Move ANTLR4 dependency inclusion to top-level CMakeLists.txt (#127)

MSVC has a hard 260 character limit which (unfortunately) isn't
immediately fixable (note: this is independent of the "infamous" Windows
path length limit).

`substrait-cpp` can become an issue when it itself is being used as a
dependency of other projects ([see e.g. this
discussion](substrait-io/substrait-mlir-contrib#92 (comment)).

Here, it's the `antlr4` build that's generating the longest paths, e.g.:
>
`substrait-cpp/src/substrait/textplan/parser/grammar/antlr4_runtime/src/antlr4_runtime/runtime/Cpp/CMakeFiles/CMakeScratch/TryCompile-s3o9if/CMakeFiles/cmTC_945a8.dir/./`

To give ourselves a bit of breathing room here,
`src/substrait/textplan/parser/grammar` seems fairly unnecessary, and is
simply a side-effect of `include(ExternalAntlr4Cpp)` being called at
that nested location.
This PR moves that inclusion up to the top-level, giving MSVC builds an
extra 38 characters to work with 🙂.

Another possible work around could be to rewrite the ANTLR4 dependency
management s.t. `FetchContent` is used to manage the build.
This would mean that the dependency would be built in e.g. the root
`build/_deps/antlr4` location, irrespective of any git submodule
nesting.

* Also allows setting `ANTLR_BUILD_CPP_TESTS` for the ANTLR build
(avoids ANTLR pulling in gtest which may conflict with other things when
`substrait-cpp` is a dependency of other projects).
* Removes the assumption that antlr libraries are located in the `dist`
folder within the ANTLR checked-out source hierarchy (we're doing an
in-source build). AFAICT, [this behavior places a hard assumption on
`ANTLR_BUILD_CPP_TESTS` being
set](https://github.com/antlr/antlr4/blob/dev/runtime/Cpp/runtime/CMakeLists.txt#L159-L177)
which is what triggers the behaviour for copying binaries to the `dist`
folder.

---------

Co-authored-by: David Sisson <[email protected]>

Author: mortbopet

CMakeLists.txt

@@ -31,6 +31,26 @@ option(
 
 set(SUBSTRAIT_CPP_INCLUDE_DIR ${CMAKE_CURRENT_SOURCE_DIR}/include)
 
+# Setup ANTLR4 dependency. It's important that this is done at the
+# top-level CMakeLists.txt, given that the ANTLR4 build may generate deeply nested
+# paths - which may be a problem for MSVC, with its 260 character path limit.
+list(APPEND CMAKE_MODULE_PATH
+     "${CMAKE_CURRENT_SOURCE_DIR}/third_party/antlr4/cmake")
+
+set(CMAKE_CXX_STANDARD 17)
+add_definitions(-DANTLR4CPP_STATIC)
+# using /MD flag for antlr4_runtime (for Visual C++ compilers only)
+set(ANTLR4_WITH_STATIC_CRT OFF)
+set(ANTLR4_TAG 4.13.2)
+set(ANTLR4_ZIP_REPOSITORY
+    https://github.com/antlr/antlr4/archive/refs/tags/${ANTLR4_TAG}.zip)
+include(ExternalAntlr4Cpp)
+include_directories(${ANTLR4_INCLUDE_DIRS})
+set(ANTLR_EXECUTABLE_DIR ${CMAKE_CURRENT_BINARY_DIR})
+file(DOWNLOAD https://www.antlr.org/download/antlr-4.13.2-complete.jar
+     "${ANTLR_EXECUTABLE_DIR}/antlr.jar")
+set(ANTLR_EXECUTABLE "${ANTLR_EXECUTABLE_DIR}/antlr.jar")
+
 # Local files come first.
 include_directories(include)
 include_directories(src)

scripts/run-clang-tidy.sh

@@ -8,7 +8,7 @@ rm -rf tmp && mkdir tmp && cmake -Btmp -DCMAKE_EXPORT_COMPILE_COMMANDS=ON -DSUBS
 # Build substrait protobuf
 pushd tmp/src/substrait/proto && make -j 2 && popd || exit
 # Build textplan grammar
-pushd tmp/src/substrait/textplan/parser/grammar && make -j antlr4_runtime textplan_grammar_headers && popd || exit
+pushd tmp/ && make -j antlr4_runtime textplan_grammar_headers && popd || exit
 # Run clang-tidy
 if [ "$1" == "fix" ]; then
    python3 scripts/run-clang-tidy.py "$WORKDIR" "tmp" "third_party" "h,hpp,cc,cpp" "--quiet --fix"

