First draft of new trending scores for solr.

conf/solr/conf/managed-schema.xml

@@ -200,6 +200,44 @@
     <field name="ratings_count_4" type="pint"/>
     <field name="ratings_count_5" type="pint"/>
 
+
+    <!-- Trending related values-->
+    <field name = "trending_z_score" type="pfloat" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_0" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_1" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_2" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_3" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_4" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_5" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_6" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_7" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_8" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_9" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_10" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_11" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_12" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_13" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_14" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_15" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_16" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_17" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_18" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_19" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_20" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_21" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_22" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_23" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_daily_0" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_daily_1" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_daily_2" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_daily_3" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_daily_4" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_daily_5" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_daily_6" type="pint" indexed="false" stored ="false"/>
+    <field name = "trending_score_hourly_sum" type="pint" indexed = "false" stored = "false" />
+
+
+
     <!-- Reading Log -->
     <field name="readinglog_count" type="pint"/>
     <field name="want_to_read_count" type="pint"/>

openlibrary/solr/solr_types.py

@@ -90,36 +90,37 @@ class SolrDocument(TypedDict):
     lending_edition_s: Optional[str]
     ia_collection_s: Optional[str]
     ebook_count_i: Optional[int]
-    trending_hourly_0: Optional[int]
-    trending_hourly_1: Optional[int]
-    trending_hourly_2: Optional[int]
-    trending_hourly_3: Optional[int]
-    trending_hourly_4: Optional[int]
-    trending_hourly_5: Optional[int]
-    trending_hourly_6: Optional[int]
-    trending_hourly_7: Optional[int]
-    trending_hourly_8: Optional[int]
-    trending_hourly_9: Optional[int]
-    trending_hourly_10: Optional[int]
-    trending_hourly_11: Optional[int]
-    trending_hourly_12: Optional[int]
-    trending_hourly_13: Optional[int]
-    trending_hourly_14: Optional[int]
-    trending_hourly_15: Optional[int]
-    trending_hourly_16: Optional[int]
-    trending_hourly_17: Optional[int]
-    trending_hourly_18: Optional[int]
-    trending_hourly_19: Optional[int]
-    trending_hourly_20: Optional[int]
-    trending_hourly_21: Optional[int]
-    trending_hourly_22: Optional[int]
-    trending_hourly_23: Optional[int]
-    trending_daily_0: Optional[int]
-    trending_daily_1: Optional[int]
-    trending_daily_2: Optional[int]
-    trending_daily_3: Optional[int]
-    trending_daily_4: Optional[int]
-    trending_daily_5: Optional[int]
-    trending_daily_6: Optional[int]
+    trending_score_hourly_0: Optional[int]
+    trending_score_hourly_1: Optional[int]
+    trending_score_hourly_2: Optional[int]
+    trending_score_hourly_3: Optional[int]
+    trending_score_hourly_4: Optional[int]
+    trending_score_hourly_5: Optional[int]
+    trending_score_hourly_6: Optional[int]
+    trending_score_hourly_7: Optional[int]
+    trending_score_hourly_8: Optional[int]
+    trending_score_hourly_9: Optional[int]
+    trending_score_hourly_10: Optional[int]
+    trending_score_hourly_11: Optional[int]
+    trending_score_hourly_12: Optional[int]
+    trending_score_hourly_13: Optional[int]
+    trending_score_hourly_14: Optional[int]
+    trending_score_hourly_15: Optional[int]
+    trending_score_hourly_16: Optional[int]
+    trending_score_hourly_17: Optional[int]
+    trending_score_hourly_18: Optional[int]
+    trending_score_hourly_19: Optional[int]
+    trending_score_hourly_20: Optional[int]
+    trending_score_hourly_21: Optional[int]
+    trending_score_hourly_22: Optional[int]
+    trending_score_hourly_23: Optional[int]
+    trending_score_hourly_sum: Optional[int]
+    trending_score_daily_0: Optional[int]
+    trending_score_daily_1: Optional[int]
+    trending_score_daily_2: Optional[int]
+    trending_score_daily_3: Optional[int]
+    trending_score_daily_4: Optional[int]
+    trending_score_daily_5: Optional[int]
+    trending_score_daily_6: Optional[int]
     trending_z_score: Optional[float]
 # fmt: on

openlibrary/solr/update.py

@@ -53,6 +53,10 @@ def can_update_key(key: str) -> bool:
     return any(updater.key_test(key) for updater in get_solr_updaters())
 
 
+async def in_place_update():
+    pass
+
+
 async def update_keys(
     keys: list[str],
     commit=True,

openlibrary/solr/updater/work.py

@@ -684,6 +684,9 @@ def build_trending_scores(self) -> dict:
             )
             for index in range(24)
         }
+        doc |= {
+            "trending_score_hourly_sum": self._work.get("trending_score_hourly_sum", 0)
+        }
         doc |= {
             f'trending_score_daily_{index}': self._work.get(
                 f'trending_score_daily_{index}', 0

openlibrary/utils/solr.py

@@ -79,6 +79,16 @@ def get_many(
         ).json()
         return [doc_wrapper(doc) for doc in resp['response']['docs']]
 
+    def update_in_place(
+        self,
+        request,
+    ):
+        resp = requests.post(
+            f'{self.base_url}/update?update.partial.requireInPlace=true,commit=true',
+            json=request,
+        ).json()
+        return resp
+
     def select(
         self,
         query,

scripts/calculate_trending_scores_daily.py

@@ -0,0 +1,52 @@
+import _init_path
+import os
+from openlibrary.config import load_config
+from openlibrary.core import db
+from openlibrary.plugins.worksearch.code import execute_solr_query
+import datetime
+
+from openlibrary.plugins.worksearch.search import get_solr
+
+
+def fetch_works(current_day: int):
+    resp = execute_solr_query(
+        '/export',
+        {
+            "q": f'trending_score_hourly_sum:[1 TO *] OR trending_score_daily_{current_day}:[1 TO *]',
+            "fl": "key,trending_score_hourly_sum",
+            "sort": "trending_score_hourly_sum desc",
+        },
+    )
+    doc_data = {}
+    if resp:
+        data = resp.json()
+        try:
+            docs = data["response"]["docs"]
+        except KeyError:
+            raise KeyError
+        print(docs)
+        doc_data = {doc["key"]: doc.get("trending_score_hourly_sum", 0) for doc in docs}
+    return doc_data
+
+
+def form_inplace_updates(work_id: str, current_day: int, new_value: int):
+    return {"key": work_id, f'trending_score_daily_{current_day}': {"set": new_value}}
+
+
+if __name__ == '__main__':
+    from contextlib import redirect_stdout
+
+    ol_config = os.getenv("OL_CONFIG")
+    if ol_config:
+        with open(os.devnull, 'w') as devnull, redirect_stdout(devnull):
+            load_config(ol_config)
+    current_day = datetime.datetime.now().weekday()
+    work_data = fetch_works(current_day)
+    print(work_data)
+    request_body = [
+        form_inplace_updates(work_id, current_day, work_data[work_id])
+        for work_id in work_data
+    ]
+    print(request_body)
+    resp = get_solr().update_in_place(request_body)
+    print(resp)

scripts/calculate_trending_scores_hourly.py

@@ -1,17 +1,100 @@
 import _init_path
 import os
+import datetime
 from openlibrary.config import load_config
 from openlibrary.core import db
-from openlibrary.plugins.worksearch.code import get_solr_works
+from openlibrary.plugins.worksearch.search import get_solr
+from openlibrary.plugins.worksearch.code import execute_solr_query
+from math import sqrt
 
+# This script handles hourly updating of each works' z-score. The 'trending_score_hourly_23' field is
+# ignored, and the current count of bookshelves events in the last hour is the new trending_score_hourly[0]
+# value, with each other value incrementing up by one.
 
-def fetch_works(work_ids: list[int]):
-    fields = [f'trending_score_hourly_{index}' for index in range(23)]
-    fields.extend([f'trending_score_daily_{index}' for index in range(7)])
+# Trending_score_daily values are all fetched in order to calculate their mean and standard deviation.
+# The formula for the new z-score is [ z = sum(trending_score_hourly_*) - mean(trending_score_daily_*) /  (standard_deviation(trending_score_daily)] + 1).
+# The incrementation of the denominator by 1 serves several purposes:
+#       - It prevents divide-by-zero errors on objects with 0s in all trending_score_daily_* values, which is important,
+#         as they are instantiated to that upon the first time solr reindexes them with this new information.
+#       - It serves to deemphasize such values, as we're not particularly interested in a shift from
+#         0 events to 1 event.
 
 
-def update_hourly_count_and_z_score(work_id: int, count: int):
-    pass
+def fetch_works(current_hour: int):
+    ol_db = db.get_db()
+    query = "select work_id, Count(updated)"
+    query += "from bookshelves_events group by bookshelves_events.updated, work_id "
+    query += "having updated >= localtimestamp - interval '1 hour'"
+    db_data = {
+        f'/works/OL{storage.work_id}W': storage.count
+        for storage in list(ol_db.query(query))
+    }
+    print(db_data)
+    query = f'trending_score_hourly_{current_hour}:[1 TO *] OR {" OR ".join(["key:\"" + key + "\"" for key in db_data])}'
+    print(query)
+    resp = execute_solr_query(
+        '/export',
+        {
+            "q": query,
+            "fl": ",".join(
+                [
+                    "key",
+                    "trending_score_hourly_sum",
+                    f'trending_score_hourly_{current_hour}',
+                    "trending_score_daily_*",
+                ]
+            ),
+            "sort": "trending_score_hourly_sum desc",
+        },
+    )
+    doc_data = {}
+    if resp:
+        data = resp.json()
+        try:
+            docs = data["response"]["docs"]
+
+        except KeyError:
+            raise KeyError
+        print(docs)
+        doc_data = {
+            doc["key"]: {"count": db_data.get(doc["key"], 0), "solr_doc": doc}
+            for doc in docs
+        }
+    return doc_data
+
+
+def get_z_score(solr_doc: dict, count: int, current_hour: int):
+    arith_mean = sum([solr_doc[f'trending_score_daily_{i}'] for i in range(7)]) / 7
+    last_24_hours_value = (
+        solr_doc['trending_score_hourly_sum']
+        + count
+        - solr_doc[f'trending_score_hourly_{current_hour}']
+    )
+    st_dev = sqrt(
+        sum(
+            [
+                pow(solr_doc[f'trending_score_daily_{i}'] - arith_mean, 2)
+                for i in range(7)
+            ]
+        )
+        / 7
+    )
+    return (last_24_hours_value - arith_mean) / (st_dev + 1.0)
+
+
+def form_inplace_updates(work_id: str, count: int, solr_doc: dict, current_hour: int):
+    request_body = {
+        "key": work_id,
+        f'trending_score_hourly_{current_hour}': {"set": count},
+        "trending_score_hourly_sum": {
+            "set": solr_doc["trending_score_hourly_sum"]
+            - solr_doc[f'trending_score_hourly_{current_hour}']
+            + count
+        },
+        "trending_z_score": {"set": get_z_score(solr_doc, count, current_hour)},
+    }
+
+    return request_body
 
 
 if __name__ == '__main__':
@@ -21,12 +104,21 @@ def update_hourly_count_and_z_score(work_id: int, count: int):
     if ol_config:
         with open(os.devnull, 'w') as devnull, redirect_stdout(devnull):
             load_config(ol_config)
-        ol_db = db.get_db()
-        query = "select work_id, Count(updated)"
-        query += "from bookshelves_events group by bookshelves_events.updated, work_id"
-        query += "having updated >= localtimestamp - interval '1 hour'"
-        print(ol_db.__dir__())
-        ids_needing_changes = list(ol_db.query(query))
-        print(ids_needing_changes[0].work_id)
+
+        current_hour = datetime.datetime.now().hour
+        work_data = fetch_works(current_hour)
+        if work_data:
+            request_body = [
+                form_inplace_updates(
+                    work_id,
+                    work_data[work_id]["count"],
+                    work_data[work_id]["solr_doc"],
+                    current_hour,
+                )
+                for work_id in work_data
+            ]
+            print(request_body)
+            resp = get_solr().update_in_place(request_body)
+            print(resp)
     else:
         print("failure")