container_info: switch to boost::regex (#465)

Summary:

OI was previously using `std::regex_match` to match container names. This was bad because `libstdc++`'s implementation of regex is awful. In the case of limited inlining it was causing a stack overflow when running CodeGen for large types (I think types with large names but I never got to the bottom of it).

Replace this with the competent `boost::regex_match` that we already have a dependency on.

Reviewed By: ajor

Differential Revision: D53002752

oi/CMakeLists.txt

@@ -34,6 +34,8 @@ add_library(container_info
 )
 target_link_libraries(container_info
   features
+
+  Boost::regex
   glog::glog
   toml
 )

oi/ContainerInfo.cpp

@@ -75,12 +75,12 @@ const char* containerTypeEnumToStr(ContainerTypeEnum ty) {
     return nullptr;
   }
 
-  std::regex matcher;
+  boost::regex matcher;
   if (std::optional<std::string> str =
           (*info)["matcher"].value<std::string>()) {
-    matcher = std::regex(*str, std::regex_constants::grep);
+    matcher = boost::regex(*str, boost::regex_constants::grep);
   } else {
-    matcher = std::regex("^" + typeName, std::regex_constants::grep);
+    matcher = boost::regex("^" + typeName, boost::regex_constants::grep);
   }
 
   std::optional<size_t> numTemplateParams =
@@ -194,8 +194,9 @@ namespace {
  *
  * The type name "name" should match "name" and "name<xxx>".
  */
-std::regex getMatcher(const std::string& typeName) {
-  return std::regex("^" + typeName + "$|^" + typeName + "<.*>$");
+boost::regex getMatcher(const std::string& typeName) {
+  return boost::regex("^" + typeName + "$|^" + typeName + "<.*>$",
+                      boost::regex_constants::extended);
 }
 }  // namespace
 
@@ -225,7 +226,7 @@ ContainerInfo::ContainerInfo(const fs::path& path) {
     throw ContainerInfoError(path, "`info.type_name` is a required field");
   }
 
-  matcher = getMatcher(typeName);
+  matcher_ = getMatcher(typeName);
 
   if (std::optional<std::string> str = info["ctype"].value<std::string>()) {
     ctype = containerTypeEnumFromStr(*str);
@@ -334,9 +335,9 @@ ContainerInfo::ContainerInfo(std::string typeName_,
                              ContainerTypeEnum ctype_,
                              std::string header_)
     : typeName(std::move(typeName_)),
-      matcher(getMatcher(typeName)),
       ctype(ctype_),
       header(std::move(header_)),
       codegen(Codegen{
-          "// DummyDecl %1%\n", "// DummyFunc %1%\n", "// DummyFunc\n"}) {
+          "// DummyDecl %1%\n", "// DummyFunc %1%\n", "// DummyFunc\n"}),
+      matcher_(getMatcher(typeName)) {
 }

oi/ContainerInfo.h

@@ -14,9 +14,9 @@
  * limitations under the License.
  */
 #pragma once
+#include <boost/regex.hpp>
 #include <filesystem>
 #include <optional>
-#include <regex>
 #include <set>
 #include <string>
 #include <vector>
@@ -49,7 +49,7 @@ struct ContainerInfo {
   // Old ctors, remove with OICodeGen:
   ContainerInfo() = default;
   ContainerInfo(std::string typeName_,
-                std::regex matcher_,
+                boost::regex matcher,
                 std::optional<size_t> numTemplateParams_,
                 ContainerTypeEnum ctype_,
                 std::string header_,
@@ -61,7 +61,6 @@ struct ContainerInfo {
                 oi::detail::FeatureSet requiredFeatures,
                 ContainerInfo::Codegen codegen_)
       : typeName(std::move(typeName_)),
-        matcher(std::move(matcher_)),
         numTemplateParams(numTemplateParams_),
         ctype(ctype_),
         header(std::move(header_)),
@@ -71,7 +70,8 @@ struct ContainerInfo {
         underlyingContainerIndex(underlyingContainerIndex_),
         stubTemplateParams(std::move(stubTemplateParams_)),
         requiredFeatures(requiredFeatures),
-        codegen(std::move(codegen_)) {
+        codegen(std::move(codegen_)),
+        matcher_(std::move(matcher)) {
   }
 
   ContainerInfo(ContainerInfo&&) = default;
@@ -83,8 +83,11 @@ struct ContainerInfo {
     return copy;
   }
 
+  bool matches(std::string_view sv) const {
+    return boost::regex_search(sv.begin(), sv.end(), matcher_);
+  }
+
   std::string typeName;
-  std::regex matcher;
   std::optional<size_t> numTemplateParams;
   ContainerTypeEnum ctype = UNKNOWN_TYPE;
   std::string header;
@@ -110,6 +113,8 @@ struct ContainerInfo {
  private:
   ContainerInfo(const ContainerInfo&) = default;
   ContainerInfo& operator=(const ContainerInfo& other) = default;
+
+  boost::regex matcher_;
 };
 
 class ContainerInfoError : public std::runtime_error {

oi/OICodeGen.cpp

@@ -150,7 +150,7 @@ OICodeGen::getContainerInfo(drgn_type* type) {
   for (auto it = containerInfoList.rbegin(); it != containerInfoList.rend();
        ++it) {
     const ContainerInfo& info = **it;
-    if (std::regex_search(nameStr, info.matcher)) {
+    if (info.matches(nameStr)) {
       return info;
     }
   }

oi/type_graph/ClangTypeParser.cpp

@@ -24,7 +24,6 @@
 #include <clang/Sema/Sema.h>
 #include <glog/logging.h>
 
-#include <regex>
 #include <stdexcept>
 
 #include "oi/type_graph/Types.h"
@@ -423,7 +422,7 @@ bool ClangTypeParser::chasePointer() const {
 ContainerInfo* ClangTypeParser::getContainerInfo(
     const std::string& fqName) const {
   for (const auto& containerInfo : containers_) {
-    if (std::regex_search(fqName, containerInfo->matcher)) {
+    if (containerInfo->matches(fqName)) {
       return containerInfo.get();
     }
   }

oi/type_graph/IdentifyContainers.cpp

@@ -15,8 +15,6 @@
  */
 #include "IdentifyContainers.h"
 
-#include <regex>
-
 #include "TypeGraph.h"
 #include "oi/ContainerInfo.h"
 
@@ -53,7 +51,7 @@ Type& IdentifyContainers::mutate(Type& type) {
 
 Type& IdentifyContainers::visit(Class& c) {
   for (const auto& containerInfo : containers_) {
-    if (!std::regex_search(c.fqName(), containerInfo->matcher)) {
+    if (!containerInfo->matches(c.fqName())) {
       continue;
     }
 

oi/type_graph/TypeIdentifier.cpp

@@ -71,7 +71,7 @@ void TypeIdentifier::visit(Container& c) {
     if (Class* paramClass = dynamic_cast<Class*>(&param.type())) {
       bool replaced = false;
       for (const auto& info : passThroughTypes_) {
-        if (std::regex_search(paramClass->fqName(), info.matcher)) {
+        if (info.matches(paramClass->fqName())) {
           // Create dummy containers. Use a map so previously deduplicated nodes
           // remain deduplicated.
           Container* dummy;

test/test_container_info.cpp

@@ -5,19 +5,17 @@
 TEST(ContainerInfoTest, matcher) {
   ContainerInfo info{"std::vector", SEQ_TYPE, "vector"};
 
-  EXPECT_TRUE(std::regex_search("std::vector<int>", info.matcher));
-  EXPECT_TRUE(std::regex_search("std::vector<std::list<int>>", info.matcher));
-  EXPECT_TRUE(std::regex_search("std::vector", info.matcher));
+  EXPECT_TRUE(info.matches("std::vector<int>"));
+  EXPECT_TRUE(info.matches("std::vector<std::list<int>>"));
+  EXPECT_TRUE(info.matches("std::vector"));
 
-  EXPECT_FALSE(std::regex_search("vector", info.matcher));
-  EXPECT_FALSE(std::regex_search("non_std::vector<int>", info.matcher));
-  EXPECT_FALSE(std::regex_search("std::vector_other<int>", info.matcher));
-  EXPECT_FALSE(std::regex_search("std::list<std::vector<int>>", info.matcher));
-  EXPECT_FALSE(std::regex_search("std::vector::value_type", info.matcher));
-  EXPECT_FALSE(std::regex_search("std::vector<int>::value_type", info.matcher));
-  EXPECT_FALSE(std::regex_search("std::vector<std::vector<int>>::value_type",
-                                 info.matcher));
+  EXPECT_FALSE(info.matches("vector"));
+  EXPECT_FALSE(info.matches("non_std::vector<int>"));
+  EXPECT_FALSE(info.matches("std::vector_other<int>"));
+  EXPECT_FALSE(info.matches("std::list<std::vector<int>>"));
+  EXPECT_FALSE(info.matches("std::vector::value_type"));
+  EXPECT_FALSE(info.matches("std::vector<int>::value_type"));
+  EXPECT_FALSE(info.matches("std::vector<std::vector<int>>::value_type"));
   // Uh-oh, here's a case that I don't think regexes are powerful enough to
-  // match: EXPECT_FALSE(std::regex_search("std::vector<int>::subtype<bool>",
-  // info.matcher));
+  // match: EXPECT_FALSE(info.matches("std::vector<int>::subtype<bool>"));
 }