google-nomulus/java/google/registry/rdap/RdapDomainSearchAction.java

// Copyright 2017 The Nomulus Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

package google.registry.rdap;

import static google.registry.model.EppResourceUtils.loadByForeignKey;
import static google.registry.model.index.ForeignKeyIndex.loadAndGetKey;
import static google.registry.model.ofy.ObjectifyService.ofy;
import static google.registry.request.Action.Method.GET;
import static google.registry.request.Action.Method.HEAD;
import static google.registry.util.DateTimeUtils.START_OF_TIME;

import com.google.common.collect.ImmutableList;
import com.google.common.collect.ImmutableMap;
import com.google.common.collect.ImmutableSortedSet;
import com.google.common.collect.Iterables;
import com.google.common.collect.Streams;
import com.google.common.flogger.FluentLogger;
import com.google.common.net.InetAddresses;
import com.google.common.primitives.Booleans;
import com.googlecode.objectify.Key;
import com.googlecode.objectify.cmd.Query;
import google.registry.model.domain.DomainResource;
import google.registry.model.host.HostResource;
import google.registry.rdap.RdapJsonFormatter.BoilerplateType;
import google.registry.rdap.RdapJsonFormatter.OutputDataType;
import google.registry.rdap.RdapMetrics.EndpointType;
import google.registry.rdap.RdapMetrics.SearchType;
import google.registry.rdap.RdapMetrics.WildcardType;
import google.registry.rdap.RdapSearchResults.IncompletenessWarningType;
import google.registry.request.Action;
import google.registry.request.HttpException.BadRequestException;
import google.registry.request.HttpException.NotFoundException;
import google.registry.request.HttpException.UnprocessableEntityException;
import google.registry.request.Parameter;
import google.registry.request.auth.Auth;
import google.registry.util.Idn;
import google.registry.util.NonFinalForTesting;
import java.net.InetAddress;
import java.util.ArrayList;
import java.util.Comparator;
import java.util.List;
import java.util.Optional;
import java.util.stream.Stream;
import javax.inject.Inject;
import org.joda.time.DateTime;

/**
 * RDAP (new WHOIS) action for domain search requests.
 *
 * <p>All commands and responses conform to the RDAP spec as defined in RFCs 7480 through 7485.
 *
 * @see <a href="http://tools.ietf.org/html/rfc7482">RFC 7482: Registration Data Access Protocol
 *     (RDAP) Query Format</a>
 * @see <a href="http://tools.ietf.org/html/rfc7483">RFC 7483: JSON Responses for the Registration
 *     Data Access Protocol (RDAP)</a>
 */
@Action(
  path = RdapDomainSearchAction.PATH,
  method = {GET, HEAD},
  auth = Auth.AUTH_PUBLIC
)
public class RdapDomainSearchAction extends RdapSearchActionBase {

  static final String PATH = "/rdap/domains";

  static final int RESULT_SET_SIZE_SCALING_FACTOR = 30;

  @NonFinalForTesting
  static int maxNameserversInFirstStage = 300;

  private static final FluentLogger logger = FluentLogger.forEnclosingClass();

  @Inject @Parameter("name") Optional<String> nameParam;
  @Inject @Parameter("nsLdhName") Optional<String> nsLdhNameParam;
  @Inject @Parameter("nsIp") Optional<String> nsIpParam;
  @Inject public RdapDomainSearchAction() {}

  @Override
  public String getHumanReadableObjectTypeName() {
    return "domain search";
  }

  @Override
  public EndpointType getEndpointType() {
    return EndpointType.DOMAINS;
  }

