add enrich most references option (#35)

* Add option to pull most references

* Change to enrich references

* Update version

Author: odarbelaeze

src/bibx/__init__.py

@@ -5,7 +5,7 @@
 
 from bibx.algorithms.sap import Sap
 from bibx.article import Article
-from bibx.builders.openalex import HandleReferences, OpenAlexCollectionBuilder
+from bibx.builders.openalex import EnrichReferences, OpenAlexCollectionBuilder
 from bibx.builders.scopus_bib import ScopusBibCollectionBuilder
 from bibx.builders.scopus_ris import ScopusRisCollectionBuilder
 from bibx.builders.wos import WosCollectionBuilder
@@ -17,7 +17,7 @@
 __all__ = [
     "Article",
     "Collection",
-    "HandleReferences",
+    "EnrichReferences",
     "Sap",
     "query_openalex",
     "read_any",
@@ -26,16 +26,16 @@
     "read_wos",
 ]
 
-__version__ = "0.4.1"
+__version__ = "0.5.0"
 
 
 def query_openalex(
     query: str,
     limit: int = 600,
-    references: HandleReferences = HandleReferences.BASIC,
+    enrich: EnrichReferences = EnrichReferences.BASIC,
 ) -> Collection:
     """Query OpenAlex and return a collection."""
-    return OpenAlexCollectionBuilder(query, limit, references=references).build()
+    return OpenAlexCollectionBuilder(query, limit, enrich=enrich).build()
 
 
 def read_scopus_bib(*files: TextIO) -> Collection:

src/bibx/builders/openalex.py

@@ -12,14 +12,16 @@
 
 logger = logging.getLogger(__name__)
 
-MAX_REFERENCES = 400
+_COMMON_REFERENCES = 400
+_MOST_REFERENCES = 2000
 
 
-class HandleReferences(Enum):
+class EnrichReferences(Enum):
     """How to handle references when building an openalex collection."""
 
     BASIC = "basic"
     COMMON = "common"
+    MOST = "most"
     FULL = "full"
 
 
@@ -30,12 +32,12 @@ def __init__(
         self,
         query: str,
         limit: int = 600,
-        references: HandleReferences = HandleReferences.BASIC,
+        enrich: EnrichReferences = EnrichReferences.BASIC,
         client: Optional[OpenAlexClient] = None,
     ) -> None:
         self.query = query
         self.limit = limit
-        self.references = references
+        self.enrich = enrich
         self.client = client or OpenAlexClient()
 
     def build(self) -> Collection:
@@ -44,21 +46,23 @@ def build(self) -> Collection:
         works = self.client.list_recent_articles(self.query, self.limit)
         cache = {work.id: work for work in works}
         references: list[str] = []
+        missing = set()
         for work in works:
             references.extend(work.referenced_works)
-        if self.references == HandleReferences.COMMON:
+        if self.enrich in (EnrichReferences.COMMON, EnrichReferences.MOST):
             counter = Counter(references)
-            most_common = {key for key, _ in counter.most_common(MAX_REFERENCES)}
+            count = (
+                _MOST_REFERENCES
+                if self.enrich == EnrichReferences.MOST
+                else _COMMON_REFERENCES
+            )
+            most_common = {key for key, _ in counter.most_common(count)}
             missing = most_common - set(cache.keys())
-            logger.info("fetching %d missing references", len(missing))
-            missing_works = self.client.list_articles_by_openalex_id(list(missing))
-            cache.update({work.id: work for work in missing_works})
-        if self.references == HandleReferences.FULL:
+        if self.enrich == EnrichReferences.FULL:
             missing = set(references) - set(cache.keys())
-            logger.info("fetching %d missing references", len(missing))
-            missing_works = self.client.list_articles_by_openalex_id(list(missing))
-            cache.update({work.id: work for work in missing_works})
-
+        logger.info("fetching %d missing references", len(missing))
+        missing_works = self.client.list_articles_by_openalex_id(list(missing))
+        cache.update({work.id: work for work in missing_works})
         article_cache = {
             openalexid: self._work_to_article(work)
             for openalexid, work in cache.items()

src/bibx/cli.py

@@ -16,7 +16,7 @@
 )
 from bibx.algorithms.preprocess import Preprocess
 from bibx.algorithms.sap import Sap
-from bibx.builders.openalex import HandleReferences
+from bibx.builders.openalex import EnrichReferences
 from bibx.collection import Collection
 
 app = typer.Typer()
@@ -80,9 +80,9 @@ def sap(filename: str) -> None:
 @app.command()
 def openalex(
     query: list[str],
-    references: HandleReferences = typer.Option(
+    enrich: EnrichReferences = typer.Option(
         help="how to handle references",
-        default=HandleReferences.BASIC,
+        default=EnrichReferences.BASIC,
     ),
     verbose: bool = typer.Option(
         help="be more verbose",
@@ -92,7 +92,7 @@ def openalex(
     """Run the sap algorithm on a seed file of any supported format."""
     if verbose:
         logging.basicConfig(level=logging.INFO)
-    c = query_openalex(" ".join(query), references=references)
+    c = query_openalex(" ".join(query), enrich=enrich)
     s = Sap()
     graph = s.create_graph(c)
     graph = s.clean_graph(graph)

src/bibx/clients/openalex.py

@@ -171,6 +171,8 @@ def list_recent_articles(self, query: str, limit: int = 600) -> list[Work]:
 
     def list_articles_by_openalex_id(self, ids: list[str]) -> list[Work]:
         """List articles by openalex id."""
+        if not ids:
+            return []
         select = ",".join(Work.model_fields.keys())
         results: list[Work] = []
         with ThreadPoolExecutor(max_workers=5) as executor: