diff --git a/.github/workflows/java-wasm-bindings.yml b/.github/workflows/java-wasm-bindings.yml index 7fe24455a7..de90a4173c 100644 --- a/.github/workflows/java-wasm-bindings.yml +++ b/.github/workflows/java-wasm-bindings.yml @@ -27,7 +27,7 @@ jobs: bundler-cache: true - name: rake templates - run: PRISM_EXCLUDE_PRETTYPRINT=1 PRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS=1 bundle exec rake templates + run: PRISM_EXCLUDE_PRETTYPRINT=1 PRISM_OMIT_NODE_ID=1 bundle exec rake templates - name: Set up WASI-SDK run: | diff --git a/Makefile b/Makefile index 0f6f5264d1..5e83dfa146 100644 --- a/Makefile +++ b/Makefile @@ -31,7 +31,7 @@ all: shared static shared: build/libprism.$(SOEXT) static: build/libprism.a wasm: javascript/src/prism.wasm -java-wasm: java/wasm/src/main/wasm/prism.wasm +java-wasm: java/wasm/src/main/wasm/prism.wasm java/wasm-full/src/main/wasm/prism-full.wasm build/libprism.$(SOEXT): $(SHARED_OBJECTS) $(ECHO) "linking $@ with $(CC)" @@ -52,6 +52,17 @@ javascript/src/prism.wasm: Makefile $(SOURCES) $(HEADERS) -o $@ $(SOURCES) java/wasm/src/main/wasm/prism.wasm: Makefile $(SOURCES) $(HEADERS) + $(ECHO) "building $@" + $(Q) $(MAKEDIRS) $(@D) + $(Q) $(WASI_SDK_PATH)/bin/clang \ + $(DEBUG_FLAGS) \ + -DPRISM_EXCLUDE_PRETTYPRINT -DPRISM_EXPORT_SYMBOLS -D_WASI_EMULATED_MMAN -DPRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS=1 \ + -DPRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS \ + -lwasi-emulated-mman $(CPPFLAGS) $(JAVA_WASM_CFLAGS) \ + -Wl,--export-all -Wl,--no-entry -mexec-model=reactor -lc++ -lc++abi \ + -o $@ $(SOURCES) + +java/wasm-full/src/main/wasm/prism-full.wasm: Makefile $(SOURCES) $(HEADERS) $(ECHO) "building $@" $(Q) $(MAKEDIRS) $(@D) $(Q) $(WASI_SDK_PATH)/bin/clang \ diff --git a/Rakefile b/Rakefile index eb96249985..998640eede 100644 --- a/Rakefile +++ b/Rakefile @@ -55,6 +55,7 @@ CLOBBER.concat(Prism::Template::TEMPLATES) CLOBBER.concat(["build"]) CLOBBER << "lib/prism/prism.#{RbConfig::CONFIG["DLEXT"]}" CLOBBER << "java/wasm/src/main/resources/prism.wasm" +CLOBBER << "java/wasm-full/src/main/resources/prism-full.wasm" Prism::Template::TEMPLATES.each do |filepath| desc "Generate #{filepath}" diff --git a/Steepfile b/Steepfile index e6e1a8efb5..9db20dfa1e 100644 --- a/Steepfile +++ b/Steepfile @@ -15,4 +15,7 @@ target :lib do # Ignored because we do not want to overlap with the C extension. ignore "lib/prism/ffi.rb" + ignore "lib/prism/ffi/common.rb" + ignore "lib/prism/ffi/native_ffi.rb" + ignore "lib/prism/ffi/wasm_ffi.rb" end diff --git a/java/README.md b/java/README.md index 0dd4215777..bd06639f67 100644 --- a/java/README.md +++ b/java/README.md @@ -15,7 +15,7 @@ Some files need to be generated before the Maven artifacts can build: Sources under `api` are generated from templates in `../templates`. Those sources are generated using the follow command line: ``` -$ PRISM_EXCLUDE_PRETTYPRINT=1 PRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS=1 bundle exec rake templates +$ PRISM_EXCLUDE_PRETTYPRINT=1 bundle exec rake templates ``` The files are generated under `api/src/main/java-templates` and will not be removed with `mvn clean`. diff --git a/java/api/pom.xml b/java/api/pom.xml index 396fa33161..4fe0252d08 100644 --- a/java/api/pom.xml +++ b/java/api/pom.xml @@ -9,7 +9,7 @@ prism-parser-api - Java Prism + Java Prism API Java API for the Prism Ruby language parser https://github.com/ruby/prism diff --git a/java/pom.xml b/java/pom.xml index f405af5304..8b71d4efee 100644 --- a/java/pom.xml +++ b/java/pom.xml @@ -6,7 +6,7 @@ prism-parser 0.0.2-SNAPSHOT pom - Java Prism + Java Prism parent Java API for the Prism Ruby language parser https://github.com/ruby/prism @@ -51,6 +51,7 @@ api wasm + wasm-full diff --git a/java/wasm-full/pom.xml b/java/wasm-full/pom.xml new file mode 100644 index 0000000000..e70172d838 --- /dev/null +++ b/java/wasm-full/pom.xml @@ -0,0 +1,80 @@ + + + 4.0.0 + + + org.ruby-lang + prism-parser + 0.0.2-SNAPSHOT + + + prism-parser-wasm-full + Java Prism WASM with full parsed content + Java WASM bindings for the Prism parser shared library + https://github.com/ruby/prism + + + 1.7.5 + + + + + + com.dylibso.chicory + bom + ${chicory.version} + pom + import + + + + + + + com.dylibso.chicory + runtime + + + com.dylibso.chicory + log + + + com.dylibso.chicory + wasi + + + com.dylibso.chicory + wasm + + + org.junit.jupiter + junit-jupiter-engine + ${junit.version} + test + + + + + + + com.dylibso.chicory + chicory-compiler-maven-plugin + ${chicory.version} + + + prism + + compile + + + org.ruby_lang.prism.wasm.full.PrismParser + org.ruby_lang.prism.wasm.full.Prism + src/main/wasm/prism-full.wasm + + + + + + + + diff --git a/java/wasm-full/src/main/java/org/ruby_lang/prism/wasm/full/Prism.java b/java/wasm-full/src/main/java/org/ruby_lang/prism/wasm/full/Prism.java new file mode 100644 index 0000000000..68c9707f7a --- /dev/null +++ b/java/wasm-full/src/main/java/org/ruby_lang/prism/wasm/full/Prism.java @@ -0,0 +1,153 @@ +package org.ruby_lang.prism.wasm.full; + +import com.dylibso.chicory.runtime.ByteArrayMemory; +import com.dylibso.chicory.runtime.ImportValues; +import com.dylibso.chicory.runtime.Instance; +import com.dylibso.chicory.wasi.WasiOptions; +import com.dylibso.chicory.wasi.WasiPreview1; + +import java.nio.charset.StandardCharsets; + +public class Prism implements AutoCloseable { + private final WasiPreview1 wasi; + protected final Prism_ModuleExports exports; + private final Instance instance; + + public Prism() { + this(WasiOptions.builder().build()); + } + + public Prism(WasiOptions wasiOpts) { + wasi = WasiPreview1.builder().withOptions(wasiOpts).build(); + instance = Instance.builder(PrismParser.load()) + .withMemoryFactory(ByteArrayMemory::new) + .withMachineFactory(PrismParser::create) + .withImportValues(ImportValues.builder().addFunction(wasi.toHostFunctions()).build()) + .build(); + exports = new Prism_ModuleExports(instance); + } + + public String version() { + int versionPointer = exports.pmVersion(); + int length = exports.strchr(versionPointer, 0); + + return new String(instance.memory().readBytes(versionPointer, length - versionPointer)); + } + + public byte[] parse(byte[] sourceBytes, byte[] packedOptions) { + try ( + Buffer buffer = new Buffer(); + Source source = new Source(sourceBytes, 0, sourceBytes.length); + Options options = new Options(packedOptions)) { + + return parse(buffer, source, options); + } + } + + public byte[] lex(byte[] sourceBytes, byte[] packedOptions) { + try ( + Buffer buffer = new Buffer(); + Source source = new Source(sourceBytes, 0, sourceBytes.length); + Options options = new Options(packedOptions)) { + + return lex(buffer, source, options); + } + } + + public byte[] parse(byte[] sourceBytes, int sourceOffset, int sourceLength, byte[] packedOptions) { + try ( + Buffer buffer = new Buffer(); + Source