  @Override
  public String getActionPath() {
    return PATH;
  }

  /**
   * Parses the parameters and calls the appropriate search function.
   *
   * <p>The RDAP spec allows for domain search by domain name, nameserver name or nameserver IP.
   */
  @Override
  public ImmutableMap<String, Object> getJsonObjectForResource(
      String pathSearchString, boolean isHeadRequest) {
    DateTime now = clock.nowUtc();
    // RDAP syntax example: /rdap/domains?name=exam*.com.
    // The pathSearchString is not used by search commands.
    if (pathSearchString.length() > 0) {
      throw new BadRequestException("Unexpected path");
    }
    if (Booleans.countTrue(nameParam.isPresent(), nsLdhNameParam.isPresent(), nsIpParam.isPresent())
        != 1) {
      throw new BadRequestException(
          "You must specify either name=XXXX, nsLdhName=YYYY or nsIp=ZZZZ");
    }
    decodeCursorToken();
    RdapSearchResults results;
    if (nameParam.isPresent()) {
      metricInformationBuilder.setSearchType(SearchType.BY_DOMAIN_NAME);
      // syntax: /rdap/domains?name=exam*.com
      String asciiName;
      try {
        asciiName = Idn.toASCII(nameParam.get());
      } catch (Exception e) {
        throw new BadRequestException("Invalid value of nsLdhName parameter");
      }
      results = searchByDomainName(
          recordWildcardType(RdapSearchPattern.create(asciiName, true)), now);
    } else if (nsLdhNameParam.isPresent()) {
      metricInformationBuilder.setSearchType(SearchType.BY_NAMESERVER_NAME);
      // syntax: /rdap/domains?nsLdhName=ns1.exam*.com
      // RFC 7482 appears to say that Unicode domains must be specified using punycode when
      // passed to nsLdhName, so IDN.toASCII is not called here.
      if (!LDH_PATTERN.matcher(nsLdhNameParam.get()).matches()) {
        throw new BadRequestException("Invalid value of nsLdhName parameter");
      }
      results = searchByNameserverLdhName(
          recordWildcardType(RdapSearchPattern.create(nsLdhNameParam.get(), true)), now);
    } else {
      metricInformationBuilder.setSearchType(SearchType.BY_NAMESERVER_ADDRESS);
      metricInformationBuilder.setWildcardType(WildcardType.NO_WILDCARD);
      metricInformationBuilder.setPrefixLength(nsIpParam.get().length());
      // syntax: /rdap/domains?nsIp=1.2.3.4
      InetAddress inetAddress;
      try {
        inetAddress = InetAddresses.forString(nsIpParam.get());
      } catch (IllegalArgumentException e) {
        throw new BadRequestException("Invalid value of nsIp parameter");
      }
      results = searchByNameserverIp(inetAddress, now);
    }
    if (results.jsonList().isEmpty()) {
      throw new NotFoundException("No domains found");
    }
    ImmutableMap.Builder<String, Object> builder = new ImmutableMap.Builder<>();
    builder.put("domainSearchResults", results.jsonList());
    rdapJsonFormatter.addTopLevelEntries(
        builder,
        BoilerplateType.DOMAIN,
        getNotices(results),
        ImmutableList.of(),
        fullServletPath);
    return builder.build();
  }

  /**
   * Searches for domains by domain name, returning a JSON array of domain info maps.
   *
   * <p>Domain query strings with wildcards are allowed to have a suffix after the wildcard, which
   * must be a TLD. If the TLD is not present, the wildcard must be preceded by at least two
   * characters (e.g. "ex*"), to avoid queries for all domains in the system. If the TLD is present,
   * the initial string is not required (e.g. "*.tld" is valid), because the search will be
   * restricted to a single TLD.
   *
   * <p>Searches which include deleted entries are effectively treated as if they have a wildcard,
   * since the same name can return multiple results.
   */
  private RdapSearchResults searchByDomainName(
      final RdapSearchPattern partialStringQuery, final DateTime now) {
    // Handle queries without a wildcard -- just load by foreign key. We can't do this if deleted
    // entries are included, because there may be multiple nameservers with the same name.
    if (!partialStringQuery.getHasWildcard() && !shouldIncludeDeleted()) {
      return searchByDomainNameWithoutWildcard(partialStringQuery, now);
    }
    // Handle queries with a wildcard and initial search string. We require either a TLD or an
    // initial string at least MIN_INITIAL_STRING_LENGTH long.
    if (!partialStringQuery.getInitialString().isEmpty()) {
      if ((partialStringQuery.getSuffix() == null)
          && (partialStringQuery.getInitialString().length()
              < RdapSearchPattern.MIN_INITIAL_STRING_LENGTH)) {
        throw new UnprocessableEntityException(
            String.format(
                "Initial search string must be at least %d characters for wildcard domain searches"
                    + " without a TLD suffix",
                RdapSearchPattern.MIN_INITIAL_STRING_LENGTH));
      }
      return searchByDomainNameWithInitialString(partialStringQuery, now);
    }
    if (partialStringQuery.getSuffix() == null) {
      throw new UnprocessableEntityException(
          "Initial search string is required for wildcard domain searches without a TLD suffix");
    }
    return searchByDomainNameByTld(partialStringQuery.getSuffix(), now);
  }

  /**
   * Searches for domains by domain name without a wildcard or interest in deleted entries.
   */
  private RdapSearchResults searchByDomainNameWithoutWildcard(
      final RdapSearchPattern partialStringQuery, final DateTime now) {
    DomainResource domainResource =
        loadByForeignKey(DomainResource.class, partialStringQuery.getInitialString(), now);
    ImmutableList<DomainResource> results =
        ((domainResource == null) || !shouldBeVisible(domainResource, now))
            ? ImmutableList.of()
            : ImmutableList.of(domainResource);
    return makeSearchResults(results, now);
  }

  /** Searches for domains by domain name with an initial string, wildcard and possible suffix. */
  private RdapSearchResults searchByDomainNameWithInitialString(
      final RdapSearchPattern partialStringQuery, final DateTime now) {
    // We can't query for undeleted domains as part of the query itself; that would require an
    // inequality query on deletion time, and we are already using inequality queries on
    // fullyQualifiedDomainName. So we instead pick an arbitrary limit of
    // RESULT_SET_SIZE_SCALING_FACTOR times the result set size limit, fetch up to that many, and
    // weed out all deleted domains. If there still isn't a full result set's worth of domains, we
    // give up and return just the ones we found. Don't use queryItems, because it checks that the
    // initial string is at least a certain length, which we don't need in this case. Query the
    // domains directly, rather than the foreign keys, because then we have an index on TLD if we
    // need it.
    int querySizeLimit = RESULT_SET_SIZE_SCALING_FACTOR * rdapResultSetMaxSize;
    Query<DomainResource> query =
        ofy()
            .load()
            .type(DomainResource.class)
            .filter("fullyQualifiedDomainName <", partialStringQuery.getNextInitialString())
            .filter("fullyQualifiedDomainName >=", partialStringQuery.getInitialString());
    if (cursorString.isPresent()) {
      query = query.filter("fullyQualifiedDomainName >", cursorString.get());
    }
    if (partialStringQuery.getSuffix() != null) {
      query = query.filter("tld", partialStringQuery.getSuffix());
    }
    query = query.limit(querySizeLimit);
    // Always check for visibility, because we couldn't look at the deletionTime in the query.
    return makeSearchResults(getMatchingResources(query, true, now, querySizeLimit), now);
  }

  /** Searches for domains by domain name with a TLD suffix. */
  private RdapSearchResults searchByDomainNameByTld(String tld, DateTime now) {
    // Even though we are not searching on fullyQualifiedDomainName, we want the results to come
    // back ordered by name, so we are still in the same boat as
    // searchByDomainNameWithInitialString, unable to perform an inequality query on deletion time.
    // Don't use queryItems, because it doesn't handle pending deletes.
    int querySizeLimit = RESULT_SET_SIZE_SCALING_FACTOR * rdapResultSetMaxSize;
    Query<DomainResource> query =
        ofy()
            .load()
            .type(DomainResource.class)
            .filter("tld", tld);
    if (cursorString.isPresent()) {
      query = query.filter("fullyQualifiedDomainName >", cursorString.get());
    }
    query = query.order("fullyQualifiedDomainName").limit(querySizeLimit);
    return makeSearchResults(getMatchingResources(query, true, now, querySizeLimit), now);
  }

  /**
   * Searches for domains by nameserver name, returning a JSON array of domain info maps.
   *
   * <p>This is a two-step process: get a list of host references by host name, and then look up
   * domains by host reference.
   *
   * <p>The includeDeleted parameter does NOT cause deleted nameservers to be searched, only deleted
   * domains which used to be connected to an undeleted nameserver.
   */
  private RdapSearchResults searchByNameserverLdhName(
      final RdapSearchPattern partialStringQuery, final DateTime now) {
    Iterable<Key<HostResource>> hostKeys = getNameserverRefsByLdhName(partialStringQuery, now);
    if (Iterables.isEmpty(hostKeys)) {
      metricInformationBuilder.setNumHostsRetrieved(0);
      throw new NotFoundException("No matching nameservers found");
    }
    return searchByNameserverRefs(hostKeys, now);
  }

