basename(
fbitext.name))
src_file, trg_file = process_bitext(bitext_file)
source_db_file = process_file(src_file, src_lang, corpus,
'source')
target_db_file = process_file(trg_file, trg_lang, corpus,
'target')
else:
with open(src_path, 'rb') as fsrctext:
src_file = FileStorage(fsrctext,
filename=os.path.basename(
fsrctext.name))
source_db_file = process_file(src_file, src_lang, corpus,
'source')
if type == "bilingual":
with open(trg_path, 'rb') as ftrgtext:
trg_file = FileStorage(ftrgtext,
filename=os.path.basename(
ftrgtext.name))
target_db_file = process_file(trg_file, trg_lang, corpus,
'target')
db.session.add(corpus)
user = User.query.filter_by(id=user_id).first()
user.user_corpora.append(LibraryCorpora(corpus=corpus, user=user))
except Exception as e:
db.session.rollback()
raise Exception(
"Something went wrong on our end... Please, try again later")
if target_db_file:
source_lines = utils.file_length(source_db_file.path)
target_lines = utils.file_length(target_db_file.path)
if source_lines != target_lines:
db.session.rollback()
raise Exception(
"Source and target file should have the same length")
db.session.commit()
return True
from app.models import ChangeRecord, WordToken, Corpus, ControlLists
from .base import TestModels
import copy
SimilarityFixtures = [
ControlLists(id=1, name="CL Fixture"),
Corpus(id=1, name="Fixtures !", control_lists_id=1),
WordToken(corpus=1,
form="Cil",
lemma="celui",
left_context="_",
right_context="_",
label_uniform="celui",
morph="smn",
POS="p"), # 1
WordToken(corpus=1,
form="Cil",
lemma="celle",
left_context="_",
right_context="_",
label_uniform="celle",
morph="smn",
POS="n"), # 2
WordToken(corpus=1,
form="Cil",
lemma="cil",
left_context="_",
right_context="_",
label_uniform="cil",
morph="smn",
POS="p"), # 3
from app.models import Corpus, WordToken, AllowedLemma, AllowedPOS, AllowedMorph, Column
from app.models import ControlLists, ControlListsUser
Floovant = Corpus(
name="Floovant",
id=2,
control_lists_id=2
)
FloovantColumns = [
Column(heading="Lemma", corpus_id=2),
Column(heading="POS", corpus_id=2),
Column(heading="Morph", corpus_id=2),
Column(heading="Similar", corpus_id=2),
]
FCL = ControlLists(id=2, name="Floovant")
FloovantTokens = [
WordToken(corpus=Floovant.id, form="SOIGNORS", lemma="seignor", left_context="", right_context="or escoutez que",
label_uniform="seignor", morph="NOMB.=p|GENRE=m|CAS=n"),
WordToken(corpus=Floovant.id, form="or", lemma="or4", left_context="SOIGNORS", right_context="escoutez que Dés",
label_uniform="or4", morph="DEGRE=-"),
WordToken(corpus=Floovant.id, form="escoutez", lemma="escouter", left_context="SOIGNORS or",
right_context="que Dés vos", label_uniform="escouter", morph="MODE=imp|PERS.=2|NOMB.=p"),
WordToken(corpus=Floovant.id, form="que", lemma="que4", left_context="SOIGNORS or escoutez",
right_context="Dés vos soit", label_uniform="que4", morph="_"),
WordToken(corpus=Floovant.id, form="Dés", lemma="dieu", left_context="or escoutez que",
right_context="vos soit amis", label_uniform="dieu", morph="NOMB.=s|GENRE=m|CAS=n"),
WordToken(corpus=Floovant.id, form="vos", lemma="vos1", left_context="escoutez que Dés",
right_context="soit amis III", label_uniform="vos1", morph="PERS.=2|NOMB.=p|GENRE=m|CAS=r"),
WordToken(corpus=Floovant.id, form="soit", lemma="estre1", left_context="que Dés vos",
right_context="amis III vers", label_uniform="estre1", morph="MODE=sub|TEMPS=pst|PERS.=3|NOMB.=s"),
from app.models import Corpus, WordToken, Column
from app.models import ControlLists
control_list = ControlLists(id=3, name="Latin")
corpus = Corpus(