source = new Source(sourceBytes, sourceOffset, sourceLength); + Options options = new Options(packedOptions)) { + + return parse(buffer, source, options); + } + } + + public byte[] parse(Buffer buffer, Source source, Options options) { + exports.pmSerializeParse( + buffer.pointer, source.pointer, source.length, options.pointer); + + return buffer.read(); + } + + public byte[] lex(Buffer buffer, Source source, Options options) { + exports.pmSerializeLex( + buffer.pointer, source.pointer, source.length, options.pointer); + + return buffer.read(); + } + + public class Buffer implements AutoCloseable { + final int pointer; + + Buffer() { + pointer = exports.pmBufferNew(); + clear(); + } + + public void clear() { + exports.pmBufferClear(pointer); + } + + public void close() { + exports.pmBufferFree(pointer); + } + + public byte[] read() { + return instance.memory().readBytes( + exports.pmBufferValue(pointer), + exports.pmBufferLength(pointer)); + } + } + + public class Source implements AutoCloseable{ + final int pointer; + final int length; + + public Source(int length) { + pointer = exports.calloc(1, length); + this.length = length; + } + + public Source(byte[] bytes, int offset, int length) { + this(length + 1); + write(bytes, offset, length); + } + + public Source(byte[] bytes) { + this(bytes, 0, bytes.length); + } + + public void write(byte[] bytes, int offset, int length) { + assert length + 1 <= this.length; + instance.memory().write(pointer, bytes, offset, length); + instance.memory().writeByte(pointer + length, (byte) 0); + } + + public void close() { + exports.free(pointer); + } + } + + class Options implements AutoCloseable { + final int pointer; + + Options(byte[] packedOptions) { + int pointer = exports.calloc(1, packedOptions.length); + instance.memory().write(pointer, packedOptions); + this.pointer = pointer; + } + + public void close() { + exports.free(pointer); + } + } + + @Override + public void close() { + if (wasi != null) { + wasi.close(); + } + } +} diff --git a/java/wasm-full/src/main/wasm/prism-full.wasm b/java/wasm-full/src/main/wasm/prism-full.wasm new file mode 100755 index 0000000000..5af8299c37 Binary files /dev/null and b/java/wasm-full/src/main/wasm/prism-full.wasm differ diff --git a/java/wasm/pom.xml b/java/wasm/pom.xml index b389c57ed1..c0772a7269 100644 --- a/java/wasm/pom.xml +++ b/java/wasm/pom.xml @@ -9,7 +9,7 @@ prism-parser-wasm - Java Prism WASM + Java Prism WASM with semantic-only content Java WASM bindings for the Prism parser shared library https://github.com/ruby/prism diff --git a/lib/prism/ffi.rb b/lib/prism/ffi.rb index 6b9bde51ea..93efcd0836 100644 --- a/lib/prism/ffi.rb +++ b/lib/prism/ffi.rb @@ -1,286 +1,57 @@ # frozen_string_literal: true # :markup: markdown +# -- # typed: ignore -# This file is responsible for mirroring the API provided by the C extension by -# using FFI to call into the shared library. - -require "rbconfig" -require "ffi" - -# We want to eagerly load this file if there are Ractors so that it does not get -# autoloaded from within a non-main Ractor. -require "prism/serialize" if defined?(Ractor) - -module Prism # :nodoc: - module LibRubyParser # :nodoc: - extend FFI::Library - - # Define the library that we will be pulling functions from. Note that this - # must align with the build shared library from make/rake. - libprism_in_build = File.expand_path("../../build/libprism.#{RbConfig::CONFIG["SOEXT"]}", __dir__) - libprism_in_libdir = "#{RbConfig::CONFIG["libdir"]}/prism/libprism.#{RbConfig::CONFIG["SOEXT"]}" - - if File.exist?(libprism_in_build) - INCLUDE_DIR = File.expand_path("../../include", __dir__) - ffi_lib libprism_in_build - else - INCLUDE_DIR = "#{RbConfig::CONFIG["libdir"]}/prism/include" - ffi_lib libprism_in_libdir - end - - # Convert a native C type declaration into a symbol that FFI understands. - # For example: - # - # const char * -> :pointer - # bool -> :bool - # size_t -> :size_t - # void -> :void - # - def self.resolve_type(type, callbacks) - type = type.strip - - if !type.end_with?("*") - type.delete_prefix("const ").to_sym - else - type = type.delete_suffix("*").rstrip - callbacks.include?(type.to_sym) ? type.to_sym : :pointer - end - end - - # Read through the given header file and find the declaration of each of the - # given functions. For each one, define a function with the same name and - # signature as the C function. - def self.load_exported_functions_from(header, *functions, callbacks) - File.foreach("#{INCLUDE_DIR}/#{header}") do |line| - # We only want to attempt to load exported functions. - next unless line.start_with?("PRISM_EXPORTED_FUNCTION ") - - # We only want to load the functions that we are interested in. - next unless functions.any? { |function| line.include?(function) } - - # Strip trailing attributes (PRISM_NODISCARD, PRISM_NONNULL(...), etc.) - line = line.sub(/\)(\s+PRISM_\w+(?:\([^)]*\))?)+\s*;/, ");") - - # Parse the function declaration. - unless /^PRISM_EXPORTED_FUNCTION (?.+) (?\w+)\((?.+)\);$/ =~ line - raise "Could not parse #{line}" - end - - # Delete the function from the list of functions we are looking for to - # mark it as having been found. - functions.delete(name) - - # Split up the argument types into an array, ensure we handle the case - # where there are no arguments (by explicit void). - arg_types = arg_types.split(",").map(&:strip) - arg_types = [] if arg_types == %w[void] - - # Resolve the type of the argument by dropping the name of the argument - # first if it is present. - arg_types.map! { |type| resolve_type(type.sub(/\w+$/, ""), callbacks) } - - # Attach the function using the FFI library. - attach_function name, arg_types, resolve_type(return_type, []) - end - - # If we didn't find all of the functions, raise an error. - raise "Could not find functions #{functions.inspect}" unless functions.empty? - end - - callback :pm_source_stream_fgets_t, [:pointer, :int, :pointer], :pointer - callback :pm_source_stream_feof_t, [:pointer], :int - pm_source_init_result_values = %i[PM_SOURCE_INIT_SUCCESS PM_SOURCE_INIT_ERROR_GENERIC PM_SOURCE_INIT_ERROR_DIRECTORY PM_SOURCE_INIT_ERROR_NON_REGULAR] - enum :pm_source_init_result_t, pm_source_init_result_values - enum :pm_string_query_t, [:PM_STRING_QUERY_ERROR, -1, :PM_STRING_QUERY_FALSE, :PM_STRING_QUERY_TRUE] - - # Ractor-safe lookup table for pm_source_init_result_t, since FFI's - # enum_type accesses module instance variables that are not shareable. - SOURCE_INIT_RESULT = pm_source_init_result_values.freeze - - load_exported_functions_from( - "prism/version.h", - "pm_version", - [] - ) - - load_exported_functions_from( - "prism/serialize.h", - "pm_serialize_parse", - "pm_serialize_parse_stream", - "pm_serialize_parse_comments", - "pm_serialize_lex", - "pm_serialize_parse_lex", - "pm_serialize_parse_success_p", - [] - ) - - load_exported_functions_from( - "prism/string_query.h", - "pm_string_query_local", - "pm_string_query_constant", - "pm_string_query_method_name", - [] - ) - - load_exported_functions_from( - "prism/buffer.h", - "pm_buffer_new", - "pm_buffer_value", - "pm_buffer_length", - "pm_buffer_free", - [] - ) - - load_exported_functions_from( - "prism/source.h", - "pm_source_file_new", - "pm_source_mapped_new", - "pm_source_stream_new", - "pm_source_free", - "pm_source_source", - "pm_source_length", - [:pm_source_stream_fgets_t, :pm_source_stream_feof_t] - ) - - # This object represents a pm_buffer_t. We only use it as an opaque pointer, - # so it doesn't need to know the fields of pm_buffer_t. - class PrismBuffer # :nodoc: - attr_reader :pointer - - def initialize(pointer) - @pointer = pointer - end - - def value - LibRubyParser.pm_buffer_value(pointer) - end - - def length - LibRubyParser.pm_buffer_length(pointer) - end - - def read - value.read_string(length) - end - - # Initialize a new buffer and yield it to the block. The buffer will be - # automatically freed when the block returns. - def self.with - buffer = LibRubyParser.pm_buffer_new - raise unless buffer - - begin - yield new(buffer) - ensure - LibRubyParser.pm_buffer_free(buffer) - end - end - end - - # This object represents source code to be parsed. For strings it wraps a - # pointer directly; for files it uses a pm_source_t under the hood. - class PrismSource # :nodoc: - PLATFORM_EXPECTS_UTF8 = - RbConfig::CONFIG["host_os"].match?(/bccwin|cygwin|djgpp|mingw|mswin|wince|darwin/i) - - attr_reader :pointer, :length - - def initialize(pointer, length, from_string) - @pointer = pointer - @length = length - @from_string = from_string - end - - def read - raise "should use the original String instead" if @from_string - @pointer.read_string(@length) - end - - # Yields a PrismSource backed by the given string to the block. - def self.with_string(string) - raise TypeError unless string.is_a?(String) - - length = string.bytesize - # + 1 to never get an address of 0, which pm_parser_init() asserts - FFI::MemoryPointer.new(:char, length + 1, false) do |pointer| - pointer.write_string(string) - # since we have the extra byte we might as well \0-terminate - pointer.put_char(length, 0) - return yield new(pointer, length, true) - end - end - - # Yields a PrismSource to the given block, backed by a pm_source_t. - def self.with_file(filepath) - raise TypeError unless filepath.is_a?(String) - - # On Windows and Mac, it's expected that filepaths will be encoded in - # UTF-8. If they are not, we need to convert them to UTF-8 before - # passing them into pm_source_mapped_new. - if PLATFORM_EXPECTS_UTF8 && (encoding = filepath.encoding) != Encoding::ASCII_8BIT && encoding != Encoding::UTF_8 - filepath = filepath.encode(Encoding::UTF_8) - end - - FFI::MemoryPointer.new(:int) do |result_ptr| - pm_source = LibRubyParser.pm_source_mapped_new(filepath, 0, result_ptr) - - case SOURCE_INIT_RESULT[result_ptr.read_int] - when :PM_SOURCE_INIT_SUCCESS - pointer = LibRubyParser.pm_source_source(pm_source) - length = LibRubyParser.pm_source_length(pm_source) - return yield new(pointer, length, false) - when :PM_SOURCE_INIT_ERROR_GENERIC - raise SystemCallError.new(filepath, FFI.errno) - when :PM_SOURCE_INIT_ERROR_DIRECTORY - raise Errno::EISDIR.new(filepath) - when :PM_SOURCE_INIT_ERROR_NON_REGULAR - # Fall back to reading the file through Ruby IO for non-regular - # files (pipes, character devices, etc.) - return with_string(File.read(filepath)) { |string| yield string } - else - raise "Unknown error initializing pm_source_t: #{result_ptr.read_int}" - end - ensure - LibRubyParser.pm_source_free(pm_source) if pm_source && !pm_source.null? - end - end - end +# This file is responsible for mirroring the API provided by the C extension. There +# are two backends: +# +# * Native FFI based on the 'ffi' gem +# * WASM compiled to JVM bytecode (JRuby only) + +require_relative "ffi/common" + +begin + require_relative "ffi/native_ffi.rb" +rescue LoadError + if RUBY_ENGINE == "jruby" + require_relative "ffi/wasm_ffi.rb" + else + raise end +end - # Mark the LibRubyParser module as private as it should only be called through - # the prism module. - private_constant :LibRubyParser +module Prism # :nodoc: # The version constant is set by reading the result of calling pm_version. - VERSION = LibRubyParser.pm_version.read_string.freeze + VERSION = FFICommon.version class << self # Mirror the Prism.dump API by using the serialization API. def dump(source, **options) - LibRubyParser::PrismSource.with_string(source) { |string| dump_common(string, options) } + FFICommon.with_string(source) { |string| FFICommon.dump(string, options) } end # Mirror the Prism.dump_file API by using the serialization API. def dump_file(filepath, **options) options[:filepath] = filepath - LibRubyParser::PrismSource.with_file(filepath) { |string| dump_common(string, options) } + FFICommon.with_file(filepath) { |string| FFICommon.dump(string, options) } end # Mirror the Prism.lex API by using the serialization API. def lex(code, **options) - LibRubyParser::PrismSource.with_string(code) { |string| lex_common(string, code, options) } + FFICommon.with_string(code) { |string| FFICommon.lex(string, code, options) } end # Mirror the Prism.lex_file API by using the serialization API. def lex_file(filepath, **options) options[:filepath] = filepath - LibRubyParser::PrismSource.with_file(filepath) { |string| lex_common(string, string.read, options) } + FFICommon.with_file(filepath) { |string| FFICommon.lex(string, string.read, options) } end # Mirror the Prism.parse API by using the serialization API. def parse(code, **options) - LibRubyParser::PrismSource.with_string(code) { |string| parse_common(string, code, options) } + FFICommon.with_string(code) { |string| FFICommon.parse(string, code, options) } end # Mirror the Prism.parse_file API by using the serialization API. This uses @@ -288,12 +59,12 @@ def parse(code, **options) # when it is available. def parse_file(filepath, **options) options[:filepath] = filepath - LibRubyParser::PrismSource.with_file(filepath) { |string| parse_common(string, string.read, options) } + FFICommon.with_file(filepath) { |string| FFICommon.parse(string, string.read, options) } end # Mirror the Prism.parse_stream API by using the serialization API. def parse_stream(stream, **options) - LibRubyParser::PrismBuffer.with do |buffer| + FFICommon.with_buffer do |buffer| source = +"" callback = -> (string, size, _) { raise "Expected size to be >= 0, got: #{size}" if size <= 0 @@ -306,19 +77,13 @@ def parse_stream(stream, **options) eof_callback = -> (_) { stream.eof? } - pm_source = LibRubyParser.pm_source_stream_new(nil, callback, eof_callback) - begin - LibRubyParser.pm_serialize_parse_stream(buffer.pointer, pm_source, dump_options(options)) - Prism.load(source, buffer.read, options.fetch(:freeze, false)) - ensure - LibRubyParser.pm_source_free(pm_source) if pm_source && !pm_source.null? - end + FFICommon.parse_stream(buffer, callback, eof_callback, options, source) end end # Mirror the Prism.parse_comments API by using the serialization API. def parse_comments(code, **options) - LibRubyParser::PrismSource.with_string(code) { |string| parse_comments_common(string, code, options) } + FFICommon.with_string(code) { |string| FFICommon.parse_comments(string, code, options) } end # Mirror the Prism.parse_file_comments API by using the serialization @@ -326,23 +91,23 @@ def parse_comments(code, **options) # to use mmap when it is available. def parse_file_comments(filepath, **options) options[:filepath] = filepath - LibRubyParser::PrismSource.with_file(filepath) { |string| parse_comments_common(string, string.read, options) } + FFICommon.with_file(filepath) { |string| FFICommon.parse_comments(string, string.read, options) } end # Mirror the Prism.parse_lex API by using the serialization API. def parse_lex(code, **options) - LibRubyParser::PrismSource.with_string(code) { |string| parse_lex_common(string, code, options) } + FFICommon.with_string(code) { |string| FFICommon.parse_lex(string, code, options) } end # Mirror the Prism.parse_lex_file API by using the serialization API. def parse_lex_file(filepath, **options) options[:filepath] = filepath - LibRubyParser::PrismSource.with_file(filepath) { |string| parse_lex_common(string, string.read, options) } + FFICommon.with_file(filepath) { |string| FFICommon.parse_lex(string, string.read, options) } end # Mirror the Prism.parse_success? API by using the serialization API. def parse_success?(code, **options) - LibRubyParser::PrismSource.with_string(code) { |string| parse_file_success_common(string, options) } + FFICommon.with_string(code) { |string| FFICommon.parse_file_success(string, options) } end # Mirror the Prism.parse_failure? API by using the serialization API. @@ -353,7 +118,7 @@ def parse_failure?(code, **options) # Mirror the Prism.parse_file_success? API by using the serialization API. def parse_file_success?(filepath, **options) options[:filepath] = filepath - LibRubyParser::PrismSource.with_file(filepath) { |string| parse_file_success_common(string, options) } + FFICommon.with_file(filepath) { |string| FFICommon.parse_file_success(string, options) } end # Mirror the Prism.parse_file_failure? API by using the serialization API. @@ -363,9 +128,9 @@ def parse_file_failure?(filepath, **options) # Mirror the Prism.profile API by using the serialization API. def profile(source, **options) - LibRubyParser::PrismSource.with_string(source) do |string| - LibRubyParser::PrismBuffer.with do |buffer| - LibRubyParser.pm_serialize_parse(buffer.pointer, string.pointer, string.length, dump_options(options)) + FFICommon.with_string(source) do |string| + FFICommon.with_buffer do |buffer| + FFICommon.parse_only(buffer, string, options) nil end end @@ -373,205 +138,15 @@ def profile(source, **options) # Mirror the Prism.profile_file API by using the serialization API. def profile_file(filepath, **options) - LibRubyParser::PrismSource.with_file(filepath) do |string| - LibRubyParser::PrismBuffer.with do |buffer| + FFICommon.with_file(filepath) do |string| + FFICommon.with_buffer do |buffer| options[:filepath] = filepath - LibRubyParser.pm_serialize_parse(buffer.pointer, string.pointer, string.length, dump_options(options)) + FFICommon.parse_only(buffer, string, options) nil end end end - private - - def dump_common(string, options) # :nodoc: - LibRubyParser::PrismBuffer.with do |buffer| - LibRubyParser.pm_serialize_parse(buffer.pointer, string.pointer, string.length, dump_options(options)) - - dumped = buffer.read - dumped.freeze if options.fetch(:freeze, false) - - dumped - end - end - - def lex_common(string, code, options) # :nodoc: - LibRubyParser::PrismBuffer.with do |buffer| - LibRubyParser.pm_serialize_lex(buffer.pointer, string.pointer, string.length, dump_options(options)) - Serialize.load_lex(code, buffer.read, options.fetch(:freeze, false)) - end - end - - def parse_common(string, code, options) # :nodoc: - serialized = dump_common(string, options) - Serialize.load_parse(code, serialized, options.fetch(:freeze, false)) - end - - def parse_comments_common(string, code, options) # :nodoc: - LibRubyParser::PrismBuffer.with do |buffer| - LibRubyParser.pm_serialize_parse_comments(buffer.pointer, string.pointer, string.length, dump_options(options)) - Serialize.load_parse_comments(code, buffer.read, options.fetch(:freeze, false)) - end - end - - def parse_lex_common(string, code, options) # :nodoc: - LibRubyParser::PrismBuffer.with do |buffer| - LibRubyParser.pm_serialize_parse_lex(buffer.pointer, string.pointer, string.length, dump_options(options)) - Serialize.load_parse_lex(code, buffer.read, options.fetch(:freeze, false)) - end - end - - def parse_file_success_common(string, options) # :nodoc: - LibRubyParser.pm_serialize_parse_success_p(string.pointer, string.length, dump_options(options)) - end - - # Return the value that should be dumped for the command_line option. - def dump_options_command_line(options) - command_line = options.fetch(:command_line, "") - raise ArgumentError, "command_line must be a string" unless command_line.is_a?(String) - - command_line.each_char.inject(0) do |value, char| - case char - when "a" then value | 0b000001 - when "e" then value | 0b000010 - when "l" then value | 0b000100 - when "n" then value | 0b001000 - when "p" then value | 0b010000 - when "x" then value | 0b100000 - else raise ArgumentError, "invalid command_line option: #{char}" - end - end - end - - # Return the value that should be dumped for the version option. - def dump_options_version(version) - case version - when "current" - version_string_to_number(RUBY_VERSION) || raise(CurrentVersionError, RUBY_VERSION) - when "latest", nil - 0 # Handled in pm_parser_init - when "nearest" - dump = version_string_to_number(RUBY_VERSION) - return dump if dump - if RUBY_VERSION < "3.3" - version_string_to_number("3.3") - else - 0 # Handled in pm_parser_init - end - else - version_string_to_number(version) || raise(ArgumentError, "invalid version: #{version}") - end - end - - # Converts a version string like "4.0.0" or "4.0" into a number. - # Returns nil if the version is unknown. - def version_string_to_number(version) - case version - when /\A3\.3(\.\d+)?\z/ - 1 - when /\A3\.4(\.\d+)?\z/ - 2 - when /\A3\.5(\.\d+)?\z/, /\A4\.0(\.\d+)?\z/ - 3 - when /\A4\.1(\.\d+)?\z/ - 4 - end - end - - # Convert the given options into a serialized options string. - def dump_options(options) - template = +"" - values = [] - - template << "L" - if (filepath = options[:filepath]) - values.push(filepath.bytesize, filepath.b) - template << "A*" - else - values << 0 - end - - template << "l" - values << options.fetch(:line, 1) - - template << "L" - if (encoding = options[:encoding]) - name = encoding.is_a?