Refactor c code generator
It's been rewritten in a less functional style. String copies were actually taking significant time for large parsers.
This commit is contained in:
Max Brunsfeld
parent
1daaf4485f
commit
c312f985c8
3 changed files with 224 additions and 189 deletions
|
|
@ -1,6 +1,7 @@
|
|||
#include <map>
|
||||
#include <set>
|
||||
#include <string>
|
||||
#include <functional>
|
||||
#include <utility>
|
||||
#include <vector>
|
||||
#include "compiler/generate_code/c_code.h"
|
||||
|
|
@ -11,45 +12,18 @@
|
|||
namespace tree_sitter {
|
||||
using std::string;
|
||||
using std::to_string;
|
||||
using std::function;
|
||||
using std::map;
|
||||
using std::vector;
|
||||
using std::set;
|
||||
using std::pair;
|
||||
using util::join;
|
||||
using util::indent;
|
||||
using util::escape_char;
|
||||
|
||||
namespace generate_code {
|
||||
string _switch(string condition, string body) {
|
||||
return join({
|
||||
"switch (" + condition + ") {",
|
||||
indent(body),
|
||||
"}"
|
||||
});
|
||||
}
|
||||
|
||||
string _case(string value, string body) {
|
||||
return join({
|
||||
"case " + value + ":",
|
||||
indent(body), ""
|
||||
});
|
||||
}
|
||||
|
||||
string _default(string body) {
|
||||
return join({
|
||||
"default:",
|
||||
indent(body)
|
||||
});
|
||||
}
|
||||
|
||||
string _if(string condition, string body) {
|
||||
return join({
|
||||
"if (" + condition + ")",
|
||||
indent(body), ""
|
||||
});
|
||||
}
|
||||
|
||||
class CCodeGenerator {
|
||||
string buffer;
|
||||
size_t indent_level;
|
||||
|
||||
const string name;
|
||||
const ParseTable parse_table;
|
||||
const LexTable lex_table;
|
||||
|
|
@ -63,6 +37,7 @@ namespace tree_sitter {
|
|||
const LexTable &lex_table,
|
||||
const PreparedGrammar &syntax_grammar,
|
||||
const PreparedGrammar &lexical_grammar) :
|
||||
indent_level(0),
|
||||
name(name),
|
||||
parse_table(parse_table),
|
||||
lex_table(lex_table),
|
||||
|
|
@ -70,21 +45,56 @@ namespace tree_sitter {
|
|||
lexical_grammar(lexical_grammar) {}
|
||||
|
||||
string code() {
|
||||
return join({
|
||||
includes(),
|
||||
state_and_symbol_counts(),
|
||||
symbol_enum(),
|
||||
symbol_names_list(),
|
||||
ubiquitous_symbols_list(),
|
||||
hidden_symbols_list(),
|
||||
lex_function(),
|
||||
lex_states_list(),
|
||||
parse_table_array(),
|
||||
parser_export(),
|
||||
}, "\n\n") + "\n";
|
||||
buffer = "";
|
||||
|
||||
includes();
|
||||
state_and_symbol_counts();
|
||||
symbol_enum();
|
||||
symbol_names_list();
|
||||
ubiquitous_symbols_list();
|
||||
hidden_symbols_list();
|
||||
lex_function();
|
||||
lex_states_list();
|
||||
parse_table_array();
|
||||
parser_export();
|
||||
|
||||
return buffer;
|
||||
}
|
||||
|
||||
private:
|
||||
void _switch(string condition, function<void()> body) {
|
||||
line("switch (" + condition + ") {");
|
||||
indent(body);
|
||||
line("}");
|
||||
}
|
||||
|
||||
void _case(string value, function<void()> body) {
|
||||
line("case " + value + ":");
|
||||
indent(body);
|
||||
}
|
||||
|
||||
void _default(function<void()> body) {
|
||||
line("default:");
|
||||
indent(body);
|
||||
}
|
||||
|
||||
void _if(function<void()> condition, function<void()> body) {
|
||||
line("if (");
|
||||
indent(condition);
|
||||
add(")");
|
||||
indent(body);
|
||||
}
|
||||
|
||||
void indent(function<void()> body) {
|
||||
indent(body, 1);
|
||||
}
|
||||
|
||||
void indent(function<void()> body, size_t n) {
|
||||
indent_level += n;
|
||||
body();
|
||||
indent_level -= n;
|
||||
}
|
||||
|
||||
const PreparedGrammar & grammar_for_symbol(const rules::Symbol &symbol) {
|
||||
return symbol.is_token() ? lexical_grammar : syntax_grammar;
|
||||
}
|
||||
|
|
@ -161,177 +171,226 @@ namespace tree_sitter {
|
|||
}
|
||||
}
|
