C++20 compatibility (#1479)

Summary:
This diff fixes a few incompatibility issues with C++ 20.

1) fmt::format checks at compile time that the format string is constexpr or implicitly constructed from a string literal.  There are a couple violations, wrapping them in fmt::runtime is less performant than making them constexpr, but they're in tests/uninteresting parts of benchmarks so it's fine.
2) A char8_t type was added for UTF-8 strings, this breaks implicit conversion of string literals using the u8 literal to strings.  This affected a few tests, but the u8 literal was unnecessary
3) Ambiguous equality expression warnings when comparing two different types.  With an expression like a == b the compiler will look up the equality function for a and the equality function for b and issue a warning if they're different functions.  We were explicitly doing this to test the inequality of different types in a few tests, explicitly calling the equality operator on the object produces the original behavior.

I also had to update the fmt dependency used in the setup-circleci.sh script.  The one available in dnf appears to be out of date, so I pulled the code from GitHub and built it, using the same version and in the same manner the other setup scripts prepare fmt.

Pull Request resolved: #1479

Reviewed By: kgpai

Differential Revision: D35939375

Pulled By: kevinwilfong

fbshipit-source-id: 3e9d8eb0bd63b6e911bb8eaf29ff37bef2a00097

Author: kevinwilfong

.circleci/config.yml

@@ -55,7 +55,7 @@ workflows:
 executors:
   build:
     docker:
-      - image : prestocpp/velox-circleci:kpai-20220401
+      - image : prestocpp/velox-circleci:kevinwilfong-20220426
     resource_class: 2xlarge
     environment:
         CC:  /opt/rh/gcc-toolset-9/root/bin/gcc

scripts/setup-circleci.sh

@@ -28,7 +28,7 @@ dnf_install epel-release dnf-plugins-core # For ccache, ninja
 dnf config-manager --set-enabled powertools
 dnf_install ninja-build ccache gcc-toolset-9 git wget which libevent-devel \
   openssl-devel re2-devel libzstd-devel lz4-devel double-conversion-devel \
-  protobuf-devel fmt-devel libdwarf-devel curl-devel
+  protobuf-devel libdwarf-devel curl-devel
 
 dnf remove -y gflags
 
@@ -64,6 +64,7 @@ wget_and_untar http://www.oberhumer.com/opensource/lzo/download/lzo-2.10.tar.gz
 wget_and_untar https://boostorg.jfrog.io/artifactory/main/release/1.72.0/source/boost_1_72_0.tar.gz boost &
 wget_and_untar https://github.com/google/snappy/archive/1.1.8.tar.gz snappy &
 wget_and_untar https://github.com/facebook/folly/archive/v2022.03.14.00.tar.gz folly &
+wget_and_untar https://github.com/fmtlib/fmt/archive/8.0.0.tar.gz fmt &
 #  wget_and_untar https://github.com/ericniebler/range-v3/archive/0.11.0.tar.gz ranges-v3 &
 
 wait  # For cmake and source downloads to complete.
@@ -85,6 +86,8 @@ wait  # For cmake and source downloads to complete.
 cmake_install gflags -DBUILD_SHARED_LIBS=ON -DBUILD_STATIC_LIBS=ON -DBUILD_gflags_LIB=ON -DLIB_SUFFIX=64 -DCMAKE_INSTALL_PREFIX:PATH=/usr
 cmake_install glog -DBUILD_SHARED_LIBS=ON -DCMAKE_INSTALL_PREFIX:PATH=/usr
 cmake_install snappy -DSNAPPY_BUILD_TESTS=OFF
+# folly depends on fmt so this needs to come first.
+cmake_install fmt -DFMT_TEST=OFF
 cmake_install folly
 # cmake_install ranges-v3
 

velox/common/base/tests/ExceptionTest.cpp

@@ -354,7 +354,7 @@ TEST(ExceptionTest, notNull) {
 TEST(ExceptionTest, expressionString) {
   size_t i = 1;
   size_t j = 100;
-  std::string msgTemplate =
+  constexpr auto msgTemplate =
       "Exception: VeloxRuntimeError"
       "\nError Source: RUNTIME"
       "\nError Code: INVALID_STATE"

velox/dwio/parquet/tests/ParquetTpchTest.cpp

@@ -69,7 +69,7 @@ class ParquetTpchTest : public testing::Test {
       fs::create_directory(tableDirectory);
       auto filePath = fmt::format("{}/file.parquet", tableDirectory);
       auto query = fmt::format(
-          duckDbParquetWriteSQL_.at(tableName),
+          fmt::runtime(duckDbParquetWriteSQL_.at(tableName)),
           tableName,
           filePath,
           kRowGroupSize);