src/substrait/textplan/parser/grammar/CMakeLists.txt

@@ -1,28 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
 
-list(APPEND CMAKE_MODULE_PATH
-     "${CMAKE_CURRENT_SOURCE_DIR}/../../../../../third_party/antlr4/cmake")
-
-set(CMAKE_CXX_STANDARD 17)
-
-add_definitions(-DANTLR4CPP_STATIC)
-
-set(GRAMMAR_DIR ${CMAKE_CURRENT_BINARY_DIR})
-
-# using /MD flag for antlr4_runtime (for Visual C++ compilers only)
-set(ANTLR4_WITH_STATIC_CRT OFF)
-
-set(ANTLR4_TAG 4.13.2)
-set(ANTLR4_ZIP_REPOSITORY
-    https://github.com/antlr/antlr4/archive/refs/tags/${ANTLR4_TAG}.zip)
-
-include(ExternalAntlr4Cpp)
-
-include_directories(${ANTLR4_INCLUDE_DIRS})
-
-file(DOWNLOAD https://www.antlr.org/download/antlr-4.13.2-complete.jar
-     "${GRAMMAR_DIR}/antlr.jar")
-set(ANTLR_EXECUTABLE "${GRAMMAR_DIR}/antlr.jar")
 find_package(ANTLR REQUIRED)
 
 antlr_target(SubstraitPlanLexer SubstraitPlanLexer.g4 LEXER PACKAGE
@@ -53,6 +30,7 @@ message(
 add_library(textplan_grammar ${ANTLR_SubstraitPlanLexer_CXX_OUTPUTS}
                              ${ANTLR_SubstraitPlanParser_CXX_OUTPUTS})
 
+set(GRAMMAR_DIR ${CMAKE_CURRENT_BINARY_DIR})
 message(STATUS "generated dir:  ${GRAMMAR_DIR}/antlr4cpp_generated_src")
 
 target_include_directories(

third_party/antlr4/cmake/ExternalAntlr4Cpp.cmake

@@ -20,11 +20,11 @@ endif()
 file(MAKE_DIRECTORY "${ANTLR4_INCLUDE_DIRS}")
 
 if(${CMAKE_GENERATOR} MATCHES "Visual Studio.*")
-  set(ANTLR4_OUTPUT_DIR ${ANTLR4_ROOT}/runtime/Cpp/dist/$(Configuration))
+  set(ANTLR4_OUTPUT_DIR ${ANTLR4_ROOT}/runtime/Cpp/runtime/$(Configuration))
 elseif(${CMAKE_GENERATOR} MATCHES "Xcode.*")
-  set(ANTLR4_OUTPUT_DIR ${ANTLR4_ROOT}/runtime/Cpp/dist/$(CONFIGURATION))
+  set(ANTLR4_OUTPUT_DIR ${ANTLR4_ROOT}/runtime/Cpp/runtime/$(CONFIGURATION))
 else()
-  set(ANTLR4_OUTPUT_DIR ${ANTLR4_ROOT}/runtime/Cpp/dist)
+  set(ANTLR4_OUTPUT_DIR ${ANTLR4_ROOT}/runtime/Cpp/runtime)
 endif()
 
 if(MSVC)
@@ -96,6 +96,7 @@ if(ANTLR4_ZIP_REPOSITORY)
       CMAKE_CACHE_ARGS
           -DCMAKE_BUILD_TYPE:STRING=${CMAKE_BUILD_TYPE}
           -DWITH_STATIC_CRT:BOOL=${ANTLR4_WITH_STATIC_CRT}
+          -DANTLR_BUILD_CPP_TESTS:BOOL=${ANTLR4_BUILD_CPP_TESTS}
           -DDISABLE_WARNINGS:BOOL=ON
           -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
           -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
@@ -117,6 +118,7 @@ else()
       CMAKE_CACHE_ARGS
           -DCMAKE_BUILD_TYPE:STRING=${CMAKE_BUILD_TYPE}
           -DWITH_STATIC_CRT:BOOL=${ANTLR4_WITH_STATIC_CRT}
+          -DANTLR_BUILD_CPP_TESTS:BOOL=${ANTLR4_BUILD_CPP_TESTS}
           -DDISABLE_WARNINGS:BOOL=ON
           -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
           -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}