||||
|
||||
string condition_for_character_set(const rules::CharacterSet &set) {
|
||||
vector<string> parts;
|
||||
void condition_for_character_set(const rules::CharacterSet &set) {
|
||||
if (set.ranges.size() == 1) {
|
||||
return condition_for_character_range(*set.ranges.begin());
|
||||
add(condition_for_character_range(*set.ranges.begin()));
|
||||
} else {
|
||||
for (auto &match : set.ranges)
|
||||
parts.push_back("(" + condition_for_character_range(match) + ")");
|
||||
return join(parts, " ||\n ");
|
||||
bool first = true;
|
||||
for (auto &match : set.ranges) {
|
||||
string part = "(" + condition_for_character_range(match) + ")";
|
||||
if (first) {
|
||||
add(part);
|
||||
} else {
|
||||
add(" ||");
|
||||
line(part);
|
||||
}
|
||||
first = false;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
string condition_for_character_rule(const rules::CharacterSet &rule) {
|
||||
vector<string> parts;
|
||||
void condition_for_character_rule(const rules::CharacterSet &rule) {
|
||||
pair<rules::CharacterSet, bool> representation = rule.most_compact_representation();
|
||||
if (representation.second)
|
||||
return condition_for_character_set(representation.first);
|
||||
else
|
||||
return "!(" + condition_for_character_set(rule.complement()) + ")";
|
||||
if (representation.second) {
|
||||
condition_for_character_set(representation.first);
|
||||
} else {
|
||||
add("!(");
|
||||
condition_for_character_set(rule.complement());
|
||||
add(")");
|
||||
}
|
||||
}
|
||||
|
||||
string code_for_parse_action(const ParseAction &action) {
|
||||
void code_for_parse_action(const ParseAction &action) {
|
||||
switch (action.type) {
|
||||
case ParseActionTypeAccept:
|
||||
return "ACCEPT_INPUT()";
|
||||
add("ACCEPT_INPUT()");
|
||||
break;
|
||||
case ParseActionTypeShift:
|
||||
return "SHIFT(" + to_string(action.state_index) + ")";
|
||||
add("SHIFT(" + to_string(action.state_index) + ")");
|
||||
break;
|
||||
case ParseActionTypeReduce:
|
||||
return "REDUCE(" +
|
||||
add("REDUCE(" +
|
||||
symbol_id(action.symbol) + ", " +
|
||||
to_string(action.consumed_symbol_count) + ")";
|
||||
default:
|
||||
return "";
|
||||
to_string(action.consumed_symbol_count) + ")");
|
||||
break;
|
||||
default:;
|
||||
}
|
||||
}
|
||||
|
||||
string code_for_lex_actions(const LexAction &action,
|
||||
void code_for_lex_actions(const LexAction &action,
|
||||
const set<rules::CharacterSet> &expected_inputs) {
|
||||
switch (action.type) {
|
||||
case LexActionTypeAdvance:
|
||||
return "ADVANCE(" + lex_state_index(action.state_index) + ");";
|
||||
line("ADVANCE(" + lex_state_index(action.state_index) + ");");
|
||||
break;
|
||||
case LexActionTypeAccept:
|
||||
return "ACCEPT_TOKEN(" + symbol_id(action.symbol) + ");";
|
||||
line("ACCEPT_TOKEN(" + symbol_id(action.symbol) + ");");
|
||||
break;
|
||||
case LexActionTypeError:
|
||||
return "LEX_ERROR();";
|
||||
default:
|
||||
return "";
|
||||
line("LEX_ERROR();");
|
||||
break;
|
||||
default:;
|
||||
}
|
||||
}
|
||||
|
||||
string code_for_lex_state(const LexState &lex_state) {
|
||||
string result = "";
|
||||
void code_for_lex_state(const LexState &lex_state) {
|
||||
auto expected_inputs = lex_state.expected_inputs();
|
||||
if (lex_state.is_token_start)
|
||||
result += "START_TOKEN();" "\n";
|
||||
line("START_TOKEN();");
|
||||
for (auto pair : lex_state.actions)
|
||||
if (!pair.first.is_empty())
|
||||
result += _if(condition_for_character_rule(pair.first),
|
||||
code_for_lex_actions(pair.second, expected_inputs));
|
||||
result += code_for_lex_actions(lex_state.default_action, expected_inputs);
|
||||
return result;
|
||||
_if([&]() { condition_for_character_rule(pair.first); },
|
||||
[&]() { code_for_lex_actions(pair.second, expected_inputs); });
|
||||
code_for_lex_actions(lex_state.default_action, expected_inputs);
|
||||
}
|
||||
|
||||
string switch_on_lex_state() {
|
||||
string body = "";
|
||||
for (size_t i = 0; i < lex_table.states.size(); i++)
|
||||
body += _case(lex_state_index(i), code_for_lex_state(lex_table.states[i]));
|
||||
body += _case("ts_lex_state_error", code_for_lex_state(lex_table.error_state));
|
||||
body += _default("LEX_PANIC();");
|
||||
return _switch("lex_state", body);
|
||||
}
|
||||
|
||||
string state_and_symbol_counts() {
|
||||
return join({
|
||||
"#define STATE_COUNT " + to_string(parse_table.states.size()),
|
||||
"#define SYMBOL_COUNT " + to_string(parse_table.symbols.size())
|
||||
void switch_on_lex_state() {
|
||||
_switch("lex_state", [&]() {
|
||||
for (size_t i = 0; i < lex_table.states.size(); i++)
|
||||
_case(lex_state_index(i), [&]() {
|
||||
code_for_lex_state(lex_table.states[i]);
|
||||
});
|
||||
_case("ts_lex_state_error", [&]() {
|
||||
code_for_lex_state(lex_table.error_state);
|
||||
});
|
||||
_default([&]() {
|
||||
line("LEX_PANIC();");
|
||||
});
|
||||
});
|
||||
}
|
||||
|
||||
string symbol_enum() {
|
||||
string result = "enum {\n";
|
||||
bool at_start = true;
|
||||
for (auto symbol : parse_table.symbols)
|
||||
if (!symbol.is_built_in()) {
|
||||
if (at_start)
|
||||
result += indent(symbol_id(symbol)) + " = ts_start_sym,\n";
|
||||
else
|
||||
result += indent(symbol_id(symbol)) + ",\n";
|
||||
at_start = false;
|
||||
}
|
||||
return result + "};";
|
||||
void state_and_symbol_counts() {
|
||||
line("#define STATE_COUNT " + to_string(parse_table.states.size()));
|
||||
line("#define SYMBOL_COUNT " + to_string(parse_table.symbols.size()));
|
||||
line();
|
||||
}
|
||||
|
||||
string symbol_names_list() {
|
||||
void symbol_enum() {
|
||||
line("enum {");
|
||||
indent([&]() {
|
||||
bool at_start = true;
|
||||
for (auto symbol : parse_table.symbols)
|
||||
if (!symbol.is_built_in()) {
|
||||
if (at_start)
|
||||
line(symbol_id(symbol) + " = ts_start_sym,");
|
||||
else
|
||||
line(symbol_id(symbol) + ",");
|
||||
at_start = false;
|
||||
}
|
||||
});
|
||||
line("};");
|
||||
line();
|
||||
}
|
||||
|
||||
void symbol_names_list() {
|
||||
set<rules::Symbol> symbols(parse_table.symbols);
|
||||
symbols.insert(rules::END_OF_INPUT());
|
||||
symbols.insert(rules::ERROR());
|
||||
|
||||
string result = "SYMBOL_NAMES = {\n";
|
||||
for (auto symbol : parse_table.symbols)
|
||||
result += indent("[" + symbol_id(symbol) + "] = \"" + symbol_name(symbol)) + "\",\n";
|
||||
return result + "};";
|
||||
}
|
||||
|
||||
string ubiquitous_symbols_list() {
|
||||
string result = "UBIQUITOUS_SYMBOLS = {\n";
|
||||
for (auto &symbol : syntax_grammar.ubiquitous_tokens())
|
||||
result += indent("[" + symbol_id(symbol) + "] = 1,") + "\n";
|
||||
return result + "};";
|
||||
}
|
||||
|
||||
string hidden_symbols_list() {
|
||||
string result = "HIDDEN_SYMBOLS = {\n";
|
||||
for (auto &symbol : parse_table.symbols)
|
||||
if (!symbol.is_built_in() && (symbol.is_auxiliary() || grammar_for_symbol(symbol).rule_name(symbol)[0] == '_'))
|
||||
result += indent("[" + symbol_id(symbol) + "] = 1,") + "\n";
|
||||
return result + "};";
|
||||
}
|
||||
|
||||
string includes() {
|
||||
return "#include \"tree_sitter/parser.h\"";
|
||||
}
|
||||
|
||||
string lex_function() {
|
||||
return join({
|
||||
"LEX_FN() {",
|
||||
indent("START_LEXER();"),
|
||||
indent(switch_on_lex_state()),
|
||||
"}"
|
||||
line("SYMBOL_NAMES = {");
|
||||
indent([&]() {
|
||||
for (auto symbol : parse_table.symbols)
|
||||
line("[" + symbol_id(symbol) + "] = \"" + symbol_name(symbol) + "\",");
|
||||
});
|
||||
line("};");
|
||||
line();
|
||||
}
|
||||
|
||||
template<typename T>