  /**
   * Assembles a list of {@link HostResource} keys by name.
   *
   * <p>Nameserver query strings with wildcards are allowed to have a suffix after the wildcard,
   * which must be a domain. If the domain is not specified, or is not an existing domain in one of
   * our TLDs, the wildcard must be preceded by at least two characters (e.g. "ns*"), to avoid
   * queries for all nameservers in the system. If the suffix specifies an existing domain, the
   * initial string is not required (e.g. "*.example.tld" is valid), because we can look up the
   * domain and just list all of its subordinate hosts.
   */
  private Iterable<Key<HostResource>> getNameserverRefsByLdhName(
      final RdapSearchPattern partialStringQuery, final DateTime now) {
    // Handle queries without a wildcard.
    if (!partialStringQuery.getHasWildcard()) {
      return getNameserverRefsByLdhNameWithoutWildcard(partialStringQuery, now);
    }
    // Handle queries with a wildcard and suffix (specifying a suprerordinate domain).
    if (partialStringQuery.getSuffix() != null) {
      return getNameserverRefsByLdhNameWithSuffix(partialStringQuery, now);
    }
    // If there's no suffix, query the host resources. Query the resources themselves, rather than
    // the foreign key indexes, because then we have an index on fully qualified host name and
    // deletion time, so we can check the deletion status in the query itself. The initial string
    // must be present, to avoid querying every host in the system. This restriction is enforced by
    // {@link queryItems}.
    //
    // Only return the first maxNameserversInFirstStage nameservers. This could result in an
    // incomplete result set if a search asks for something like "ns*", but we need to enforce a
    // limit in order to avoid arbitrarily long-running queries.
    Query<HostResource> query =
        queryItems(
            HostResource.class,
            "fullyQualifiedHostName",
            partialStringQuery,
            DeletedItemHandling.EXCLUDE,
            maxNameserversInFirstStage);
    Optional<String> desiredRegistrar = getDesiredRegistrar();
    if (desiredRegistrar.isPresent()) {
      query = query.filter("currentSponsorClientId", desiredRegistrar.get());
    }
    return query.keys();
  }

  /** Assembles a list of {@link HostResource} keys by name when the pattern has no wildcard. */
  private Iterable<Key<HostResource>> getNameserverRefsByLdhNameWithoutWildcard(
      final RdapSearchPattern partialStringQuery, final DateTime now) {
    // If we need to check the sponsoring registrar, we need to load the resource rather than just
    // the key.
    Optional<String> desiredRegistrar = getDesiredRegistrar();
    if (desiredRegistrar.isPresent()) {
      HostResource host =
          loadByForeignKey(
              HostResource.class,
              partialStringQuery.getInitialString(),
              shouldIncludeDeleted() ? START_OF_TIME : now);
      return ((host == null)
              || !desiredRegistrar.get().equals(host.getPersistedCurrentSponsorClientId()))
          ? ImmutableList.of()
          : ImmutableList.of(Key.create(host));
    } else {
      Key<HostResource> hostKey =
          loadAndGetKey(
              HostResource.class,
              partialStringQuery.getInitialString(),
              shouldIncludeDeleted() ? START_OF_TIME : now);
      return (hostKey == null) ? ImmutableList.of() : ImmutableList.of(hostKey);
    }
  }

