def sentence_stream(limit=None, skip=None, skip_errors=True):
""" Generator of a sentence stream consisting of `limit` sentences (or less) from the
most recently parsed articles. Each sentence is a list of token dicts. """
with SessionContext(commit=True, read_only=True) as session:
q = (session.query(
ArticleRow.url, ArticleRow.parsed,
ArticleRow.tokens).filter(ArticleRow.tokens != None).order_by(
desc(ArticleRow.parsed)).yield_per(200))
count = 0
skipped = 0
for a in q:
doc = json.loads(a.tokens)
for pg in doc:
for sent in pg:
if not sent:
continue
if skip_errors and any("err" in t for t in sent):
# Skip error sentences
continue
if skip is not None and skipped < skip:
# If requested, skip sentences from the front (useful for test set)
skipped += 1
continue
# Yield the sentence as a fresh token list
yield [t for t in sent]
# Are we done?
count += 1
if limit is not None and count >= limit:
return
def token_stream(limit=None, skip_errors=True):
""" Generator of a token stream consisting of `limit` sentences (or less) from the
most recently parsed articles. After each sentence, None is yielded. """
with SessionContext(commit=True, read_only=True) as session:
q = (session.query(
ArticleRow.url, ArticleRow.parsed,
ArticleRow.tokens).filter(ArticleRow.tokens != None).order_by(
desc(ArticleRow.parsed)).yield_per(200))
count = 0
for a in q:
doc = json.loads(a.tokens)
for pg in doc:
for sent in pg:
if not sent:
continue
if skip_errors and any("err" in t for t in sent):
# Skip error sentences
continue
for t in sent:
# Yield the tokens
yield t
yield None # End-of-sentence marker
# Are we done?
count += 1
if limit is not None and count >= limit:
return
def top_persons(limit=_TOP_PERSONS_LENGTH):
""" Return a list of names and titles appearing recently in the news """
toplist = dict()
bindb = BIN_Db.get_db()
with SessionContext(commit=True) as session:
q = session.query(Person.name, Person.title, Person.article_url, Article.id) \
.join(Article).join(Root) \
.filter(Root.visible) \
.order_by(desc(Article.timestamp))[0:limit * 2] # Go through up to 2 * N records
for p in q:
# Insert the name into the list if it's not already there,
# or if the new title is longer than the previous one
if p.name not in toplist or len(p.title) > len(toplist[p.name][0]):
toplist[p.name] = (correct_spaces(p.title), p.article_url,
p.id, bindb.lookup_name_gender(p.name))
if len(toplist) >= limit:
# We now have as many names as we initially wanted: terminate the loop
break
with changedlocale() as strxfrm:
# Convert the dictionary to a sorted list of dicts
return sorted([
dict(name=name, title=tu[0], gender=tu[3], url=tu[1], uuid=tu[2])
for name, tu in toplist.items()
],
key=lambda x: strxfrm(x["name"]))
def top_persons(limit=_TOP_PERSONS_LENGTH):
""" Return a list of names and titles appearing recently in the news """
toplist = dict()
MAX_TITLE_LENGTH = 64
with SessionContext(commit=True) as session:
q = (
session.query(Person.name, Person.title, Person.article_url,
Article.id).join(Article).join(Root).filter(
Root.visible).order_by(desc(Article.timestamp))
[0:limit * 2] # Go through up to 2 * N records
)
def is_better_title(new_title, old_title):
len_new = len(new_title)
len_old = len(old_title)
if len_old >= MAX_TITLE_LENGTH:
# Too long: we want a shorter one
return len_new < len_old
if len_new >= MAX_TITLE_LENGTH:
# This one is too long: we don't want it
return False
# Otherwise, longer is better
return len_new > len_old
with BIN_Db.get_db() as bindb:
for p in q:
# Insert the name into the list if it's not already there,
# or if the new title is longer than the previous one
if p.name not in toplist or is_better_title(
p.title, toplist[p.name][0]):
toplist[p.name] = (
correct_spaces(p.title),
p.article_url,
p.id,
bindb.lookup_name_gender(p.name),
)
if len(toplist) >= limit:
# We now have as many names as we initially wanted: terminate the loop
break
with changedlocale() as strxfrm:
# Convert the dictionary to a sorted list of dicts
return sorted(
[
dict(name=name,
title=tu[0],
gender=tu[3],
url=tu[1],
uuid=tu[2]) for name, tu in toplist.items()
],
key=lambda x: strxfrm(x["name"]),
)
def suggest(limit=10):
""" Return suggestions for query field autocompletion """
limit = request.args.get("limit", limit)
txt = request.args.get("q", "").strip()
suggestions = list()
whois_prefix = "hver er "
whatis_prefix = "hvað er "
prefix = None
if txt.lower().startswith(whois_prefix):
prefix = whois_prefix
elif txt.lower().startswith(whatis_prefix):
prefix = whatis_prefix
if not txt or not prefix:
return better_jsonify(suggestions=suggestions)
with SessionContext(commit=False) as session:
name = txt[len(prefix):].strip()
model_col = None
# Hver er Jón Jónsson ?
if prefix is whois_prefix and name[0].isupper():
model_col = Person.name
# Hver er seðlabankastjóri?
elif prefix is whois_prefix:
model_col = Person.title
# Hvað er UNESCO?
elif prefix is whatis_prefix:
model_col = Entity.name
q = (session.query(model_col,
dbfunc.count(Article.id).label("total")).filter(
model_col.ilike(name + "%")).join(
Article).group_by(model_col).order_by(
desc("total")).limit(limit).all())
prefix = prefix[:1].upper() + prefix[1:].lower()
suggestions = [{"value": (prefix + p[0] + "?"), "data": ""} for p in q]
return better_jsonify(suggestions=suggestions)
def articles(cls, criteria, enclosing_session=None):
""" Generator of Article objects from the database that meet the given criteria """
# The criteria are currently "timestamp", "author" and "domain",
# as well as "order_by_parse" which if True indicates that the result
# should be ordered with the most recently parsed articles first.
with SessionContext(commit=True,
read_only=True,
session=enclosing_session) as session:
# Only fetch articles that have a parse tree
q = session.query(ArticleRow).filter(ArticleRow.tree != None)
# timestamp is assumed to contain a tuple: (from, to)
if criteria and "timestamp" in criteria:
ts = criteria["timestamp"]
q = (q.filter(ArticleRow.timestamp >= ts[0]).filter(
ArticleRow.timestamp < ts[1]))
if criteria and "author" in criteria:
author = criteria["author"]
q = q.filter(ArticleRow.author == author)
if criteria and ("visible" in criteria or "domain" in criteria):
# Need a join with Root for these criteria
q = q.join(Root)
if "visible" in criteria:
# Return only articles from roots with the specified visibility
visible = criteria["visible"]
assert isinstance(visible, bool)
q = q.filter(Root.visible == visible)
if "domain" in criteria:
# Return only articles from the specified domain
domain = criteria["domain"]
assert isinstance(domain, str)
q = q.filter(Root.domain == domain)
if criteria and criteria.get("order_by_parse"):
# Order with newest parses first
q = q.order_by(desc(ArticleRow.parsed))
for arow in q.yield_per(500):
yield cls._init_from_row(arow)
def top_news(topic=None, offset=0, limit=_TOP_NEWS_LENGTH):
""" Return a list of articles (with a particular topic) in
chronologically reversed order. """
toplist = []
topdict = dict()
with SessionContext(commit=True) as session:
q = (session.query(Article).join(Root).filter(
Article.tree != None).filter(Article.timestamp != None).filter(
Article.timestamp <= datetime.utcnow()).filter(
Article.heading > "").filter(
Article.num_sentences > 0).filter(
Root.visible == True))
if topic is not None:
# Filter by topic identifier
q = q.join(ArticleTopic).join(Topic).filter(
Topic.identifier == topic)
q = q.order_by(desc(Article.timestamp)).offset(offset).limit(limit)
class ArticleDisplay:
""" Utility class to carry information about an article to the web template """
def __init__(
self,
heading,
timestamp,
url,
uuid,
num_sentences,
num_parsed,
icon,
localized_date,
source,
):
self.heading = heading
self.timestamp = timestamp
self.url = url
self.uuid = uuid
self.num_sentences = num_sentences
self.num_parsed = num_parsed
self.icon = icon
self.localized_date = localized_date
self.source = source
@property
def width(self):
""" The ratio of parsed sentences to the total number of sentences,
expressed as a percentage string """
if self.num_sentences == 0:
return "0%"
return "{0}%".format(
(100 * self.num_parsed) // self.num_sentences)
@property
def time(self):
return self.timestamp.isoformat()[11:16]
@property
def date(self):
if datetime.today().year == self.timestamp.year:
return self.localized_date
return self.fulldate
@property
def fulldate(self):
return self.localized_date + self.timestamp.strftime(" %Y")
with changedlocale(category="LC_TIME"):
for a in q:
# Instantiate article objects from results
source = a.root.domain
icon = source + ".png"
locdate = a.timestamp.strftime("%-d. %b")
d = ArticleDisplay(
heading=a.heading,
timestamp=a.timestamp,
url=a.url,
uuid=a.id,
num_sentences=a.num_sentences,
num_parsed=a.num_parsed,
icon=icon,
localized_date=locdate,
source=source,
)
toplist.append(d)
return toplist
def top_news(topic=None, start=None, limit=_TOP_NEWS_LENGTH):
""" Return a list of top recent news, of a particular topic,
up to a particular start time, having a specified length """
toplist = []
topdict = dict()
if start is None:
start = datetime.utcnow()
MARGIN = 10 # Get more articles than requested in case there are duplicates
with SessionContext(commit=True) as session:
q = session.query(Article).join(Root) \
.filter(Article.tree != None) \
.filter(Article.timestamp != None) \
.filter(Article.timestamp < start) \
.filter(Article.heading > "") \
.filter(Root.visible == True)
if topic is not None:
# Filter by topic identifier
q = q.join(ArticleTopic).join(Topic).filter(
Topic.identifier == topic)
q = q.order_by(desc(Article.timestamp))[0:limit + MARGIN]
class ArticleDisplay:
""" Utility class to carry information about an article to the web template """
def __init__(self, heading, timestamp, url, uuid, num_sentences,
num_parsed, icon):
self.heading = heading
self.timestamp = timestamp
self.url = url
self.uuid = uuid
self.num_sentences = num_sentences
self.num_parsed = num_parsed
self.icon = icon
@property
def width(self):
""" The ratio of parsed sentences to the total number of sentences,
expressed as a percentage string """
if self.num_sentences == 0:
return "0%"
return "{0}%".format(
(100 * self.num_parsed) // self.num_sentences)
@property
def time(self):
return self.timestamp.isoformat()[11:16]
@property
def date(self):
return self.timestamp.isoformat()[0:10]
for a in q:
# Collect and count the titles
icon = a.root.domain + ".ico"
d = ArticleDisplay(heading=a.heading,
timestamp=a.timestamp,
url=a.url,
uuid=a.id,
num_sentences=a.num_sentences,
num_parsed=a.num_parsed,
icon=icon)
# Have we seen the same heading on the same domain?
t = (a.root.domain, a.heading)
if t in topdict:
# Same domain+heading already in the list
i = topdict[t]
if d.timestamp > toplist[i].timestamp:
# The new entry is newer: replace the old one
toplist[i] = d
# Otherwise, ignore the new entry and continue
else:
# New heading: note its index in the list
llist = len(toplist)
topdict[t] = llist
toplist.append(d)
if llist + 1 >= limit:
break
return toplist[0:limit]