|
||||
vector<string> map_to_string(const vector<T> &inputs, std::function<string(T)> f) {
|
||||
vector<string> result;
|
||||
for (auto &item : inputs)
|
||||
result.push_back(f(item));
|
||||
return result;
|
||||
void ubiquitous_symbols_list() {
|
||||
line("UBIQUITOUS_SYMBOLS = {");
|
||||
indent([&]() {
|
||||
for (auto &symbol : syntax_grammar.ubiquitous_tokens())
|
||||
line("[" + symbol_id(symbol) + "] = 1,");
|
||||
});
|
||||
line("};");
|
||||
line();
|
||||
}
|
||||
|
||||
string lex_states_list() {
|
||||
void hidden_symbols_list() {
|
||||
line("HIDDEN_SYMBOLS = {");
|
||||
indent([&]() {
|
||||
for (auto &symbol : parse_table.symbols)
|
||||
if (!symbol.is_built_in() && (symbol.is_auxiliary() || grammar_for_symbol(symbol).rule_name(symbol)[0] == '_'))
|
||||
line("[" + symbol_id(symbol) + "] = 1,");
|
||||
});
|
||||
line("};");
|
||||
line();
|
||||
}
|
||||
|
||||
void includes() {
|
||||
add("#include \"tree_sitter/parser.h\"");
|
||||
line();
|
||||
}
|
||||
|
||||
void lex_function() {
|
||||
line("LEX_FN() {");
|
||||
indent([&]() {
|
||||
line("START_LEXER();");
|
||||
switch_on_lex_state();
|
||||
});
|
||||
line("}");
|
||||
line();
|
||||
}
|
||||
|
||||
void lex_states_list() {
|
||||
line("LEX_STATES = {");
|
||||
indent([&]() {
|
||||
size_t state_id = 0;
|
||||
for (auto &state : parse_table.states)
|
||||
line("[" + to_string(state_id++) + "] = " + lex_state_index(state.lex_state_id) + ",");
|
||||
});
|
||||
line("};");
|
||||
line();
|
||||
}
|
||||
|
||||
void parse_table_array() {
|
||||
size_t state_id = 0;
|
||||
return join({
|
||||
"LEX_STATES = {",
|
||||
indent(join(map_to_string<ParseState>(parse_table.states, [&](ParseState state) {
|
||||
return "[" + to_string(state_id++) + "] = " + lex_state_index(state.lex_state_id) + ",";
|
||||
}))),
|
||||
"};"
|
||||
line("#pragma GCC diagnostic push");
|
||||
line("#pragma GCC diagnostic ignored \"-Wmissing-field-initializers\"");
|
||||
line();
|
||||
line("PARSE_TABLE = {");
|
||||
|
||||
indent([&]() {
|
||||
for (auto &state : parse_table.states) {
|
||||
line("[" + to_string(state_id++) + "] = {");
|
||||
indent([&]() {
|
||||
for (auto &pair : state.actions) {
|
||||
line("[" + symbol_id(pair.first) + "] = ");
|
||||
code_for_parse_action(pair.second);
|
||||
add(",");
|
||||
}
|
||||
});
|
||||
line("},");
|
||||
}
|
||||
});
|
||||
|
||||
line("};");
|
||||
line();
|
||||
line("#pragma GCC diagnostic pop");
|
||||
line();
|
||||
}
|
||||
|
||||
string parse_table_array() {
|
||||
size_t state_id = 0;
|
||||
return join({
|
||||
"#pragma GCC diagnostic push",
|
||||
"#pragma GCC diagnostic ignored \"-Wmissing-field-initializers\"",
|
||||
"",
|
||||
"PARSE_TABLE = {",
|
||||
indent(join(map_to_string<ParseState>(parse_table.states, [&](ParseState state) {
|
||||
string result = "[" + to_string(state_id++) + "] = {\n";
|
||||
for (auto &pair : state.actions)
|
||||
result += indent("[" + symbol_id(pair.first) + "] = " + code_for_parse_action(pair.second) + ",") + "\n";
|
||||
return result + "},";
|
||||
}), "\n")),
|
||||
"};",
|
||||
"",
|
||||
"#pragma GCC diagnostic pop"
|
||||
});
|
||||
void parser_export() {
|
||||
line("EXPORT_PARSER(ts_parser_" + name + ");");
|
||||
line();
|
||||
}
|
||||
|
||||
string parser_export() {
|
||||
return "EXPORT_PARSER(ts_parser_" + name + ");";
|
||||
void line() {
|
||||
line("");
|
||||
}
|
||||
|
||||
void line(string input) {
|
||||
add("\n");
|
||||
if (!input.empty()) {
|
||||
string space;
|
||||
for (size_t i = 0; i < indent_level; i++)
|
||||
space += " ";
|
||||
add(space + input);
|
||||
}
|
||||
}
|
||||
|
||||
void add(string input) {
|
||||
buffer += input;
|
||||
}
|
||||
};
|
||||
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue