📄 parser.java
字号:
private void initStack() throws IOException
{
if (states == null || states.length < Parser.this.states.length)
{
states = new short[Parser.this.states.length];
min_top = 0;
}
System.arraycopy(Parser.this.states, min_top, states, min_top, (top = Parser.this.top) + 1);
}
private void increaseStackCapacity()
{
short[] new_states = new short[states.length * 2];
System.arraycopy(states, 0, new_states, 0, states.length);
states = new_states;
}
private void shift(short state)
{
if (++top == states.length)
increaseStackCapacity();
states[top] = state;
}
private short reduce(int rule_id)
{
int rule_info = tables.rule_infos[rule_id];
int rhs_size = rule_info & 0xFFFF;
top -= rhs_size;
min_top = Math.min(min_top, top);
return (short) (rule_info >>> 16);
}
}
/** The automaton tables. */
private final ParsingTables tables;
/** Cached ID of the ACCEPT action. */
private final short accept_action_id;
/** The parser's stack. */
private short[] states;
/** Index of the stack's top element, i.e. it's = -1 when the stack is empty; */
private int top;
/** The stack of shifted symbols. */
protected Symbol[] _symbols;
/** Parsing events notification "gateway" */
protected Events report;
protected Parser(ParsingTables tables)
{
this.tables = tables;
this.accept_action_id = (short) ~tables.rule_infos.length;
this.states = new short[256];
}
/**
* Parses a source and returns a semantic value of the accepted nonterminal
*
* @param source of tokens - a Scanner
* @return semantic value of the accepted nonterminal
*/
public Object parse(Scanner source) throws IOException, Parser.Exception
{
init();
return parse(new TokenStream(source));
}
/**
* Parses a source and returns a semantic value of the accepted nonterminal.
* Before parsing starts injects alternative goal marker into the source to
* indicate that an alternative goal should be matched.
*
* @param source of tokens - a Scanner
* @param alt_goal_marker_id ID of a token like symbol that will be used as a marker
* @return semantic value of the accepted nonterminal
*/
public Object parse(Scanner source, short alt_goal_marker_id) throws IOException, Parser.Exception
{
init();
TokenStream in = new TokenStream(source, new Symbol(alt_goal_marker_id));
return parse(in);
}
private Object parse(TokenStream in) throws IOException, Parser.Exception
{
while (true)
{
Symbol token = in.nextToken();
while (true)
{
short act = tables.findParserAction(states[top], token.id);
if (act > 0)
{
shift(token, act);
break;
}
else if (act == accept_action_id)
{
Symbol goal = _symbols[top];
_symbols = null; // drop this stack to prevent loitering
return goal.value;
}
else if (act < 0)
{
Symbol nt = reduce(~act);
act = tables.findNextState(states[top], nt.id);
if (act > 0)
{
shift(nt, act);
}
else if (act == accept_action_id)
{
_symbols = null; // no loitering
return nt.value;
}
else
{
throw new IllegalStateException("Cannot shift a nonterminal");
}
}
else // act == 0, i.e. this is an error
{
report.syntaxError(token);
recoverFromError(token, in);
break; // because error recovery altered token stream - parser needs to refetch the next token
}
}
}
}
/**
* Invoke actual reduce action routine.
* Method must be implemented by a generated parser
*
* @param rule_num ID of a reduce action routine to invoke
* @param offset to the symbol before first action routine argument
* @return reduced nonterminal
*/
protected abstract Symbol invokeReduceAction(int rule_num, int offset);
/**
* Performs stacks and, if not initialized yet, reduce actions array initialization.
*/
private void init()
{
if (report == null) report = new Events();
_symbols = new Symbol[states.length];
top = 0; // i.e. it's not empty
_symbols[top] = new Symbol("none"); // need a symbol here for a default reduce on the very first erroneous token
states[top] = 1; // initial/first state
}
/**
* Increases the stack capacity if it has no room for new entries.
*/
private void increaseStackCapacity()
{
short[] new_states = new short[states.length * 2];
System.arraycopy(states, 0, new_states, 0, states.length);
states = new_states;
Symbol[] new_stack = new Symbol[states.length];
System.arraycopy(_symbols, 0, new_stack, 0, _symbols.length);
_symbols = new_stack;
}
/**
* Shift a symbol to stack and go to a new state
*
* @param sym
* symbol that will be shifted
* @param goto_state
* to switch to
*/
private void shift(Symbol sym, short goto_state)
{
if (++top == states.length)
increaseStackCapacity();
_symbols[top] = sym;
states[top] = goto_state;
}
/**
* Perform a reduce action.
