More text_all search tweaks

- on reprovision, first do migration, then delete+insert so the
correct schema is ensured
- remove the fuzzy operator which doesn't play well with multiple
tokens in a query
- don't split words on numbers, retaining those unique words like
`a56bd3e` used in our tests

Author: eikek

components/renku_data_services/search/reprovision.py

@@ -102,6 +102,9 @@ def log_counter(c: int) -> None:
             logger.info(f"Starting reprovisioning with ID {reprovisioning.id}")
             started = datetime.now()
             await self._search_updates_repo.clear_all()
+            if migrate_solr_schema:
+                await migrator.migrate(entity_schema.all_migrations)
+
             async with DefaultSolrClient(self._solr_config) as client:
                 res = await client.delete("_type:*")
                 if res.status_code != 200:
@@ -119,9 +122,6 @@ def log_counter(c: int) -> None:
                             exc_info=False,
                         )
 
-            if migrate_solr_schema:
-                await migrator.migrate(entity_schema.all_migrations)
-
             all_users = self._user_repo.get_all_users(requested_by=admin)
             counter = await self.__update_entities(all_users, "user", started, counter, log_counter)
             logger.info(f"Done adding user entities to search_updates table. Record count: {counter}.")

components/renku_data_services/search/solr_token.py

@@ -186,7 +186,7 @@ def public_only() -> SolrToken:
 
 def content_all(text: str) -> SolrToken:
     """Search the content_all field."""
-    terms_str = "(" + __escape_query(text) + "~)"
+    terms_str = "(" + __escape_query(text) + ")"
     return SolrToken(f"{Fields.content_all}:{terms_str}")
 
 

components/renku_data_services/solr/solr_schema.py

@@ -83,7 +83,13 @@ class Filters:
     flattenGraph = Filter(name="flattenGraph")
     word = Filter(
         name="wordDelimiterGraph",
-        settings={"splitOnCaseChange": "1", "catenateNumbers": "1", "catenateAll": "1", "preserveOriginal": "1"},
+        settings={
+            "splitOnCaseChange": "1",
+            "catenateNumbers": "1",
+            "catenateAll": "1",
+            "preserveOriginal": "1",
+            "splitOnNumerics": "0",
+        },
     )
 
     @classmethod

test/bases/renku_data_services/data_api/test_search.py

@@ -236,31 +236,6 @@ async def test_projects(
     assert_search_result(result, [p1, p3, p2], check_order=True)
 
 
-# TODO: figure out how to run search tests fully parallel
-@pytest.mark.xdist_group("search")
-@pytest.mark.asyncio
-async def test_distance(
-    search_reprovision: SearchReprovisionCall,
-    create_project_model: CreateProjectCall,
-    search_query: SearchQueryCall,
-    sanic_client_with_solr: SanicASGITestClient,
-    app_manager_instance: TestDependencyManager,
-) -> None:
-    """Search should be lenient to simple typos, distance=2."""
-    p1 = await create_project_model(
-        sanic_client_with_solr, "Project Bike Z", visibility="public", description="a bike with a bike"
-    )
-    await search_reprovision(app_manager_instance)
-
-    result = await search_query(sanic_client_with_solr, "mikin type:project")
-    assert result.items == []
-
-    result = await search_query(sanic_client_with_solr, "mike type:project")
-    assert result.items is not None
-    assert len(result.items) == 1
-    assert __entity_id(result.items[0]) == p1.id
-
-
 # TODO: figure out how to run search tests fully parallel
 @pytest.mark.xdist_group("search")
 @pytest.mark.asyncio
@@ -303,15 +278,28 @@ async def test_search_project_with_dash(
     sanic_client_with_solr: SanicASGITestClient,
     app_manager_instance: TestDependencyManager,
 ) -> None:
-    p1 = await create_project_model(sanic_client_with_solr, "project-with-dash")
+    p1 = await create_project_model(sanic_client_with_solr, "project-with-dash-a0ie84eb")
     await search_reprovision(app_manager_instance)
 
+    result = await search_query(sanic_client_with_solr, "project-with-dash-a0ie84eb", regular_user)
+    assert_search_result(result, [p1], check_order=False)
     result = await search_query(sanic_client_with_solr, "project-with-dash", regular_user)
     assert_search_result(result, [p1], check_order=False)
-    result = await search_query(sanic_client_with_solr, "project-with-hash", regular_user)
+    result = await search_query(sanic_client_with_solr, "project-with", regular_user)
+    assert_search_result(result, [p1], check_order=False)
+
+    result = await search_query(sanic_client_with_solr, "project", regular_user)
+    assert_search_result(result, [p1], check_order=False)
+    result = await search_query(sanic_client_with_solr, "with", regular_user)
+    assert_search_result(result, [p1], check_order=False)
+    result = await search_query(sanic_client_with_solr, "dash", regular_user)
+    assert_search_result(result, [p1], check_order=False)
+    result = await search_query(sanic_client_with_solr, "a0ie84eb", regular_user)
     assert_search_result(result, [p1], check_order=False)
 
 
+
+
 # TODO: figure out how to run search tests fully parallel
 @pytest.mark.xdist_group("search")
 @pytest.mark.asyncio
@@ -344,6 +332,7 @@ async def test_search_domain_and_email(
     assert_search_result(result, [p1], check_order=False)
 
 
+
 def __entity_id(e: SearchEntity) -> str:
     match e.root:
         case SearchProject() as p:

test/components/renku_data_services/search/test_solr_token.py

@@ -130,7 +130,7 @@ def test_created_by_exists() -> None:
 
 
 def test_content_all() -> None:
-    assert st.content_all("abc") == "content_all:(abc~)"
-    assert st.content_all("a+b+c") == "content_all:(a\\+b\\+c~)"
-    assert st.content_all("ab cd") == "content_all:(ab\\ cd~)"
-    assert st.content_all("ab    cd") == "content_all:(ab\\ \\ \\ \\ cd~)"
+    assert st.content_all("abc") == "content_all:(abc)"
+    assert st.content_all("a+b+c") == "content_all:(a\\+b\\+c)"
+    assert st.content_all("ab cd") == "content_all:(ab\\ cd)"
+    assert st.content_all("ab    cd") == "content_all:(ab\\ \\ \\ \\ cd)"