velox/functions/lib/tests/Re2FunctionsTest.cpp

@@ -380,16 +380,13 @@ TEST_F(Re2FunctionsTest, likePattern) {
 
   EXPECT_EQ(
       like(
-          u8"\u4FE1\u5FF5 \u7231 \u5E0C\u671B \u2028 abc",
-          u8"\u4FE1\u5FF5 \u7231%"),
+          "\u4FE1\u5FF5 \u7231 \u5E0C\u671B \u2028 abc",
+          "\u4FE1\u5FF5 \u7231%"),
       true);
   EXPECT_EQ(
-      like(u8"\u4FE1\u5FF5 \u7231 \u5E0C\u671B \u2028 ", u8"\u4FE1%\u7231%"),
-      true);
+      like("\u4FE1\u5FF5 \u7231 \u5E0C\u671B \u2028 ", "\u4FE1%\u7231%"), true);
   EXPECT_EQ(
-      like(
-          u8"\u4FE1\u5FF5 \u7231 \u5E0C\u671B \u2028 ",
-          u8"\u7231\u4FE1%\u7231%"),
+      like("\u4FE1\u5FF5 \u7231 \u5E0C\u671B \u2028 ", "\u7231\u4FE1%\u7231%"),
       false);
 
   EXPECT_EQ(like("abc", "MEDIUM POLISHED%"), false);

velox/functions/prestosql/aggregates/tests/ChecksumAggregateTest.cpp

@@ -152,7 +152,7 @@ TEST_F(ChecksumAggregateTest, varchars) {
   assertSingleGroupChecksum<StringView>({{}}, "h8rrhbF5N54=");
   assertSingleGroupChecksum<StringView>({"abcd"_sv}, "lGFxgnIYgPw=");
   assertSingleGroupChecksum<StringView>(
-      {u8"Thanks \u0020\u007F"_sv}, "oEh7YyEV+dM=");
+      {"Thanks \u0020\u007F"_sv}, "oEh7YyEV+dM=");
 }
 
 TEST_F(ChecksumAggregateTest, arrays) {

velox/functions/prestosql/aggregates/tests/PrestoHasherTest.cpp

@@ -211,7 +211,7 @@ TEST_F(PrestoHasherTest, floats) {
 
 TEST_F(PrestoHasherTest, varchars) {
   assertHash<StringView>(
-      {"abcd"_sv, ""_sv, std::nullopt, u8"Thanks \u0020\u007F"_sv},
+      {"abcd"_sv, ""_sv, std::nullopt, "Thanks \u0020\u007F"_sv},
       {-2449070131962342708, -1205034819632174695, 0, 2911531567394159200});
 }
 

velox/functions/prestosql/aggregates/tests/VarianceAggregationTest.cpp

@@ -24,7 +24,7 @@ namespace {
 
 // Replace the given query's placeholders '{0}' with the given aggregation name.
 std::string genAggrQuery(const char* query, const char* aggrName) {
-  return fmt::format(query, aggrName);
+  return fmt::format(fmt::runtime(query), aggrName);
 }
 
 // Helper generates aggregation over column string.

velox/functions/prestosql/benchmarks/URLBenchmark.cpp

@@ -207,8 +207,8 @@ class UrlBenchmark : public functions::test::FunctionBenchmarkBase {
         : vectorMaker_.rowVector({vectorUrls});
 
     auto queryString = isParameter ? "{}(c0, c1)" : "{}(c0)";
-    auto exprSet =
-        compileExpression(fmt::format(queryString, fnName), rowVector->type());
+    auto exprSet = compileExpression(
+        fmt::format(fmt::runtime(queryString), fnName), rowVector->type());
 
     suspender.dismiss();
 

velox/functions/prestosql/tests/JsonExtractScalarTest.cpp

@@ -65,15 +65,15 @@ TEST_F(JsonExtractScalarTest, simple) {
 TEST_F(JsonExtractScalarTest, utf8) {
   EXPECT_EQ(
       json_extract_scalar(R"({"k1":"I \u2665 UTF-8"})", "$.k1"),
-      u8"I \u2665 UTF-8");
+      "I \u2665 UTF-8");
   EXPECT_EQ(
-      json_extract_scalar(u8"{\"k1\":\"I \u2665 UTF-8\"}", "$.k1"),
-      u8"I \u2665 UTF-8");
+      json_extract_scalar("{\"k1\":\"I \u2665 UTF-8\"}", "$.k1"),
+      "I \u2665 UTF-8");
 
   EXPECT_EQ(
       json_extract_scalar(
-          u8"{\"k1\":\"I \U0001D11E playing in G-clef\"}", "$.k1"),
-      u8"I \U0001D11E playing in G-clef");
+          "{\"k1\":\"I \U0001D11E playing in G-clef\"}", "$.k1"),
+      "I \U0001D11E playing in G-clef");
 }
 
 TEST_F(JsonExtractScalarTest, invalidPath) {