*
* @param rule_id
* Number of the production by which to reduce
* @return nonterminal created by a reduction
*/
private Symbol reduce(int rule_id)
{
int rule_info = tables.rule_infos[rule_id];
int rhs_size = rule_info & 0xFFFF;
top -= rhs_size;
Symbol lhs_sym = invokeReduceAction(rule_id, top);
lhs_sym.id = (short) (rule_info >>> 16);
if (rhs_size == 0)
{
lhs_sym.start = lhs_sym.end = _symbols[top].end;
}
else
{
lhs_sym.start = _symbols[top + 1].start;
lhs_sym.end = _symbols[top + rhs_size].end;
}
return lhs_sym;
}
/**
* Implements parsing error recovery. Tries several simple approches first, like deleting "bad" token
* or replacing the latter with one of the expected in his state (if possible). If simple methods did
* not work tries to do error phrase recovery.
*
* It is expected that normally descendand parsers do not need to alter this method. In same cases though
* they may want to override it if they need a different error recovery strategy.
*
* @param token a lookahead terminal symbol that messed parsing
* @param in token stream
* @throws IOException propagated from a scanner if it has issues with the source
* @throws Parser.Exception if Parser cannot recover
*/
protected void recoverFromError(Symbol token, TokenStream in) throws IOException, Parser.Exception
{
if (token.id == 0) // end of input
throw new Parser.Exception("Cannot recover from the syntax error");
Simulator sim = new Simulator();
in.mark(3);
if (sim.parse(in)) // just delete "token" from the stream
{
in.reset();
report.unexpectedTokenRemoved(token);
return;
}
short current_state = states[top];
if (!tables.compressed) // then try other simple recoveries
{
short first_term_id = tables.findFirstTerminal(current_state);
if (first_term_id >= 0)
{
Symbol term = new Symbol(first_term_id, _symbols[top].end, token.start);
in.insert(term, token); // insert expected terminal before unexpected one
in.reset();
if (sim.parse(in))
{
in.reset();
report.missingTokenInserted(term);
return;
}
int offset = tables.actn_offsets[current_state];
for (short term_id = (short) (first_term_id + 1); term_id < tables.n_term; term_id++)
{
int index = offset + term_id;
if (index >= tables.lookaheads.length)
break;
if (tables.lookaheads[index] == term_id)
{
term.id = term_id;
in.reset();
if (sim.parse(in))
{
in.reset();
report.missingTokenInserted(term);
return;
}
}
}
in.remove(1); // unexpected token, i.e. alter stream as if we replaced
// an unexpected token to an expected terminal
term.start = token.start;
term.end = token.end;
for (short term_id = first_term_id; term_id < tables.n_term; term_id++)
{
int index = offset + term_id;
if (index >= tables.lookaheads.length)
break;
if (tables.lookaheads[index] == term_id)
{
term.id = term_id;
in.reset();
if (sim.parse(in))
{
in.reset();
report.misspelledTokenReplaced(term);
return;
}
}
}
in.remove(0); // simple recoveries failed - remove all stream changes
}
}
// Simple recoveries failed or are not applicable. Next step is an error phrase recovery.
/*
* Find a state where parser can shift "error" symbol. Discard already reduced (and shifted)
* productions, which are part of a phrase where unexpected terminal is found. (Note that if
* "error" symbol was not used by a grammar, in the end the entire input becomes an error phrase,
* and ... parser won't recover from it :)
*/
Symbol first_sym = token, last_sym = token;
short goto_state;
while ((goto_state = tables.findNextState(states[top], tables.error_symbol_id)) <= 0)
{
// parser cannot shift "error" in this state, so use the top symbol
// as the leftmost symbol of an error phrase
first_sym = _symbols[top];
// and go to the previous state
if (--top < 0)
throw new Parser.Exception("Cannot recover from the syntax error");
}
Symbol error = new Symbol(tables.error_symbol_id, first_sym.start, last_sym.end); // the end is temporary
shift(error, goto_state);
in.reset();
while (!sim.parse(in))
{
last_sym = in.remove(0);
if (last_sym.id == 0) // EOF
throw new Parser.Exception("Cannot recover from the syntax error");
in.reset();
}
error.end = last_sym.end;
in.reset();
report.errorPhraseRemoved(error);
}
}
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -