fix(nimble): Make the shared dictionary export build and pass checks

#18640 brought D116132982's shared dictionary work across byte-identical to
fbcode, which is right for the export but leaves it not compiling here and
failing pre-commit. These are the repairs, kept separate so the export itself
stays a clean mirror.

Two struct members arrived without default initializers while their neighbours
have them. Velox builds -Wextra -Werror and fbcode does not promote
-Wmissing-field-initializers, so the omissions are invisible internally and
fatal here:

- Encoding::Options::sharedDictionaryAlphabet, 196 instances across
  ConstantEncodingTest, DictionaryEncodingTest, EncodingTest, FsstEncodingTest
  and HuffmanEncodingViewTest. Set to nullptr, matching bufferPool,
  encodingBufferPool and decodingStats beside it.
- SharedDictionaryConfig::alphabetEncodings, 5 instances in
  SharedDictionaryConfigTest. Given `{}`, matching dictionaryId and
  useExternalAlphabet beside it.

ExternalSharedDictionaryBuilder::resetImpl passed a Kind enum straight to
NIMBLE_UNSUPPORTED. Velox bundles fmt v11, which will not format an enum
implicitly:

  error: 'type_is_unformattable_for<SharedDictionaryBuilder<int>::Kind, char>'
    has incomplete type

instantiated for four value types. SharedDictionaryBuilder::kindString already
exists for this, so it is used, which also turns the message from an integer
into "External shared dictionary builder does not support reset()".

Four headers reached this tree without CMake entries, so check-header-ownership
failed: SharedDictionaryCatalog.h, SharedDictionaryReader.h,
ExternalDictionaryBuilder.h and SharedDictionaryConfig.h. Each is listed
alongside the .cpp already in its target. fbcode builds with BUCK, so nothing
internal enforces this. All 1651 headers are tracked again.

Touching those files makes gersemi reformat two add_dependencies blocks that
arrived unformatted; that is the hook's doing, not a manual edit.

Velox is GitHub-first, so the next import carries these into fbcode.

Author: Krishna Pai

velox/dwio/nimble/encodings/CMakeLists.txt

@@ -32,6 +32,7 @@ add_library(
   PrefixEncoding.cpp
   RLEEncoding.cpp
   SharedDictionaryCatalog.cpp
+  SharedDictionaryCatalog.h
   SharedDictionaryEncoding.cpp
   SparseBoolEncoding.cpp
   TrivialEncoding.cpp
@@ -134,10 +135,7 @@ target_include_directories(
   nimble_shared_dictionary_fb
   INTERFACE ${PROJECT_BINARY_DIR} ${FLATBUFFERS_INCLUDE_DIR}
 )
-add_dependencies(
-  nimble_shared_dictionary_fb
-  nimble_shared_dictionary_schema_fb
-)
+add_dependencies(nimble_shared_dictionary_fb nimble_shared_dictionary_schema_fb)
 
 target_link_libraries(
   nimble_encodings

velox/dwio/nimble/encodings/SharedDictionaryBuilder.h

@@ -289,7 +289,8 @@ class ExternalSharedDictionaryBuilder final
 
   void resetImpl() final {
     NIMBLE_UNSUPPORTED(
-        "{} shared dictionary builder does not support reset().", this->kind());
+        "{} shared dictionary builder does not support reset().",
+        SharedDictionaryBuilder<T>::kindString(this->kind()));
   }
 
  private:

velox/dwio/nimble/encodings/common/Encoding.h

@@ -162,7 +162,8 @@ class Encoding {
 
     /// Direct alphabet for SharedDictionary encodings when the read path has
     /// already resolved the dictionary bound to this value stream.
-    std::shared_ptr<const SharedDictionaryAlphabet> sharedDictionaryAlphabet;
+    std::shared_ptr<const SharedDictionaryAlphabet> sharedDictionaryAlphabet =
+        nullptr;
 
     velox::io::IoCounter* decompressCounter() const {
       return decodingStats != nullptr ? &decodingStats->decompressCPUTimeNanos

velox/dwio/nimble/tablet/CMakeLists.txt

@@ -186,6 +186,7 @@ add_library(
   nimble_tablet_reader
   FileLayout.cpp
   SharedDictionaryReader.cpp
+  SharedDictionaryReader.h
   StripeGroup.cpp
   TabletReader.cpp
   TabletReader.h

velox/dwio/nimble/tools/CMakeLists.txt

@@ -29,10 +29,7 @@ target_include_directories(
   nimble_external_dictionary_fb
   INTERFACE ${PROJECT_BINARY_DIR} ${FLATBUFFERS_INCLUDE_DIR}
 )
-add_dependencies(
-  nimble_external_dictionary_fb
-  nimble_external_dictionary_schema_fb
-)
+add_dependencies(nimble_external_dictionary_fb nimble_external_dictionary_schema_fb)
 
 add_library(nimble_dump_lib NimbleDumpLib.cpp NimbleDumpLib.h)
 target_link_libraries(
@@ -73,7 +70,11 @@ target_link_libraries(
   velox_memory
 )
 
-add_library(nimble_external_dictionary_builder ExternalDictionaryBuilder.cpp)
+add_library(
+  nimble_external_dictionary_builder
+  ExternalDictionaryBuilder.cpp
+  ExternalDictionaryBuilder.h
+)
 target_link_libraries(
   nimble_external_dictionary_builder
   nimble_external_dictionary_fb

velox/dwio/nimble/velox/CMakeLists.txt

@@ -14,7 +14,11 @@
 add_library(nimble_velox_common SchemaUtils.cpp SchemaUtils.h)
 target_link_libraries(nimble_velox_common nimble_common velox_type Folly::folly fmt::fmt)
 
-add_library(nimble_velox_shared_dictionary_config SharedDictionaryConfig.cpp)
+add_library(
+  nimble_velox_shared_dictionary_config
+  SharedDictionaryConfig.cpp
+  SharedDictionaryConfig.h
+)
 target_link_libraries(
   nimble_velox_shared_dictionary_config
   nimble_common

velox/dwio/nimble/velox/SharedDictionaryConfig.h

@@ -37,7 +37,7 @@ struct SharedDictionaryConfig {
   /// Resolves a provided alphabet instead of building one from written values.
   bool useExternalAlphabet{false};
   /// Candidate encodings for alphabets stored in this file.
-  std::vector<EncodingType> alphabetEncodings;
+  std::vector<EncodingType> alphabetEncodings{};
 };
 
 /// Shared dictionary settings for one regular column value stream.