name="Priapees",
id=3,
control_lists_id=control_list.id,
)
PriapeeColumns = [
Column(heading="Lemma", corpus_id=3),
Column(heading="POS", corpus_id=3),
Column(heading="Morph", corpus_id=3),
Column(heading="Similar", corpus_id=3),
]
tokens = [
WordToken(corpus=corpus.id, form="Carminis", lemma="carmen1", POS="NOMcom", left_context="Carminis incompti lusus lecture", right_context="procaces ,", label_uniform="carmen1", morph="Case=Gen|Numb=Sing"),
WordToken(corpus=corpus.id, form="incompti", lemma="incomptus", POS="ADJqua", left_context="Carminis incompti lusus lecture", right_context="procaces , conueniens", label_uniform="incomptus", morph="Case=Gen|Numb=Sing|Deg=Pos"),
WordToken(corpus=corpus.id, form="lusus", lemma="lusus", POS="NOMcom", left_context="Carminis incompti lusus lecture", right_context="procaces , conueniens Latio", label_uniform="lusus", morph="Case=Gen|Numb=Sing"),
WordToken(corpus=corpus.id, form="lecture", lemma="lego?", POS="VER", left_context="Carminis incompti lusus lecture", right_context="procaces , conueniens Latio pone", label_uniform="lego?", morph="Case=Voc|Numb=Sing|Mood=Par|Voice=Act"),
WordToken(corpus=corpus.id, form="procaces", lemma="procax", POS="ADJqua", left_context="Carminis incompti lusus lecture", right_context="procaces , conueniens Latio pone supercilium", label_uniform="procax", morph="Case=Acc|Numb=Plur|Deg=Pos"),
WordToken(corpus=corpus.id, form=",", lemma=",", POS="PUNC", left_context="Carminis incompti lusus lecture", right_context="procaces , conueniens Latio pone supercilium .", label_uniform=",", morph="MORPH=empty"),
WordToken(corpus=corpus.id, form="conueniens", lemma="conueniens", POS="ADJqua", left_context="incompti lusus lecture procaces", right_context=", conueniens Latio pone supercilium . non", label_uniform="conueniens", morph="Case=Nom|Numb=Sing|Deg=Pos"),
WordToken(corpus=corpus.id, form="Latio", lemma="latio", POS="NOMcom", left_context="lusus lecture procaces ,", right_context="conueniens Latio pone supercilium . non soror", label_uniform="latio", morph="Case=Nom|Numb=Sing"),
WordToken(corpus=corpus.id, form="pone", lemma="pono", POS="VER", left_context="lecture procaces , conueniens", right_context="Latio pone supercilium . non soror hoc", label_uniform="pono", morph="Numb=Sing|Mood=Imp|Tense=Pres|Voice=Act|Person=2"),
WordToken(corpus=corpus.id, form="supercilium", lemma="supercilium", POS="NOMcom", left_context="procaces , conueniens Latio", right_context="pone supercilium . non soror hoc habitat", label_uniform="supercilium", morph="Case=Acc|Numb=Sing"),
WordToken(corpus=corpus.id, form=".", lemma=".", POS="PUNC", left_context=", conueniens Latio pone", right_context="supercilium . non soror hoc habitat Phoebi", label_uniform=".", morph="MORPH=empty"),
WordToken(corpus=corpus.id, form="non", lemma="non", POS="ADVneg", left_context="conueniens Latio pone supercilium", right_context=". non soror hoc habitat Phoebi ,", label_uniform="non", morph="MORPH=empty"),
WordToken(corpus=corpus.id, form="soror", lemma="soror", POS="NOMcom", left_context="Latio pone supercilium .", right_context="non soror hoc habitat Phoebi , non", label_uniform="soror", morph="Case=Nom|Numb=Sing"),
WordToken(corpus=corpus.id, form="hoc", lemma="hic1", POS="PROdem", left_context="pone supercilium . non", right_context="soror hoc habitat Phoebi , non uesta", label_uniform="hic1", morph="Case=Nom|Numb=Sing"),
WordToken(corpus=corpus.id, form="habitat", lemma="habito", POS="VER", left_context="supercilium . non soror", right_context="hoc habitat Phoebi , non uesta sacello", label_uniform="habito", morph="Numb=Sing|Mood=Ind|Tense=Pres|Voice=Act|Person=3"),
