1
1
mirror of https://github.com/MarginaliaSearch/MarginaliaSearch.git synced 2025-10-05 21:22:39 +02:00

Compare commits

...

10 Commits

Author SHA1 Message Date
Viktor Lofgren
e34230c25b (search) Dyslexia fix 2025-01-12 20:39:59 +01:00
Viktor Lofgren
ef3f175ede (search) Don't clobber the search query URL with default values 2025-01-10 15:57:30 +01:00
Viktor Lofgren
bbe4b5d9fd Revert experimental changes 2025-01-10 15:52:02 +01:00
Viktor Lofgren
c67a635103 (search, experimental) Add a few debugging tracks to the search UI 2025-01-10 15:44:44 +01:00
Viktor Lofgren
20b24133fb (search, experimental) Add a few debugging tracks to the search UI 2025-01-10 15:34:48 +01:00
Viktor Lofgren
f2567677e8 (index-client) Clean up index client code
Improve error handling.  This should be a relatively rare case, but we don't want one bad index partition to blow up the entire query.
2025-01-10 15:17:07 +01:00
Viktor Lofgren
bc2c2061f2 (index-client) Clean up index client code
This should have the rpc stream reception be performed in parallel in separate threads, rather blocking sequentially in the main thread, hopefully giving a slight performance boost.
2025-01-10 15:14:42 +01:00
Viktor Lofgren
1c7f5a31a5 (search) Further reduce the number of db queries by adding more caching to DbDomainQueries. 2025-01-10 14:17:29 +01:00
Viktor Lofgren
59a8ea60f7 (search) Further reduce the number of db queries by adding more caching to DbDomainQueries. 2025-01-10 14:15:22 +01:00
Viktor Lofgren
aa9b1244ea (search) Reduce the number of db queries a bit by caching data that doesn't change too often 2025-01-10 13:56:04 +01:00
7 changed files with 121 additions and 91 deletions

View File

@@ -20,7 +20,10 @@ public class DbDomainQueries {
private final HikariDataSource dataSource;
private static final Logger logger = LoggerFactory.getLogger(DbDomainQueries.class);
private final Cache<EdgeDomain, Integer> domainIdCache = CacheBuilder.newBuilder().maximumSize(10_000).build();
private final Cache<Integer, EdgeDomain> domainNameCache = CacheBuilder.newBuilder().maximumSize(10_000).build();
private final Cache<String, List<DomainWithNode>> siblingsCache = CacheBuilder.newBuilder().maximumSize(10_000).build();
@Inject
public DbDomainQueries(HikariDataSource dataSource)
@@ -30,16 +33,21 @@ public class DbDomainQueries {
public Integer getDomainId(EdgeDomain domain) throws NoSuchElementException {
try (var connection = dataSource.getConnection()) {
try {
return domainIdCache.get(domain, () -> {
try (var stmt = connection.prepareStatement("SELECT ID FROM EC_DOMAIN WHERE DOMAIN_NAME=?")) {
try (var connection = dataSource.getConnection();
var stmt = connection.prepareStatement("SELECT ID FROM EC_DOMAIN WHERE DOMAIN_NAME=?")) {
stmt.setString(1, domain.toString());
var rsp = stmt.executeQuery();
if (rsp.next()) {
return rsp.getInt(1);
}
}
catch (SQLException ex) {
throw new RuntimeException(ex);
}
throw new NoSuchElementException();
});
}
@@ -49,9 +57,6 @@ public class DbDomainQueries {
catch (ExecutionException ex) {
throw new RuntimeException(ex.getCause());
}
catch (SQLException ex) {
throw new RuntimeException(ex);
}
}
public OptionalInt tryGetDomainId(EdgeDomain domain) {
@@ -84,47 +89,55 @@ public class DbDomainQueries {
}
public Optional<EdgeDomain> getDomain(int id) {
try (var connection = dataSource.getConnection()) {
EdgeDomain existing = domainNameCache.getIfPresent(id);
if (existing != null) {
return Optional.of(existing);
}
try (var connection = dataSource.getConnection()) {
try (var stmt = connection.prepareStatement("SELECT DOMAIN_NAME FROM EC_DOMAIN WHERE ID=?")) {
stmt.setInt(1, id);
var rsp = stmt.executeQuery();
if (rsp.next()) {
return Optional.of(new EdgeDomain(rsp.getString(1)));
var val = new EdgeDomain(rsp.getString(1));
domainNameCache.put(id, val);
return Optional.of(val);
}
return Optional.empty();
}
}
catch (UncheckedExecutionException ex) {
throw new RuntimeException(ex.getCause());
}
catch (SQLException ex) {
throw new RuntimeException(ex);
}
}
public List<DomainWithNode> otherSubdomains(EdgeDomain domain, int cnt) {
List<DomainWithNode> ret = new ArrayList<>();
public List<DomainWithNode> otherSubdomains(EdgeDomain domain, int cnt) throws ExecutionException {
String topDomain = domain.topDomain;
try (var conn = dataSource.getConnection();
var stmt = conn.prepareStatement("SELECT DOMAIN_NAME, NODE_AFFINITY FROM EC_DOMAIN WHERE DOMAIN_TOP = ? LIMIT ?")) {
stmt.setString(1, domain.topDomain);
stmt.setInt(2, cnt);
return siblingsCache.get(topDomain, () -> {
List<DomainWithNode> ret = new ArrayList<>();
var rs = stmt.executeQuery();
while (rs.next()) {
var sibling = new EdgeDomain(rs.getString(1));
try (var conn = dataSource.getConnection();
var stmt = conn.prepareStatement("SELECT DOMAIN_NAME, NODE_AFFINITY FROM EC_DOMAIN WHERE DOMAIN_TOP = ? LIMIT ?")) {
stmt.setString(1, topDomain);
stmt.setInt(2, cnt);
if (sibling.equals(domain))
continue;
var rs = stmt.executeQuery();
while (rs.next()) {
var sibling = new EdgeDomain(rs.getString(1));
ret.add(new DomainWithNode(sibling, rs.getInt(2)));
if (sibling.equals(domain))
continue;
ret.add(new DomainWithNode(sibling, rs.getInt(2)));
}
} catch (SQLException e) {
logger.error("Failed to get domain neighbors");
}
} catch (SQLException e) {
logger.error("Failed to get domain neighbors");
}
return ret;
});
return ret;
}
public record DomainWithNode (EdgeDomain domain, int nodeAffinity) {

View File

@@ -16,20 +16,19 @@ import org.slf4j.LoggerFactory;
import java.util.ArrayList;
import java.util.Comparator;
import java.util.Iterator;
import java.util.List;
import java.util.concurrent.CompletableFuture;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import static java.lang.Math.clamp;
import java.util.concurrent.atomic.AtomicInteger;
import java.util.function.Consumer;
@Singleton
public class IndexClient {
private static final Logger logger = LoggerFactory.getLogger(IndexClient.class);
private final GrpcMultiNodeChannelPool<IndexApiGrpc.IndexApiBlockingStub> channelPool;
private final DomainBlacklistImpl blacklist;
private static final ExecutorService executor = Executors.newVirtualThreadPerTaskExecutor();
private static final ExecutorService executor = Executors.newCachedThreadPool();
@Inject
public IndexClient(GrpcChannelPoolFactory channelPoolFactory, DomainBlacklistImpl blacklist) {
@@ -51,40 +50,37 @@ public class IndexClient {
/** Execute a query on the index partitions and return the combined results. */
public AggregateQueryResponse executeQueries(RpcIndexQuery indexRequest, Pagination pagination) {
List<CompletableFuture<Iterator<RpcDecoratedResultItem>>> futures =
channelPool.call(IndexApiGrpc.IndexApiBlockingStub::query)
.async(executor)
.runEach(indexRequest);
final int requestedMaxResults = indexRequest.getQueryLimits().getResultsTotal();
final int resultsUpperBound = requestedMaxResults * channelPool.getNumNodes();
List<RpcDecoratedResultItem> results = new ArrayList<>(resultsUpperBound);
AtomicInteger totalNumResults = new AtomicInteger(0);
for (var future : futures) {
try {
future.get().forEachRemaining(results::add);
}
catch (Exception e) {
logger.error("Downstream exception", e);
}
}
List<RpcDecoratedResultItem> results =
channelPool.call(IndexApiGrpc.IndexApiBlockingStub::query)
.async(executor)
.runEach(indexRequest)
.stream()
.map(future -> future.thenApply(iterator -> {
List<RpcDecoratedResultItem> ret = new ArrayList<>(requestedMaxResults);
iterator.forEachRemaining(ret::add);
totalNumResults.addAndGet(ret.size());
return ret;
}))
.mapMulti((CompletableFuture<List<RpcDecoratedResultItem>> fut, Consumer<List<RpcDecoratedResultItem>> c) ->{
try {
c.accept(fut.join());
} catch (Exception e) {
logger.error("Error while fetching results", e);
}
})
.flatMap(List::stream)
.filter(item -> !isBlacklisted(item))
.sorted(comparator)
.skip(Math.max(0, (pagination.page - 1) * pagination.pageSize))
.limit(pagination.pageSize)
.toList();
// Sort the results by ranking score and remove blacklisted domains
results.sort(comparator);
results.removeIf(this::isBlacklisted);
int numReceivedResults = results.size();
// pagination is typically 1-indexed, so we need to adjust the start and end indices
int indexStart = (pagination.page - 1) * pagination.pageSize;
int indexEnd = (pagination.page) * pagination.pageSize;
results = results.subList(
clamp(indexStart, 0, Math.max(0, results.size() - 1)), // from is inclusive, so subtract 1 from size()
clamp(indexEnd, 0, results.size()));
return new AggregateQueryResponse(results, pagination.page(), numReceivedResults);
return new AggregateQueryResponse(results, pagination.page(), totalNumResults.get());
}
private boolean isBlacklisted(RpcDecoratedResultItem item) {

View File

@@ -84,18 +84,33 @@ public record SearchParameters(WebsiteUrl url,
}
public String renderUrl() {
String path = String.format("/search?query=%s&profile=%s&js=%s&adtech=%s&recent=%s&searchTitle=%s&newfilter=%s&page=%d",
URLEncoder.encode(query, StandardCharsets.UTF_8),
URLEncoder.encode(profile.filterId, StandardCharsets.UTF_8),
URLEncoder.encode(js.value, StandardCharsets.UTF_8),
URLEncoder.encode(adtech.value, StandardCharsets.UTF_8),
URLEncoder.encode(recent.value, StandardCharsets.UTF_8),
URLEncoder.encode(searchTitle.value, StandardCharsets.UTF_8),
Boolean.valueOf(newFilter).toString(),
page
);
return path;
StringBuilder pathBuilder = new StringBuilder("/search?");
pathBuilder.append("query=").append(URLEncoder.encode(query, StandardCharsets.UTF_8));
if (profile != SearchProfile.NO_FILTER) {
pathBuilder.append("&profile=").append(URLEncoder.encode(profile.filterId, StandardCharsets.UTF_8));
}
if (js != SearchJsParameter.DEFAULT) {
pathBuilder.append("&js=").append(URLEncoder.encode(js.value, StandardCharsets.UTF_8));
}
if (adtech != SearchAdtechParameter.DEFAULT) {
pathBuilder.append("&adtech=").append(URLEncoder.encode(adtech.value, StandardCharsets.UTF_8));
}
if (recent != SearchRecentParameter.DEFAULT) {
pathBuilder.append("&recent=").append(URLEncoder.encode(recent.value, StandardCharsets.UTF_8));
}
if (searchTitle != SearchTitleParameter.DEFAULT) {
pathBuilder.append("&searchTitle=").append(URLEncoder.encode(searchTitle.value, StandardCharsets.UTF_8));
}
if (page != 1) {
pathBuilder.append("&page=").append(page);
}
if (newFilter) {
pathBuilder.append("&newfilter=").append(Boolean.valueOf(newFilter).toString());
}
return pathBuilder.toString();
}
public RpcTemporalBias.Bias temporalBias() {

View File

@@ -3,27 +3,22 @@ package nu.marginalia.search.command.commands;
import com.google.inject.Inject;
import io.jooby.MapModelAndView;
import io.jooby.ModelAndView;
import nu.marginalia.search.JteRenderer;
import nu.marginalia.search.SearchOperator;
import nu.marginalia.search.command.SearchCommandInterface;
import nu.marginalia.search.command.SearchParameters;
import nu.marginalia.search.model.DecoratedSearchResults;
import nu.marginalia.search.model.NavbarModel;
import java.io.IOException;
import java.util.Map;
import java.util.Optional;
public class SearchCommand implements SearchCommandInterface {
private final SearchOperator searchOperator;
private final JteRenderer jteRenderer;
@Inject
public SearchCommand(SearchOperator searchOperator,
JteRenderer jteRenderer) throws IOException {
public SearchCommand(SearchOperator searchOperator){
this.searchOperator = searchOperator;
this.jteRenderer = jteRenderer;
}
@Override