(Encoding) ? encoding.name : encoding - values.push(name.bytesize, name.b) - template << "A*" - else - values << 0 - end - - template << "C" - values << (options.fetch(:frozen_string_literal, false) ? 1 : 0) - - template << "C" - values << dump_options_command_line(options) - - template << "C" - values << dump_options_version(options[:version]) - - template << "C" - values << (options[:encoding] == false ? 1 : 0) - - template << "C" - values << (options.fetch(:main_script, false) ? 1 : 0) - - template << "C" - values << (options.fetch(:partial_script, false) ? 1 : 0) - - template << "C" - values << (options.fetch(:freeze, false) ? 1 : 0) - - template << "L" - if (scopes = options[:scopes]) - values << scopes.length - - scopes.each do |scope| - locals = nil - forwarding = 0 - - case scope - when Array - locals = scope - when Scope - locals = scope.locals - - scope.forwarding.each do |forward| - case forward - when :* then forwarding |= 0x1 - when :** then forwarding |= 0x2 - when :& then forwarding |= 0x4 - when :"..." then forwarding |= 0x8 - else raise ArgumentError, "invalid forwarding value: #{forward}" - end - end - else - raise TypeError, "wrong argument type #{scope.class.inspect} (expected Array or Prism::Scope)" - end - - template << "L" - values << locals.length - - template << "C" - values << forwarding - - locals.each do |local| - name = local.name - template << "L" - values << name.bytesize - - template << "A*" - values << name.b - end - end - else - values << 0 - end - - values.pack(template) - end end # Here we are going to patch StringQuery to put in the class-level methods so @@ -580,17 +155,17 @@ class StringQuery # :nodoc: class << self # Mirrors the C extension's StringQuery::local? method. def local?(string) - query(LibRubyParser.pm_string_query_local(string, string.bytesize, string.encoding.name)) + query(FFICommon.string_query_local(string)) end # Mirrors the C extension's StringQuery::constant? method. def constant?(string) - query(LibRubyParser.pm_string_query_constant(string, string.bytesize, string.encoding.name)) + query(FFICommon.string_query_constant(string)) end # Mirrors the C extension's StringQuery::method_name? method. def method_name?(string) - query(LibRubyParser.pm_string_query_method_name(string, string.bytesize, string.encoding.name)) + query(FFICommon.string_query_method_name(string)) end private diff --git a/lib/prism/ffi/common.rb b/lib/prism/ffi/common.rb new file mode 100644 index 0000000000..552810c059 --- /dev/null +++ b/lib/prism/ffi/common.rb @@ -0,0 +1,233 @@ +# frozen_string_literal: true +# :markup: markdown +# typed: ignore + +module Prism + + class Common # :nodoc: + def dump(string, options) # :nodoc: + with_buffer do |buffer| + parse_only(buffer, string, options) + + dumped = buffer.read + dumped.freeze if options.fetch(:freeze, false) + + dumped + end + end + + def parse(string, code, options) # :nodoc: + serialized = dump(string, options) + Serialize.load_parse(code, serialized, options.fetch(:freeze, false)) + end + + def lex(string, code, options) # :nodoc: + with_buffer do |buffer| + lex_only(buffer, string, options) + Serialize.load_lex(code, buffer.read, options.fetch(:freeze, false)) + end + end + + # Return the value that should be dumped for the command_line option. + def dump_options_command_line(options) + command_line = options.fetch(:command_line, "") + raise ArgumentError, "command_line must be a string" unless command_line.is_a?(String) + + command_line.each_char.inject(0) do |value, char| + case char + when "a" then value | 0b000001 + when "e" then value | 0b000010 + when "l" then value | 0b000100 + when "n" then value | 0b001000 + when "p" then value | 0b010000 + when "x" then value | 0b100000 + else raise ArgumentError, "invalid command_line option: #{char}" + end + end + end + + # Return the value that should be dumped for the version option. + def dump_options_version(version) + case version + when "current" + version_string_to_number(RUBY_VERSION) || raise(CurrentVersionError, RUBY_VERSION) + when "latest", nil + 0 # Handled in pm_parser_init + when "nearest" + dump = version_string_to_number(RUBY_VERSION) + return dump if dump + if RUBY_VERSION < "3.3" + version_string_to_number("3.3") + else + 0 # Handled in pm_parser_init + end + else + version_string_to_number(version) || raise(ArgumentError, "invalid version: #{version}") + end + end + + # Converts a version string like "4.0.0" or "4.0" into a number. + # Returns nil if the version is unknown. + def version_string_to_number(version) + case version + when /\A3\.3(\.\d+)?\z/ + 1 + when /\A3\.4(\.\d+)?\z/ + 2 + when /\A3\.5(\.\d+)?\z/, /\A4\.0(\.\d+)?\z/ + 3 + when /\A4\.1(\.\d+)?\z/ + 4 + end + end + + # Convert the given options into a serialized options string. + def dump_options(options) + template = +"" + values = [] + + template << "L" + if (filepath = options[:filepath]) + values.push(filepath.bytesize, filepath.b) + template << "A*" + else + values << 0 + end + + template << "l" + values << options.fetch(:line, 1) + + template << "L" + if (encoding = options[:encoding]) + name = encoding.is_a?(Encoding) ? encoding.name : encoding + values.push(name.bytesize, name.b) + template << "A*" + else + values << 0 + end + + template << "C" + values << (options.fetch(:frozen_string_literal, false) ? 1 : 0) + + template << "C" + values << dump_options_command_line(options) + + template << "C" + values << dump_options_version(options[:version]) + + template << "C" + values << (options[:encoding] == false ? 1 : 0) + + template << "C" + values << (options.fetch(:main_script, false) ? 1 : 0) + + template << "C" + values << (options.fetch(:partial_script, false) ? 1 : 0) + + template << "C" + values << (options.fetch(:freeze, false) ? 1 : 0) + + template << "L" + if (scopes = options[:scopes]) + values << scopes.length + + scopes.each do |scope| + locals = nil + forwarding = 0 + + case scope + when Array + locals = scope + when Scope + locals = scope.locals + + scope.forwarding.each do |forward| + case forward + when :* then forwarding |= 0x1 + when :** then forwarding |= 0x2 + when :& then forwarding |= 0x4 + when :"..." then forwarding |= 0x8 + else raise ArgumentError, "invalid forwarding value: #{forward}" + end + end + else + raise TypeError, "wrong argument type #{scope.class.inspect} (expected Array or Prism::Scope)" + end + + template << "L" + values << locals.length + + template << "C" + values << forwarding + + locals.each do |local| + name = local.name + template << "L" + values << name.bytesize + + template << "A*" + values << name.b + end + end + else + values << 0 + end + + values.pack(template) + end + + # Required APIs below + + def version # :nodoc: + raise NotImplementedError + end + + def with_buffer(&b) # :nodoc: + raise NotImplementedError + end + + def with_string(string, &b) # :nodoc: + raise NotImplementedError + end + + def with_file(string, &b) # :nodoc: + raise NotImplementedError + end + + def lex_only(buffer, string, options) # :nodoc: + raise NotImplementedError + end + + def parse_only(buffer, string, options) # :nodoc: + raise NotImplementedError + end + + def parse_stream(buffer, callback, eof_callback, options, source) # :nodoc: + raise NotImplementedError + end + + def parse_comments(string, code, options) # :nodoc: + raise NotImplementedError + end + + def parse_lex(string, code, options) # :nodoc: + raise NotImplementedError + end + + def parse_file_success(string, options) # :nodoc: + raise NotImplementedError + end + + def string_query_method_name(string) # :nodoc: + raise NotImplementedError + end + + def string_query_constant(string) # :nodoc: + raise NotImplementedError + end + + def string_query_local(string) # :nodoc: + raise NotImplementedError + end + end +end diff --git a/lib/prism/ffi/native_ffi.rb b/lib/prism/ffi/native_ffi.rb new file mode 100644 index 0000000000..c4fbbcb660 --- /dev/null +++ b/lib/prism/ffi/native_ffi.rb @@ -0,0 +1,323 @@ +# frozen_string_literal: true +# :markup: markdown +# -- +# typed: ignore + +require "rbconfig" +require "ffi" + +# We want to eagerly load this file if there are Ractors so that it does not get +# autoloaded from within a non-main Ractor. +require "prism/serialize" if defined?(Ractor) + +module Prism # :nodoc: + module LibRubyParser # :nodoc: + extend FFI::Library + + # Define the library that we will be pulling functions from. Note that this + # must align with the build shared library from make/rake. + libprism_in_build = File.expand_path("../../../build/libprism.#{RbConfig::CONFIG["SOEXT"]}", __dir__) + libprism_in_libdir = "#{RbConfig::CONFIG["libdir"]}/prism/libprism.#{RbConfig::CONFIG["SOEXT"]}" + + if File.exist?(libprism_in_build) + INCLUDE_DIR = File.expand_path("../../../include", __dir__) + ffi_lib libprism_in_build + else + INCLUDE_DIR = "#{RbConfig::CONFIG["libdir"]}/prism/include" + ffi_lib libprism_in_libdir + end + + # Convert a native C type declaration into a symbol that FFI understands. + # For example: + # + # const char * -> :pointer + # bool -> :bool + # size_t -> :size_t + # void -> :void + # + def self.resolve_type(type, callbacks) + type = type.strip + + if !type.end_with?("*") + type.delete_prefix("const ").to_sym + else + type = type.delete_suffix("*").rstrip + callbacks.include?(type.to_sym) ? type.to_sym : :pointer + end + end + + # Read through the given header file and find the declaration of each of the + # given functions. For each one, define a function with the same name and + # signature as the C function. + def self.load_exported_functions_from(header, *functions, callbacks) + File.foreach("#{INCLUDE_DIR}/#{header}") do |line| + # We only want to attempt to load exported functions. + next unless line.start_with?("PRISM_EXPORTED_FUNCTION ") + + # We only want to load the functions that we are interested in. + next unless functions.any? { |function| line.include?(function) } + + # Strip trailing attributes (PRISM_NODISCARD, PRISM_NONNULL(...), etc.) + line = line.sub(/\)(\s+PRISM_\w+(?:\([^)]*\))?)+\s*;/, ");") + + # Parse the function declaration. + unless /^PRISM_EXPORTED_FUNCTION (?.+) (?\w+)\((?.+)\);$/ =~ line + raise "Could not parse #{line}" + end + + # Delete the function from the list of functions we are looking for to + # mark it as having been found. + functions.delete(name) + + # Split up the argument types into an array, ensure we handle the case + # where there are no arguments (by explicit void). + arg_types = arg_types.split(",").map(&:strip) + arg_types = [] if arg_types == %w[void] + + # Resolve the type of the argument by dropping the name of the argument + # first if it is present. + arg_types.map! { |type| resolve_type(type.sub(/\w+$/, ""), callbacks) } + + # Attach the function using the FFI library. + attach_function name, arg_types, resolve_type(return_type, []) + end + + # If we didn't find all of the functions, raise an error. + raise "Could not find functions #{functions.inspect}" unless functions.empty? + end + + callback :pm_source_stream_fgets_t, [:pointer, :int, :pointer], :pointer + callback :pm_source_stream_feof_t, [:pointer], :int + pm_source_init_result_values = %i[PM_SOURCE_INIT_SUCCESS PM_SOURCE_INIT_ERROR_GENERIC PM_SOURCE_INIT_ERROR_DIRECTORY PM_SOURCE_INIT_ERROR_NON_REGULAR] + enum :pm_source_init_result_t, pm_source_init_result_values + enum :pm_string_query_t, [:PM_STRING_QUERY_ERROR, -1, :PM_STRING_QUERY_FALSE, :PM_STRING_QUERY_TRUE] + + # Ractor-safe lookup table for pm_source_init_result_t, since FFI's + # enum_type accesses module instance variables that are not shareable. + SOURCE_INIT_RESULT = pm_source_init_result_values.freeze + + load_exported_functions_from( + "prism/version.h", + "pm_version", + [] + ) + + load_exported_functions_from( + "prism/serialize.h", + "pm_serialize_parse", + "pm_serialize_parse_stream", + "pm_serialize_parse_comments", + "pm_serialize_lex", + "pm_serialize_parse_lex", + "pm_serialize_parse_success_p", + [] + ) + + load_exported_functions_from( + "prism/string_query.h", + "pm_string_query_local", + "pm_string_query_constant", + "pm_string_query_method_name", + [] + ) + + load_exported_functions_from( + "prism/buffer.h", + "pm_buffer_new", + "pm_buffer_value", + "pm_buffer_length", + "pm_buffer_free", + [] + ) + + load_exported_functions_from( + "prism/source.h", + "pm_source_file_new", + "pm_source_mapped_new", + "pm_source_stream_new", + "pm_source_free", + "pm_source_source", + "pm_source_length", + [:pm_source_stream_fgets_t, :pm_source_stream_feof_t] + ) + + # This object represents a pm_buffer_t. We only use it as an opaque pointer, + # so it doesn't need to know the fields of pm_buffer_t. + class NativeBuffer # :nodoc: + attr_reader :pointer + + def initialize(pointer) + @pointer = pointer + end + + def value + LibRubyParser.pm_buffer_value(pointer) + end + + def length + LibRubyParser.pm_buffer_length(pointer) + end + + def read + value.read_string(length) + end + + # Initialize a new buffer and yield it to the block. The buffer will be + # automatically freed when the block returns. + def self.with + buffer = LibRubyParser.pm_buffer_new + raise unless buffer + + begin + yield new(buffer) + ensure + LibRubyParser.pm_buffer_free(buffer) + end + end + end + + # This object represents source code to be parsed. For strings it wraps a + # pointer directly; for files it uses a pm_source_t under the hood. + class NativeSource # :nodoc: + PLATFORM_EXPECTS_UTF8 = + RbConfig::CONFIG["host_os"].match?