address review comments

arangodb · gnusi · Sep 10, 2021 · Sep 3, 2021 · Sep 6, 2021 · Sep 6, 2021
commit a969139e227ff079c84706827d82ca6ffdd24966
diff --git a/arangod/IResearch/IResearchFilterFactory.cpp b/arangod/IResearch/IResearchFilterFactory.cpp
@@ -65,6 +65,7 @@
 #include "IResearch/IResearchAnalyzerFeature.h"
 #include "IResearch/IResearchCommon.h"
 #include "IResearch/IResearchFeature.h"
+#include "IResearch/IResearchFilterOptimization.h"
 #include "IResearch/IResearchKludge.h"
 #include "IResearch/IResearchPDP.h"
 #include "Logger/LogMacros.h"
@@ -3572,49 +3573,10 @@ Result fromFuncStartsWith(
 
     // Try to optimize us away
     if (!isMultiPrefix && !prefixes.empty() &&
-         ctx.filterOptimization != FilterOptimization::None && filter->type() == irs::type<irs::And>::id()) {
-      for (auto& f : *filter) {
-        if (f.type() == irs::type<irs::by_edit_distance>::id()) {
-          auto& levenshtein = static_cast<irs::by_edit_distance&>(f);
-if (levenshtein.field() == name) {
-            auto options = levenshtein.mutable_options();
-            auto startsWith = prefixes.back().second;
-            if (startsWith.size() <= options->prefix.size()) {
-              if (std::memcmp(options->prefix.data(), startsWith.c_str(),
-                              startsWith.size()) == 0) {
-                // Nothing to do. We are already covered by this levenshtein prefix
-                return {};
-              }
-            } else {
-              // maybe we could enlarge prefix to cover us?
-              if (std::memcmp(options->prefix.data(), startsWith.c_str(),
-                              options->prefix.size()) == 0) {
-                // looks promising - beginning of the levenshtein prefix is ok
-                auto prefixTailSize = startsWith.size() - options->prefix.size();
-                if (options->term.size() >= prefixTailSize &&
-                    std::memcmp(options->term.data(),
-                                startsWith.c_str() + options->prefix.size(),
-                                prefixTailSize) == 0) {
-                  // we could enlarge prefix
-               
8000
   options->prefix = irs::ref_cast<irs::byte_type>(startsWith);
-                  options->term.erase(options->term.begin(),
-                                      options->term.begin() + prefixTailSize);
-                  return {};
-                }
-              }
-            }
-            if ((options->term.size() +
-                 options->prefix.size() +
-                 options->max_distance) < startsWith.size()) {
-                     // last optimization effort - we can't fulfill this conjunction.
-                     // make it empty
-                     filter->clear();
-                     filter->add<irs::empty>();
-                     return {};
-            }
-          }
-        }
-      }
+        ctx.filterOptimization != FilterOptimization::None  && 
+        arangodb::iresearch::includeStartsWithInLevenshtein(filter, name,
+                                                            prefixes.back().second)) {
+      return {};
     }
 
     if (isMultiPrefix) {

diff --git a/arangod/IResearch/IResearchFilterOptimization.cpp b/arangod/IResearch/IResearchFilterOptimization.cpp
@@ -21,3 +21,59 @@
 ////////////////////////////////////////////////////////////////////////////////
 
 #include "IResearch/IResearchFilterOptimization.h"
+#include "search/levenshtein_filter.hpp"
+#include "search/prefix_filter.hpp"
+
+namespace arangodb {
+namespace iresearch {
+
+
+bool includeStartsWithInLevenshtein(irs::boolean_filter* filter,
+                                    irs::string_ref name,
+                                    irs::string_ref startsWith) {
+  if (filter->type() == irs::type<irs::And>::id()) {
+    for (auto& f : *filter) {
+      if (f.type() == irs::type<irs::by_edit_distance>::id()) {
+        auto& levenshtein = static_cast<irs::by_edit_distance&>(f);
+        if (levenshtein.field() == name) {
+          auto options = levenshtein.mutable_options();
+          if (startsWith.size() <= options->prefix.size()) {
+            if (std::memcmp(options->prefix.data(), startsWith.c_str(),
+                            startsWith.size()) == 0) {
+              // Nothing to do. We are already covered by this levenshtein prefix
+              return true;
+            }
+          } else {
+            // maybe we could enlarge prefix to cover us?
+            if (std::memcmp(options->prefix.data(), startsWith.c_str(),
+                            options->prefix.size()) == 0) {
+              // looks promising - beginning of the levenshtein prefix is ok
+              auto prefixTailSize = startsWith.size() - options->prefix.size();
+              if (options->term.size() >= prefixTailSize &&
+                  std::memcmp(options->term.data(),
+                              startsWith.c_str() + options->prefix.size(),
+                              prefixTailSize) == 0) {
+                // we could enlarge prefix
+                options->prefix = irs::ref_cast<irs::byte_type>(startsWith);
+                options->term.erase(options->term.begin(), options->term.begin() + prefixTailSize);
+                return true;
+              }
+            }
+          }
+          if ((options->term.size() + options->prefix.size() + options->max_distance) <
+              startsWith.size()) {
+            // last optimization effort - we can't fulfill this conjunction.
+            // make it empty
+            filter->clear();
+            filter->add<irs::empty>();
+            return true;
+          }
+        }
+      }
+    }
+  }
+  return false;
+}
+
+} // namespace iresearch
+} // namespace arangodb
@@ -21,6 +21,8 @@
 ////////////////////////////////////////////////////////////////////////////////
 #pragma once
 
+#include "search/boolean_filter.hpp"
+
 namespace arangodb {
 namespace iresearch {
 
@@ -29,5 +31,9 @@ enum class FilterOptimization : int {
   None = 0
 };
 
-}
-} //namespace arangodb
+bool includeStartsWithInLevenshtein(irs::boolean_filter* filter,
+                                    irs::string_ref name,
+                                    irs::string_ref startsWith);
+
+} // namespace iresearch
+} // namespace arangodb