def process_upload_request(self,
user_id,
bitext_path,
src_path,
trg_path,
src_lang,
trg_lang,
corpus_name,
corpus_desc="",
corpus_topic=None):
type = "bitext" if bitext_path else "bilingual" if trg_path else "monolingual"
def process_file(file, language, corpus, role):
db_file = data_utils.upload_file(file, language, user_id=user_id)
if role == "source":
corpus.source_id = language
else:
corpus.target_id = language
db.session.add(db_file)
corpus.corpus_files.append(Corpus_File(db_file, role=role))
return db_file
def process_bitext(file):
file_name, file_extension = os.path.splitext(file.filename)
norm_name = utils.normname(user_id=user_id, filename=file_name)
tmp_file_fd, tmp_path = utils.tmpfile()
file.save(tmp_path)
if file_extension == ".tmx":
with open(utils.filepath('FILES_FOLDER', norm_name + "-src"), 'wb') as src_file, \
open(utils.filepath('FILES_FOLDER', norm_name + "-trg"), 'wb') as trg_file, \
open(tmp_path, 'r') as tmx_file:
tmx = etree.parse(tmx_file, etree.XMLParser())
body = tmx.getroot().find("body")
for tu in body.findall('.//tu'):
for i, tuv in enumerate(tu.findall('.//tuv')):
if i > 1: break
line = tuv.find("seg").text.strip()
line = re.sub(r'[\r\n\t\f\v]', " ", line)
dest_file = src_file if i == 0 else trg_file
dest_file.write(line.encode('utf-8'))
dest_file.write(os.linesep.encode('utf-8'))
else:
# We assume it is a TSV
with open(utils.filepath('FILES_FOLDER', norm_name + "-src"), 'wb') as src_file, \
open(utils.filepath('FILES_FOLDER', norm_name + "-trg"), 'wb') as trg_file, \
open(tmp_path, 'r') as tmp_file:
for line in tmp_file:
cols = line.strip().split('\t')
src_file.write((cols[0] + '\n').encode('utf-8'))
trg_file.write((cols[1] + '\n').encode('utf-8'))
src_file = open(utils.filepath('FILES_FOLDER', norm_name + "-src"),
'rb')
trg_file = open(utils.filepath('FILES_FOLDER', norm_name + "-trg"),
'rb')
return FileStorage(src_file, filename=file.filename + "-src"), \
FileStorage(trg_file, filename=file.filename + "-trg")
# We create the corpus, retrieve the files and attach them to that corpus
target_db_file = None
try:
corpus = Corpus(name=corpus_name,
type="bilingual" if type == "bitext" else type,
owner_id=user_id,
description=corpus_desc,
topic_id=corpus_topic)
if type == "bitext":
with open(bitext_path, 'rb') as fbitext:
bitext_file = FileStorage(fbitext,
filename=os.path.basename(
fbitext.name))
src_file, trg_file = process_bitext(bitext_file)
source_db_file = process_file(src_file, src_lang, corpus,
'source')
target_db_file = process_file(trg_file, trg_lang, corpus,
'target')
else:
with open(src_path, 'rb') as fsrctext:
src_file = FileStorage(fsrctext,
filename=os.path.basename(
fsrctext.name))
source_db_file = process_file(src_file, src_lang, corpus,
'source')
if type == "bilingual":
with open(trg_path, 'rb') as ftrgtext:
trg_file = FileStorage(ftrgtext,
filename=os.path.basename(
ftrgtext.name))
target_db_file = process_file(trg_file, trg_lang, corpus,
'target')
db.session.add(corpus)
user = User.query.filter_by(id=user_id).first()
user.user_corpora.append(LibraryCorpora(corpus=corpus, user=user))
except Exception as e:
db.session.rollback()
raise Exception(
"Something went wrong on our end... Please, try again later")
if target_db_file:
source_lines = utils.file_length(source_db_file.path)
target_lines = utils.file_length(target_db_file.path)
if source_lines != target_lines:
db.session.rollback()
raise Exception(
"Source and target file should have the same length")
db.session.commit()
return True
