From dfd853a7965d5b5a753d12a45a360ae7599e86c7 Mon Sep 17 00:00:00 2001
From: Daniel O'Connell <github@ahiru.pl>
Date: Tue, 15 Aug 2023 14:32:13 +0200
Subject: [PATCH] filter out rows with empty text

---
 align_data/common/alignment_dataset.py  |  6 ++++++
 align_data/db/models.py                 | 21 +++++++++++++++++++++
 align_data/db/session.py                | 10 +++++++---
 align_data/sources/articles/datasets.py | 11 ++++++-----
 4 files changed, 40 insertions(+), 8 deletions(-)

diff --git a/align_data/common/alignment_dataset.py b/align_data/common/alignment_dataset.py
index 8bbd224e..5ff55f63 100644
--- a/align_data/common/alignment_dataset.py
+++ b/align_data/common/alignment_dataset.py
@@ -202,6 +202,12 @@ def fetch_entries(self):
             if not entry:
                 continue
 
+            try:
+                entry.verify_id_fields()
+            except AssertionError as e:
+                logger.error(e)
+                continue
+
             yield entry
 
             if self.COOLDOWN:
diff --git a/align_data/db/models.py b/align_data/db/models.py
index afbd27fd..6cb3878f 100644
--- a/align_data/db/models.py
+++ b/align_data/db/models.py
@@ -16,10 +16,12 @@
 )
 from sqlalchemy.orm import DeclarativeBase, Mapped, mapped_column, relationship
 from sqlalchemy.dialects.mysql import LONGTEXT
+from sqlalchemy.ext.hybrid import hybrid_property
 from align_data.settings import PINECONE_METADATA_KEYS
 
 
 logger = logging.getLogger(__name__)
+OK_STATUS = None
 
 
 class Base(DeclarativeBase):
@@ -131,6 +133,25 @@ def add_meta(self, key, val):
             self.meta = {}
         self.meta[key] = val
 
+    @hybrid_property
+    def is_valid(self):
+        return (
+            self.text and self.text.strip() and
+            self.url and self.title and
+            self.authors is not None and
+            self.status == OK_STATUS
+        )
+
+    @is_valid.expression
+    def is_valid(cls):
+        return (
+            (cls.status == OK_STATUS)
+            & (cls.text != None)
+            & (cls.url != None)
+            & (cls.title != None)
+            & (cls.authors != None)
+        )
+
     @classmethod
     def before_write(cls, mapper, connection, target):
         target.verify_id_fields()
diff --git a/align_data/db/session.py b/align_data/db/session.py
index ace0ff8a..55e03e53 100644
--- a/align_data/db/session.py
+++ b/align_data/db/session.py
@@ -2,7 +2,7 @@
 import logging
 
 from contextlib import contextmanager
-from sqlalchemy import create_engine
+from sqlalchemy import create_engine, and_
 from sqlalchemy.orm import Session
 from align_data.settings import DB_CONNECTION_URI
 from align_data.db.models import Article
@@ -23,5 +23,9 @@ def make_session(auto_commit=False):
 def stream_pinecone_updates(session, custom_sources: List[str]):
     """Yield Pinecone entries that require an update."""
     yield from session.query(Article).filter(
-        Article.pinecone_update_required.is_(True)
-    ).filter(Article.source.in_(custom_sources)).yield_per(1000)
+        Article.pinecone_update_required.is_(True),
+    ).filter(
+        Article.is_valid
+    ).filter(
+        Article.source.in_(custom_sources)
+    ).yield_per(1000)
diff --git a/align_data/sources/articles/datasets.py b/align_data/sources/articles/datasets.py
index d032903f..f8e2c897 100644
--- a/align_data/sources/articles/datasets.py
+++ b/align_data/sources/articles/datasets.py
@@ -6,6 +6,7 @@
 from urllib.parse import urlparse
 
 import pandas as pd
+from align_data.sources.articles import articles
 from gdown.download import download
 from markdownify import markdownify
 from pypandoc import convert_file
@@ -123,16 +124,12 @@ def not_processed(self, item):
         )
 
     def process_entry(self, item):
-        if parse_domain(item.url) == "arxiv.org":
+        if ArxivPapers.is_arxiv(item.url):
             contents = ArxivPapers.get_contents(item)
             contents['source'] = 'arxiv'
         else:
             contents = self.get_contents(item)
 
-        # Skip items that can't be saved because missing fields
-        if not all(contents.get(key) for key in ['title', 'url']):
-            return None
-
         return self.make_data_entry(contents)
 
 
@@ -218,6 +215,10 @@ def _get_text(self, item):
 class ArxivPapers(SpreadsheetDataset):
     COOLDOWN: int = 1
 
+    @staticmethod
+    def is_arxiv(url):
+        return parse_domain(url) == "arxiv.org"
+
     @classmethod
     def get_contents(cls, item) -> Dict:
         contents = fetch_arxiv(item.url or item.source_url)