View File

@@ -28,6 +28,7 @@ import org.slf4j.LoggerFactory;
import java.sql.SQLException;
import java.util.*;
import java.util.concurrent.CompletableFuture;
import java.util.concurrent.ExecutionException;
import java.util.concurrent.Future;
import java.util.concurrent.TimeUnit;
import java.util.function.Supplier;
@@ -71,7 +72,7 @@ public class SearchSiteInfoService {
Thread.ofPlatform().name("Recently Added Domains Model Updater").start(this::modelUpdater);
}
private volatile SiteOverviewModel model = new SiteOverviewModel(List.of());
private volatile SiteOverviewModel cachedOverviewModel = new SiteOverviewModel(List.of());
@GET
@Path("/site")
@@ -83,7 +84,7 @@ public class SearchSiteInfoService {
return new MapModelAndView("siteinfo/start.jte",
Map.of("navbar", NavbarModel.SITEINFO,
"model", model));
"model", cachedOverviewModel));
}
private void modelUpdater() {
@@ -95,17 +96,26 @@ public class SearchSiteInfoService {
// the result.
try (var conn = dataSource.getConnection();
var stmt = conn.prepareStatement("SELECT DOMAIN_NAME, DISCOVER_DATE FROM EC_DOMAIN WHERE NODE_AFFINITY = 0 ORDER BY ID DESC LIMIT 10")) {
var stmt = conn.prepareStatement("""
SELECT DOMAIN_NAME, DISCOVER_DATE
FROM EC_DOMAIN
WHERE NODE_AFFINITY = 0
ORDER BY ID DESC
LIMIT 10
"""))
{
var rs = stmt.executeQuery();
while (rs.next()) {
domains.add(new SiteOverviewModel.DiscoveredDomain(rs.getString("DOMAIN_NAME"), rs.getString("DISCOVER_DATE")));
domains.add(new SiteOverviewModel.DiscoveredDomain(
rs.getString("DOMAIN_NAME"),
rs.getString("DISCOVER_DATE"))
);
}
} catch (SQLException ex) {
throw new RuntimeException();
logger.warn("Failed to get recently added domains: {}", ex.getMessage());
}
model = new SiteOverviewModel(domains);
cachedOverviewModel = new SiteOverviewModel(domains);
try {
TimeUnit.MINUTES.sleep(15);
@@ -127,7 +137,7 @@ public class SearchSiteInfoService {
@PathParam String domainName,
@QueryParam String view,
@QueryParam Integer page
) throws SQLException {
) throws SQLException, ExecutionException {
if (null == domainName || domainName.isBlank()) {
return null;
@@ -213,7 +223,7 @@ public class SearchSiteInfoService {
);
}
private SiteInfoWithContext listInfo(Context context, String domainName) {
private SiteInfoWithContext listInfo(Context context, String domainName) throws ExecutionException {
var domain = new EdgeDomain(domainName);
final int domainId = domainQueries.tryGetDomainId(domain).orElse(-1);

View File

@@ -36,10 +36,11 @@
</div>
@if (filters.showRecentOption.isSet()) <input type="hidden" name="js" value="${filters.removeJsOption.value()}"> @endif
@if (filters.reduceAdtechOption.isSet()) <input type="hidden" name="adtech" value="${filters.reduceAdtechOption.value()}"> @endif
@if (filters.searchTitleOption.isSet()) <input type="hidden" name="searchTitle" value="${filters.searchTitleOption.value()}"> @endif
@if (filters.showRecentOption.isSet()) <input type="hidden" name="recent" value="${filters.showRecentOption.value()}"> @endif
<input type="hidden" name="js" value="${filters.removeJsOption.value()}">
<input type="hidden" name="adtech" value="${filters.reduceAdtechOption.value()}">
<input type="hidden" name="searchTitle" value="${filters.searchTitleOption.value()}">
<input type="hidden" name="profile" value="${profile}">
<input type="hidden" name="recent" value="${filters.showRecentOption.value()}">
</form>

View File

@@ -36,7 +36,7 @@
<div class="text-slate-700 dark:text-white text-sm p-4">
<div class="fas fa-gift mr-1 text-margeblue dark:text-slate-200"></div>
This is the new design and home of Marginalia Search.
You can about what this entails <a href="https://about.marginalia-search.com/article/redesign/" class="underline text-liteblue dark:text-blue-200">here</a>.
You can read about what this entails <a href="https://about.marginalia-search.com/article/redesign/" class="underline text-liteblue dark:text-blue-200">here</a>.
<p class="my-4"></p>
The old version of Marginalia Search remains available at
<a href="https://old-search.marginalia.nu/" class="underline text-liteblue dark:text-blue-200">https://old-search.marginalia.nu/</a>.