From 4646bf4b9221490f1eca39f894d84a34646f5421 Mon Sep 17 00:00:00 2001 From: Itamar Date: Sun, 6 Feb 2022 23:02:34 +0200 Subject: [PATCH] LibCpp: Add SemanticSyntaxHighlighter The SemanticSyntaxHighlighter uses TokenInfo results from the language server to provide 'semantic' syntax highlighting, which provides more fin-grained text spans results. For example, the SemanticSyntaxHighlighter can color function calls, member fields references and user-defined types in different colors. With the simple lexer-only syntax highlighter, all of these tokens were given the same text highlighting span type. Since we have to provide immediate highlighting feedback to the user after each edit and before we get the result for the language server, we use a heuristic which computes the diff between the current tokens and the last known tokens with compete semantic information (We use LibDiff for this). This heuristic is not very performant, and starts feeling sluggish with bigger (~200 LOC) files. A possible future improvement would be only computing the diff for tokens in text ranges that have changes since the last commit. --- Userland/Libraries/LibCpp/CMakeLists.txt | 3 +- .../LibCpp/SemanticSyntaxHighlighter.cpp | 163 ++++++++++++++++++ .../LibCpp/SemanticSyntaxHighlighter.h | 49 ++++++ Userland/Libraries/LibCpp/SyntaxHighlighter.h | 4 + Userland/Libraries/LibSyntax/Highlighter.h | 6 + 5 files changed, 224 insertions(+), 1 deletion(-) create mode 100644 Userland/Libraries/LibCpp/SemanticSyntaxHighlighter.cpp create mode 100644 Userland/Libraries/LibCpp/SemanticSyntaxHighlighter.h diff --git a/Userland/Libraries/LibCpp/CMakeLists.txt b/Userland/Libraries/LibCpp/CMakeLists.txt index 65ebc107f2..87aee2093f 100644 --- a/Userland/Libraries/LibCpp/CMakeLists.txt +++ b/Userland/Libraries/LibCpp/CMakeLists.txt @@ -3,9 +3,10 @@ set(SOURCES Lexer.cpp Parser.cpp Preprocessor.cpp + SemanticSyntaxHighlighter.cpp SyntaxHighlighter.cpp Token.cpp ) serenity_lib(LibCpp cpp) -target_link_libraries(LibCpp LibC LibSyntax) +target_link_libraries(LibCpp LibC LibSyntax LibDiff) diff --git a/Userland/Libraries/LibCpp/SemanticSyntaxHighlighter.cpp b/Userland/Libraries/LibCpp/SemanticSyntaxHighlighter.cpp new file mode 100644 index 0000000000..30200227ea --- /dev/null +++ b/Userland/Libraries/LibCpp/SemanticSyntaxHighlighter.cpp @@ -0,0 +1,163 @@ +/* + * Copyright (c) 2022, Itamar S. + * Copyright (c) 2022, the SerenityOS developers. + * + * SPDX-License-Identifier: BSD-2-Clause + */ + +#include "SemanticSyntaxHighlighter.h" +#include "Lexer.h" +#include +#include +#include + +namespace Cpp { + +void SemanticSyntaxHighlighter::rehighlight(Palette const& palette) +{ + Vector new_tokens_info; + auto text = m_client->get_text(); + { + Threading::MutexLocker locker(m_lock); + Cpp::Lexer lexer(text); + lexer.set_ignore_whitespace(true); + auto current_tokens = lexer.lex(); + + StringBuilder current_tokens_as_lines; + StringBuilder previous_tokens_as_lines; + + for (auto& token : current_tokens) + current_tokens_as_lines.appendff("{}\n", token.type_as_string()); + + for (Cpp::Token const& token : m_saved_tokens) + previous_tokens_as_lines.appendff("{}\n", token.type_as_string()); + + auto previous = previous_tokens_as_lines.build(); + auto current = current_tokens_as_lines.build(); + + // FIXME: Computing the diff on the entire document's tokens is quite inefficient. + // An improvement over this could be only including the tokens that are in edited text ranges in the diff. + auto diff_hunks = Diff::from_text(previous.view(), current.view()); + for (auto& token : current_tokens) { + new_tokens_info.append(GUI::AutocompleteProvider::TokenInfo { GUI::AutocompleteProvider::TokenInfo::SemanticType::Unknown, + token.start().line, token.start().column, token.end().line, token.end().column }); + } + size_t previous_token_index = 0; + size_t current_token_index = 0; + for (auto const& hunk : diff_hunks) { + for (; previous_token_index < hunk.original_start_line; ++previous_token_index) { + new_tokens_info[current_token_index].type = m_tokens_info[previous_token_index].type; + ++current_token_index; + } + for (size_t i = 0; i < hunk.added_lines.size(); ++i) { + ++current_token_index; + } + for (size_t i = 0; i < hunk.removed_lines.size(); ++i) { + ++previous_token_index; + } + } + while (current_token_index < new_tokens_info.size() && previous_token_index < m_tokens_info.size()) { + new_tokens_info[current_token_index].type = m_tokens_info[previous_token_index].type; + + ++previous_token_index; + ++current_token_index; + } + } + + update_spans(new_tokens_info, palette); +} + +static Syntax::TextStyle style_for_token_type(Gfx::Palette const& palette, GUI::AutocompleteProvider::TokenInfo::SemanticType type) +{ + switch (type) { + case GUI::AutocompleteProvider::TokenInfo::SemanticType::Unknown: + return { palette.base_text(), false }; + case GUI::AutocompleteProvider::TokenInfo::SemanticType::Keyword: + return { palette.syntax_keyword(), true }; + case GUI::AutocompleteProvider::TokenInfo::SemanticType::Type: + return { palette.syntax_type(), true }; + case GUI::AutocompleteProvider::TokenInfo::SemanticType::Identifier: + return { palette.syntax_identifier(), false }; + case GUI::AutocompleteProvider::TokenInfo::SemanticType::String: + return { palette.syntax_string(), false }; + case GUI::AutocompleteProvider::TokenInfo::SemanticType::Number: + return { palette.syntax_number(), false }; + case GUI::AutocompleteProvider::TokenInfo::SemanticType::IncludePath: + return { palette.syntax_preprocessor_value(), false }; + case GUI::AutocompleteProvider::TokenInfo::SemanticType::PreprocessorStatement: + return { palette.syntax_preprocessor_statement(), false }; + case GUI::AutocompleteProvider::TokenInfo::SemanticType::Comment: + return { palette.syntax_comment(), false }; + case GUI::AutocompleteProvider::TokenInfo::SemanticType::Function: + return { palette.syntax_function(), false }; + case GUI::AutocompleteProvider::TokenInfo::SemanticType::Variable: + return { palette.syntax_variable(), false }; + case GUI::AutocompleteProvider::TokenInfo::SemanticType::CustomType: + return { palette.syntax_custom_type(), false }; + case GUI::AutocompleteProvider::TokenInfo::SemanticType::Namespace: + return { palette.syntax_namespace(), false }; + case GUI::AutocompleteProvider::TokenInfo::SemanticType::Member: + return { palette.syntax_member(), false }; + case GUI::AutocompleteProvider::TokenInfo::SemanticType::Parameter: + return { palette.syntax_parameter(), false }; + default: + VERIFY_NOT_REACHED(); + return { palette.base_text(), false }; + } +} +void SemanticSyntaxHighlighter::update_spans(Vector const& tokens_info, Gfx::Palette const& pallete) +{ + Vector spans; + for (auto& token : tokens_info) { + // FIXME: The +1 for the token end column is a quick hack due to not wanting to modify the lexer (which is also used by the parser). Maybe there's a better way to do this. + GUI::TextDocumentSpan span; + span.range.set_start({ token.start_line, token.start_column }); + span.range.set_end({ token.end_line, token.end_column + 1 }); + auto style = style_for_token_type(pallete, token.type); + span.attributes.color = style.color; + span.attributes.bold = style.bold; + span.is_skippable = token.type == GUI::AutocompleteProvider::TokenInfo::SemanticType::Whitespace; + span.data = static_cast(token.type); + spans.append(span); + } + m_client->do_set_spans(move(spans)); + + m_has_brace_buddies = false; + highlight_matching_token_pair(); + + m_client->do_update(); +} + +void SemanticSyntaxHighlighter::update_tokens_info(Vector tokens_info) +{ + { + Threading::MutexLocker locker(m_lock); + m_tokens_info = move(tokens_info); + + m_saved_tokens_text = m_client->get_text(); + Cpp::Lexer lexer(m_saved_tokens_text); + lexer.set_ignore_whitespace(true); + m_saved_tokens = lexer.lex(); + } +} + +bool SemanticSyntaxHighlighter::is_identifier(u64 token_type) const +{ + using GUI::AutocompleteProvider; + auto type = static_cast(token_type); + + return type == AutocompleteProvider::TokenInfo::SemanticType::Identifier + || type == AutocompleteProvider::TokenInfo::SemanticType::Function + || type == AutocompleteProvider::TokenInfo::SemanticType::Variable + || type == AutocompleteProvider::TokenInfo::SemanticType::CustomType + || type == AutocompleteProvider::TokenInfo::SemanticType::Namespace + || type == AutocompleteProvider::TokenInfo::SemanticType::Member + || type == AutocompleteProvider::TokenInfo::SemanticType::Parameter; +} + +bool SemanticSyntaxHighlighter::is_navigatable(u64 token_type) const +{ + return static_cast(token_type) == GUI::AutocompleteProvider::TokenInfo::SemanticType::IncludePath; +} + +} diff --git a/Userland/Libraries/LibCpp/SemanticSyntaxHighlighter.h b/Userland/Libraries/LibCpp/SemanticSyntaxHighlighter.h new file mode 100644 index 0000000000..7983c35d4f --- /dev/null +++ b/Userland/Libraries/LibCpp/SemanticSyntaxHighlighter.h @@ -0,0 +1,49 @@ +/* + * Copyright (c) 2022, Itamar S. + * Copyright (c) 2022, the SerenityOS developers. + * + * SPDX-License-Identifier: BSD-2-Clause + */ + +#pragma once + +#include "SyntaxHighlighter.h" +#include "Token.h" +#include +#include +#include + +namespace Cpp { + +class SemanticSyntaxHighlighter final : public Syntax::Highlighter { + +public: + SemanticSyntaxHighlighter() { } + virtual ~SemanticSyntaxHighlighter() override = default; + + virtual bool is_identifier(u64 token) const override; + virtual bool is_navigatable(u64 token) const override; + + virtual Syntax::Language language() const override { return Syntax::Language::Cpp; } + virtual void rehighlight(Palette const&) override; + + void update_tokens_info(Vector); + + virtual bool is_cpp_semantic_highlighter() const override { return true; } + +protected: + virtual Vector matching_token_pairs_impl() const override { return m_simple_syntax_highlighter.matching_token_pairs_impl(); }; + virtual bool token_types_equal(u64 token1, u64 token2) const override { return m_simple_syntax_highlighter.token_types_equal(token1, token2); }; + +private: + void update_spans(Vector const&, Gfx::Palette const&); + + Cpp::SyntaxHighlighter m_simple_syntax_highlighter; + Vector m_tokens_info; + String m_saved_tokens_text; + Vector m_saved_tokens; + Threading::Mutex m_lock; +}; +} +template<> +inline bool Syntax::Highlighter::fast_is() const { return is_cpp_semantic_highlighter(); } diff --git a/Userland/Libraries/LibCpp/SyntaxHighlighter.h b/Userland/Libraries/LibCpp/SyntaxHighlighter.h index 5526c02b10..d83d4b7b04 100644 --- a/Userland/Libraries/LibCpp/SyntaxHighlighter.h +++ b/Userland/Libraries/LibCpp/SyntaxHighlighter.h @@ -10,7 +10,11 @@ namespace Cpp { +class SemanticSyntaxHighlighter; + class SyntaxHighlighter final : public Syntax::Highlighter { + friend SemanticSyntaxHighlighter; + public: SyntaxHighlighter() { } virtual ~SyntaxHighlighter() override; diff --git a/Userland/Libraries/LibSyntax/Highlighter.h b/Userland/Libraries/LibSyntax/Highlighter.h index bed19b3fbd..5351b462bc 100644 --- a/Userland/Libraries/LibSyntax/Highlighter.h +++ b/Userland/Libraries/LibSyntax/Highlighter.h @@ -56,6 +56,12 @@ public: }; Vector matching_token_pairs() const; + template + bool fast_is() const = delete; + + // FIXME: When other syntax highlighters start using a language server, we should add a common base class here. + virtual bool is_cpp_semantic_highlighter() const { return false; } + protected: Highlighter() { }