  /** Assembles a list of {@link HostResource} keys by name using a superordinate domain suffix. */
  private Iterable<Key<HostResource>> getNameserverRefsByLdhNameWithSuffix(
      final RdapSearchPattern partialStringQuery, final DateTime now) {
    // The suffix must be a domain that we manage. That way, we can look up the domain and search
    // through the subordinate hosts. This is more efficient, and lets us permit wildcard searches
    // with no initial string.
    DomainResource domainResource =
        loadByForeignKey(
            DomainResource.class,
            partialStringQuery.getSuffix(),
            shouldIncludeDeleted() ? START_OF_TIME : now);
    if (domainResource == null) {
      // Don't allow wildcards with suffixes which are not domains we manage. That would risk a
      // table scan in some easily foreseeable cases.
      throw new UnprocessableEntityException(
          "A suffix in a lookup by nameserver name must be a domain defined in the system");
    }
    Optional<String> desiredRegistrar = getDesiredRegistrar();
    ImmutableList.Builder<Key<HostResource>> builder = new ImmutableList.Builder<>();
    for (String fqhn : ImmutableSortedSet.copyOf(domainResource.getSubordinateHosts())) {
      // We can't just check that the host name starts with the initial query string, because
      // then the query ns.exam*.example.com would match against nameserver ns.example.com.
      if (partialStringQuery.matches(fqhn)) {
        if (desiredRegistrar.isPresent()) {
          HostResource host =
              loadByForeignKey(
                  HostResource.class, fqhn, shouldIncludeDeleted() ? START_OF_TIME : now);
          if ((host != null)
              && desiredRegistrar.get().equals(host.getPersistedCurrentSponsorClientId())) {
            builder.add(Key.create(host));
          }
        } else {
          Key<HostResource> hostKey =
              loadAndGetKey(HostResource.class, fqhn, shouldIncludeDeleted() ? START_OF_TIME : now);
          if (hostKey != null) {
            builder.add(hostKey);
          } else {
            logger.atWarning().log("Host key unexpectedly null");
          }
        }
      }
    }
    return builder.build();
  }

  /**
   * Searches for domains by nameserver address, returning a JSON array of domain info maps.
   *
   * <p>This is a two-step process: get a list of host references by IP address, and then look up
   * domains by host reference.
   *
   * <p>In theory, we could have any number of hosts using the same IP address. To make sure we get
   * all the associated domains, we have to retrieve all of them, and use them to look up domains.
   * This could open us up to a kind of DoS attack if huge number of hosts are defined on a single
   * IP. To avoid this, fetch only the first {@link #maxNameserversInFirstStage} nameservers. In
   * all normal circumstances, this should be orders of magnitude more than there actually are. But
   * it could result in us missing some domains.
   *
   * <p>The includeDeleted parameter does NOT cause deleted nameservers to be searched, only deleted
   * domains which used to be connected to an undeleted nameserver.
   */
  private RdapSearchResults searchByNameserverIp(
      final InetAddress inetAddress, final DateTime now) {
    Query<HostResource> query =
        queryItems(
            HostResource.class,
            "inetAddresses",
            inetAddress.getHostAddress(),
            Optional.empty(),
            Optional.empty(),
            DeletedItemHandling.EXCLUDE,
            maxNameserversInFirstStage);
    Optional<String> desiredRegistrar = getDesiredRegistrar();
    if (desiredRegistrar.isPresent()) {
      query = query.filter("currentSponsorClientId", desiredRegistrar.get());
    }
    return searchByNameserverRefs(query.keys(), now);
  }

  /**
   * Locates all domains which are linked to a set of host keys.
   *
   * <p>This method is called by {@link #searchByNameserverLdhName} and {@link
   * #searchByNameserverIp} after they assemble the relevant host keys.
   */
  private RdapSearchResults searchByNameserverRefs(
      final Iterable<Key<HostResource>> hostKeys, final DateTime now) {
    // We must break the query up into chunks, because the in operator is limited to 30 subqueries.
    // Since it is possible for the same domain to show up more than once in our result list (if
    // we do a wildcard nameserver search that returns multiple nameservers used by the same
    // domain), we must create a set of resulting {@link DomainResource} objects. Use a sorted set,
    // and fetch all domains, to make sure that we can return the first domains in alphabetical
    // order.
    ImmutableSortedSet.Builder<DomainResource> domainSetBuilder =
        ImmutableSortedSet.orderedBy(
            Comparator.comparing(DomainResource::getFullyQualifiedDomainName));
    int numHostKeysSearched = 0;
    for (List<Key<HostResource>> chunk : Iterables.partition(hostKeys, 30)) {
      numHostKeysSearched += chunk.size();
      Query<DomainResource> query = ofy().load()
          .type(DomainResource.class)
          .filter("nsHosts in", chunk);
      if (!shouldIncludeDeleted()) {
        query = query.filter("deletionTime >", now);
      // If we are not performing an inequality query, we can filter on the cursor in the query.
      // Otherwise, we will need to filter the results afterward.
      } else if (cursorString.isPresent()) {
        query = query.filter("fullyQualifiedDomainName >", cursorString.get());
      }
      Stream<DomainResource> stream =
          Streams.stream(query).filter(domain -> isAuthorized(domain, now));
      if (cursorString.isPresent()) {
        stream =
            stream.filter(
                domain -> (domain.getFullyQualifiedDomainName().compareTo(cursorString.get()) > 0));
      }
      stream.forEach(domainSetBuilder::add);
    }
    List<DomainResource> domains = domainSetBuilder.build().asList();
    metricInformationBuilder.setNumHostsRetrieved(numHostKeysSearched);
    if (domains.size() > rdapResultSetMaxSize) {
      return makeSearchResults(
          domains.subList(0, rdapResultSetMaxSize),
          IncompletenessWarningType.TRUNCATED,
          Optional.of((long) domains.size()),
          now);
    } else {
      // If everything that we found will fit in the result, check whether there might have been
      // more results that got dropped because the first stage limit on number of nameservers. If
      // so, indicate the result might be incomplete.
      return makeSearchResults(
          domains,
          (numHostKeysSearched >= maxNameserversInFirstStage)
              ? IncompletenessWarningType.MIGHT_BE_INCOMPLETE
              : IncompletenessWarningType.COMPLETE,
          (numHostKeysSearched > 0) ? Optional.of((long) domains.size()) : Optional.empty(),
          now);
    }
  }

  /** Output JSON for a list of domains, with no incompleteness warnings. */
  private RdapSearchResults makeSearchResults(List<DomainResource> domains, DateTime now) {
    return makeSearchResults(
        domains, IncompletenessWarningType.COMPLETE, Optional.of((long) domains.size()), now);
  }

  /** Output JSON from data in an {@link RdapResultSet} object. */
  private RdapSearchResults makeSearchResults(
      RdapResultSet<DomainResource> resultSet, DateTime now) {
    return makeSearchResults(
        resultSet.resources(),
        resultSet.incompletenessWarningType(),
        Optional.of((long) resultSet.numResourcesRetrieved()),
        now);
  }

  /**
   * Output JSON for a list of domains.
   *
   * <p>The incompletenessWarningType should be set to TRUNCATED if the search found more results
   * than are in the list, or MIGHT_BE_INCOMPLETE if a search for domains by nameserver returned the
   * maximum number of nameservers in the first stage query.
   */
  private RdapSearchResults makeSearchResults(
      List<DomainResource> domains,
      IncompletenessWarningType incompletenessWarningType,
      Optional<Long> numDomainsRetrieved,
      DateTime now) {
    numDomainsRetrieved.ifPresent(metricInformationBuilder::setNumDomainsRetrieved);
    OutputDataType outputDataType =
        (domains.size() > 1) ? OutputDataType.SUMMARY : OutputDataType.FULL;
    RdapAuthorization authorization = getAuthorization();
    List<ImmutableMap<String, Object>> jsonList = new ArrayList<>();
    Optional<String> newCursor = Optional.empty();
    for (DomainResource domain : domains) {
      newCursor = Optional.of(domain.getFullyQualifiedDomainName());
      jsonList.add(
          rdapJsonFormatter.makeRdapJsonForDomain(
              domain, false, fullServletPath, rdapWhoisServer, now, outputDataType, authorization));
      if (jsonList.size() >= rdapResultSetMaxSize) {
        break;
      }
    }
    IncompletenessWarningType finalIncompletenessWarningType =
        (jsonList.size() < domains.size())
            ? IncompletenessWarningType.TRUNCATED
            : incompletenessWarningType;
    metricInformationBuilder.setIncompletenessWarningType(finalIncompletenessWarningType);
    return RdapSearchResults.create(
        ImmutableList.copyOf(jsonList),
        finalIncompletenessWarningType,
        (finalIncompletenessWarningType == IncompletenessWarningType.TRUNCATED)
            ? newCursor
            : Optional.empty());
  }
}