(/bccwin|cygwin|djgpp|mingw|mswin|wince|darwin/i) + + attr_reader :pointer, :length + + def initialize(pointer, length, from_string) + @pointer = pointer + @length = length + @from_string = from_string + end + + def read + raise "should use the original String instead" if @from_string + @pointer.read_string(@length) + end + + # Yields a PrismSource backed by the given string to the block. + def self.with_string(string) + raise TypeError unless string.is_a?(String) + + length = string.bytesize + # + 1 to never get an address of 0, which pm_parser_init() asserts + FFI::MemoryPointer.new(:char, length + 1, false) do |pointer| + pointer.write_string(string) + # since we have the extra byte we might as well \0-terminate + pointer.put_char(length, 0) + return yield new(pointer, length, true) + end + end + + # Yields a PrismSource to the given block, backed by a pm_source_t. + def self.with_file(filepath) + raise TypeError unless filepath.is_a?(String) + + # On Windows and Mac, it's expected that filepaths will be encoded in + # UTF-8. If they are not, we need to convert them to UTF-8 before + # passing them into pm_source_mapped_new. + if PLATFORM_EXPECTS_UTF8 && (encoding = filepath.encoding) != Encoding::ASCII_8BIT && encoding != Encoding::UTF_8 + filepath = filepath.encode(Encoding::UTF_8) + end + + FFI::MemoryPointer.new(:int) do |result_ptr| + pm_source = LibRubyParser.pm_source_mapped_new(filepath, 0, result_ptr) + + case SOURCE_INIT_RESULT[result_ptr.read_int] + when :PM_SOURCE_INIT_SUCCESS + pointer = LibRubyParser.pm_source_source(pm_source) + length = LibRubyParser.pm_source_length(pm_source) + return yield new(pointer, length, false) + when :PM_SOURCE_INIT_ERROR_GENERIC + raise SystemCallError.new(filepath, FFI.errno) + when :PM_SOURCE_INIT_ERROR_DIRECTORY + raise Errno::EISDIR.new(filepath) + when :PM_SOURCE_INIT_ERROR_NON_REGULAR + # Fall back to reading the file through Ruby IO for non-regular + # files (pipes, character devices, etc.) + return with_string(File.read(filepath)) { |string| yield string } + else + raise "Unknown error initializing pm_source_t: #{result_ptr.read_int}" + end + ensure + LibRubyParser.pm_source_free(pm_source) if pm_source && !pm_source.null? + end + end + end + end + + # Mark the LibRubyParser module as private as it should only be called through + # the prism module. + private_constant :LibRubyParser + + class NativeCommon < Common # :nodoc: + + # The version constant is set by reading the result of calling pm_version. + def version + LibRubyParser.pm_version.read_string.freeze + end + + def with_buffer(&b) # :nodoc: + LibRubyParser::NativeBuffer.with(&b) + end + + def with_string(string, &b) # :nodoc: + LibRubyParser::NativeSource.with_string(string, &b) + end + + def with_file(string, &b) # :nodoc: + LibRubyParser::NativeSource.with_file(string, &b) + end + + def lex_only(buffer, string, options) # :nodoc: + LibRubyParser.pm_serialize_lex(buffer.pointer, string.pointer, string.length, dump_options(options)) + end + + def parse_only(buffer, string, options) # :nodoc: + LibRubyParser.pm_serialize_parse(buffer.pointer, string.pointer, string.length, dump_options(options)) + end + + def parse_stream(buffer, callback, eof_callback, options, source) # :nodoc: + pm_source = LibRubyParser.pm_source_stream_new(nil, callback, eof_callback) + begin + LibRubyParser.pm_serialize_parse_stream(buffer.pointer, pm_source, dump_options(options)) + Prism.load(source, buffer.read, options.fetch(:freeze, false)) + ensure + LibRubyParser.pm_source_free(pm_source) if pm_source && !pm_source.null? + end + end + + def parse_comments(string, code, options) # :nodoc: + with_buffer do |buffer| + LibRubyParser.pm_serialize_parse_comments(buffer.pointer, string.pointer, string.length, dump_options(options)) + Serialize.load_parse_comments(code, buffer.read, options.fetch(:freeze, false)) + end + end + + def parse_lex(string, code, options) # :nodoc: + with_buffer do |buffer| + LibRubyParser.pm_serialize_parse_lex(buffer.pointer, string.pointer, string.length, dump_options(options)) + Serialize.load_parse_lex(code, buffer.read, options.fetch(:freeze, false)) + end + end + + def parse_file_success(string, options) # :nodoc: + LibRubyParser.pm_serialize_parse_success_p(string.pointer, string.length, dump_options(options)) + end + + def string_query_method_name(string) # :nodoc: + LibRubyParser.pm_string_query_method_name(string, string.bytesize, string.encoding.name) + end + + def string_query_constant(string) # :nodoc: + LibRubyParser.pm_string_query_constant(string, string.bytesize, string.encoding.name) + end + + def string_query_local(string) # :nodoc: + LibRubyParser.pm_string_query_local(string, string.bytesize, string.encoding.name) + end + end + + FFICommon = NativeCommon.new.freeze + private_constant(:FFICommon) +end diff --git a/lib/prism/ffi/wasm_ffi.rb b/lib/prism/ffi/wasm_ffi.rb new file mode 100644 index 0000000000..560001fcb0 --- /dev/null +++ b/lib/prism/ffi/wasm_ffi.rb @@ -0,0 +1,94 @@ +# frozen_string_literal: true +# :markup: markdown +# -- +# typed: ignore + +require "rbconfig" +require "ffi" + +# We want to eagerly load this file if there are Ractors so that it does not get +# autoloaded from within a non-main Ractor. +require "prism/serialize" if defined?(Ractor) + +# Load the prism-parser-wasm jar +require 'jar-dependencies' +require_jar('org.ruby-lang', 'prism-parser-wasm-full', '0.0.2-SNAPSHOT') +require_jar('com.dylibso.chicory', 'runtime', '1.6.1') +require_jar('com.dylibso.chicory', 'wasi', '1.6.1') +require_jar('com.dylibso.chicory', 'wasm', '1.6.1') +require_jar('com.dylibso.chicory', 'log', '1.6.1') + +module Prism # :nodoc: + class WASMCommon < Common # :nodoc: + java_import org.ruby_lang.prism.wasm.full.Prism + + # TODO: concurrency + PRISM = org.ruby_lang.prism.wasm.full.Prism.new + + def version + # The version constant is set by reading the result of calling pm_version. + PRISM.version + end + + def with_buffer(&b) # :nodoc: + buffer = Prism::Buffer.new + begin + b.call(buffer) + ensure + buffer.close + end + end + + def with_string(string, &b) # :nodoc: + source = Prism::Source.new(string.to_java_bytes) + begin + b.call(source) + ensure + source.close + end + end + + def with_file(string, &b) # :nodoc: + raise NotImplementedError + end + + def lex_only(buffer, string, options) # :nodoc: + String.from_java_bytes(Prism.lex(buffer, string, dump_options(options))) + end + + def parse_only(buffer, string, options) # :nodoc: + String.from_java_bytes(Prism.lex(buffer, string, dump_options(options))) + end + + def parse_stream(buffer, callback, eof_callback, options, source) # :nodoc: + raise NotImplementedError + end + + def parse_comments(string, code, options) # :nodoc: + raise NotImplementedError + end + + def parse_lex(string, code, options) # :nodoc: + raise NotImplementedError + end + + def parse_file_success(string, options) # :nodoc: + raise NotImplementedError + end + + def string_query_method_name(string) # :nodoc: + raise NotImplementedError + end + + def string_query_constant(string) # :nodoc: + raise NotImplementedError + end + + def string_query_local(string) # :nodoc: + raise NotImplementedError + end + end + + FFICommon = WASMCommon.new.freeze + private_constant(:FFICommon) +end diff --git a/prism.gemspec b/prism.gemspec index aac056b3f8..449b9c052c 100644 --- a/prism.gemspec +++ b/prism.gemspec @@ -113,6 +113,9 @@ Gem::Specification.new do |spec| "lib/prism/dot_visitor.rb", "lib/prism/dsl.rb", "lib/prism/ffi.rb", + "lib/prism/ffi/common.rb", + "lib/prism/ffi/native_ffi.rb", + "lib/prism/ffi/wasm_ffi.rb", "lib/prism/inspect_visitor.rb", "lib/prism/lex_compat.rb", "lib/prism/mutation_compiler.rb", diff --git a/rakelib/serialization.rake b/rakelib/serialization.rake index 516e8fe5ba..65c382fcc3 100644 --- a/rakelib/serialization.rake +++ b/rakelib/serialization.rake @@ -1,12 +1,9 @@ # frozen_string_literal: true task "test:java_loader" do - # Recompile with PRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS=1 - # Due to some JRuby bug this does not get propagated to the compile task, so require the caller to set the env var - # ENV["PRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS"] = "1" - raise "this task requires $SERIALIZE_ONLY_SEMANTICS_FIELDS to be set" unless ENV["PRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS"] - Rake::Task["clobber"].invoke + # All Java API consumers want semantic-only build + ENV["CFLAGS"] = "-DPRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS=1" Rake::Task["test:java_loader:internal"].invoke end diff --git a/src/prism.c b/src/prism.c index 72c49da6f2..61233b5d43 100644 --- a/src/prism.c +++ b/src/prism.c @@ -22843,7 +22843,11 @@ pm_serialize_header(pm_buffer_t *buffer) { pm_buffer_append_byte(buffer, PRISM_VERSION_MAJOR); pm_buffer_append_byte(buffer, PRISM_VERSION_MINOR); pm_buffer_append_byte(buffer, PRISM_VERSION_PATCH); - pm_buffer_append_byte(buffer, PRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS ? 1 : 0); + #ifdef PRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS + pm_buffer_append_byte(buffer, 1); + #else + pm_buffer_append_byte(buffer, 0); + #endif } /** diff --git a/templates/include/prism/ast.h.erb b/templates/include/prism/ast.h.erb index 3b3be25e76..5306c0abe4 100644 --- a/templates/include/prism/ast.h.erb +++ b/templates/include/prism/ast.h.erb @@ -267,12 +267,4 @@ typedef enum pm_<%= flag.human %> { PRISM_EXPORTED_FUNCTION pm_<%= node.human %>_t * pm_<%= node.human %>_new(pm_arena_t *arena, uint32_t node_id, pm_node_flags_t flags, pm_location_t location<%= params.empty? ? "" : ", #{params.join(", ")}" %>); <%- end -%> -/** - * When we're serializing to Java, we want to skip serializing the location - * fields as they won't be used by JRuby or TruffleRuby. This boolean allows us - * to specify that through the environment. It will never be true except for in - * those build systems. - */ -#define PRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS <%= Prism::Template::SERIALIZE_ONLY_SEMANTICS_FIELDS ? 1 : 0 %> - #endif diff --git a/templates/src/serialize.c.erb b/templates/src/serialize.c.erb index 3d9811e5db..2e2fa65251 100644 --- a/templates/src/serialize.c.erb +++ b/templates/src/serialize.c.erb @@ -91,8 +91,12 @@ pm_serialize_node(pm_parser_t *parser, pm_node_t *node, pm_buffer_t *buffer) { size_t length_offset = buffer->length; pm_buffer_append_string(buffer, "\0\0\0\0", 4); /* consume 4 bytes, updated below */ <%- end -%> - <%- unless Prism::Template::SERIALIZE_ONLY_SEMANTICS_FIELDS && !node.flags -%> + <%- if node.flags -%> pm_buffer_append_varuint(buffer, (uint32_t) node->flags); + <%- else -%> + #ifndef PRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS + pm_buffer_append_varuint(buffer, (uint32_t) node->flags); + #endif <%- end -%> <%- node.fields.each do |field| -%> <%- case field -%> @@ -121,17 +125,25 @@ pm_serialize_node(pm_parser_t *parser, pm_node_t *node, pm_buffer_t *buffer) { pm_buffer_append_varuint(buffer, pm_sizet_to_u32(((pm_<%= node.human %>_t *)node)-><%= field.name %>.ids[index])); } <%- when Prism::Template::LocationField -%> - <%- if field.should_be_serialized? -%> + <%- unless field.semantic_field? -%> + #ifndef PRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS + <%- end -%> pm_serialize_location(&((pm_<%= node.human %>_t *)node)-><%= field.name %>, buffer); + <%- unless field.semantic_field? -%> + #endif <%- end -%> <%- when Prism::Template::OptionalLocationField -%> - <%- if field.should_be_serialized? -%> + <%- unless field.semantic_field? -%> + #ifndef PRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS + <%- end -%> if (((pm_<%= node.human %>_t *)node)-><%= field.name %>.length == 0) { pm_buffer_append_byte(buffer, 0); } else { pm_buffer_append_byte(buffer, 1); pm_serialize_location(&((pm_<%= node.human %>_t *)node)-><%= field.name %>, buffer); } + <%- unless field.semantic_field? -%> + #endif <%- end -%> <%- when Prism::Template::UInt8Field -%> pm_buffer_append_byte(buffer, ((pm_<%= node.human %>_t *)node)-><%= field.name %>); @@ -261,9 +273,9 @@ pm_serialize_metadata(pm_parser_t *parser, pm_buffer_t *buffer) { pm_serialize_encoding(parser->encoding, buffer); pm_buffer_append_varsint(buffer, parser->start_line); pm_serialize_line_offset_list(&parser->line_offsets, buffer); -<%- unless Prism::Template::SERIALIZE_ONLY_SEMANTICS_FIELDS -%> - pm_serialize_comment_list(&parser->comment_list, buffer); -<%- end -%> + #ifndef PRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS + pm_serialize_comment_list(&parser->comment_list, buffer); + #endif pm_serialize_magic_comment_list(&parser->magic_comment_list, buffer); pm_serialize_data_loc(parser, buffer); pm_serialize_diagnostic_list(&parser->error_list, buffer); diff --git a/templates/template.rb b/templates/template.rb index 7638c9c058..92140851a3 100755 --- a/templates/template.rb +++ b/templates/template.rb @@ -7,12 +7,11 @@ module Prism module Template # :nodoc: all - SERIALIZE_ONLY_SEMANTICS_FIELDS = ENV.fetch("PRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS", false) CHECK_FIELD_KIND = ENV.fetch("CHECK_FIELD_KIND", false) JAVA_BACKEND = ENV["PRISM_JAVA_BACKEND"] || "default" JAVA_IDENTIFIER_TYPE = JAVA_BACKEND == "truffleruby" ? "String" : "byte[]" - INCLUDE_NODE_ID = !SERIALIZE_ONLY_SEMANTICS_FIELDS || JAVA_BACKEND == "jruby" + INCLUDE_NODE_ID = ENV.fetch("PRISM_INCLUDE_NODE_ID", "true") != "false" COMMON_FLAGS_COUNT = 2 @@ -95,10 +94,6 @@ def each_comment_java_line(&block) def semantic_field? true end - - def should_be_serialized? - SERIALIZE_ONLY_SEMANTICS_FIELDS ? semantic_field? : true - end end # Some node fields can be specialized if they point to a specific kind of