A generic ANTLR code generator. All code generators
Derive from this class.
A CodeGenerator knows about a Grammar data structure and
a grammar analyzer. The Grammar is walked to generate the
appropriate code for both a parser and lexer (if present).
This interface may change slightly so that the lexer is
itself living inside of a Grammar object (in which case,
this class generates only one recognizer). The main method
to call is
gen(), which initiates all code gen.
The interaction of the code generator with the analyzer is
simple: each subrule block calls deterministic() before generating
code for the block. Method deterministic() sets lookahead caches
in each Alternative object. Technically, a code generator
doesn't need the grammar analyzer if all lookahead analysis
is done at runtime, but this would result in a slower parser.
This class provides a set of support utilities to handle argument
list parsing and so on.
BITSET_OPTIMIZE_INIT_THRESHOLD
protected static final int BITSET_OPTIMIZE_INIT_THRESHOLD
If there are more than 8 long words to init in a bitset,
try to optimize it; e.g., detect runs of -1L and 0L.
DEBUG_CODE_GENERATOR
protected boolean DEBUG_CODE_GENERATOR
Use option "codeGenDebug" to generate debugging output
DEFAULT_BITSET_TEST_THRESHOLD
protected static final int DEFAULT_BITSET_TEST_THRESHOLD
DEFAULT_MAKE_SWITCH_THRESHOLD
protected static final int DEFAULT_MAKE_SWITCH_THRESHOLD
Default values for code-generation thresholds
TokenTypesFileExt
public static String TokenTypesFileExt
TokenTypesFileSuffix
public static String TokenTypesFileSuffix
antlrTool
protected Tool antlrTool
bitsetTestThreshold
protected int bitsetTestThreshold
This is a hint for the language-specific code generator.
A bitset membership test will be generated instead of an
ORed series of LA(k) comparisions for lookahead sets with
degree greater than or equal to this value.
This is modified by the grammar option "codeGenBitsetTestThreshold"
bitsetsUsed
protected Vector bitsetsUsed
List of all bitsets that must be dumped. These are Vectors of BitSet.
charFormatter
protected CharFormatter charFormatter
Object used to format characters in the target language.
subclass must initialize this to the language-specific formatter
currentOutput
protected PrintWriter currentOutput
Current output Stream
grammar
protected Grammar grammar
The grammar for which we generate code
makeSwitchThreshold
protected int makeSwitchThreshold
This is a hint for the language-specific code generator.
A switch() or language-specific equivalent will be generated instead
of a series of if/else statements for blocks with number of alternates
greater than or equal to this number of non-predicated LL(1) alternates.
This is modified by the grammar option "codeGenMakeSwitchThreshold"
tabs
protected int tabs
Current tab indentation for code output
_print
protected void _print(String s)
Output a String to the currentOutput stream.
Ignored if string is null.
_printAction
protected void _printAction(String s)
Print an action without leading tabs, attempting to
preserve the current indentation level for multi-line actions
Ignored if string is null.
s
- The action string to output
_println
protected void _println(String s)
Output a String followed by newline, to the currentOutput stream.
Ignored if string is null.
decodeLexerRuleName
public static String decodeLexerRuleName(String id)
elementsAreRange
public static boolean elementsAreRange(int[] elems)
Test if a set element array represents a contiguous range.
elems
- The array of elements representing the set, usually from BitSet.toArray().
- true if the elements are a contiguous range (with two or more).
encodeLexerRuleName
public static String encodeLexerRuleName(String id)
extractIdOfAction
protected String extractIdOfAction(String s,
int line,
int column)
Get the identifier portion of an argument-action.
The ID of an action is assumed to be a trailing identifier.
Specific code-generators may want to override this
if the language has unusual declaration syntax.
s
- The action textline
- Line used for error reporting.column
- Line used for error reporting.
- A string containing the text of the identifier
extractIdOfAction
protected String extractIdOfAction(Token t)
Get the identifier portion of an argument-action token.
The ID of an action is assumed to be a trailing identifier.
Specific code-generators may want to override this
if the language has unusual declaration syntax.
- A string containing the text of the identifier
extractTypeOfAction
protected String extractTypeOfAction(String s,
int line,
int column)
Get the type portion of an argument-action.
The type of an action is assumed to precede a trailing identifier
Specific code-generators may want to override this
if the language has unusual declaration syntax.
s
- The action textline
- Line used for error reporting.
- A string containing the text of the type
extractTypeOfAction
protected String extractTypeOfAction(Token t)
Get the type string out of an argument-action token.
The type of an action is assumed to precede a trailing identifier
Specific code-generators may want to override this
if the language has unusual declaration syntax.
- A string containing the text of the type
gen
public abstract void gen()
Generate the code for all grammars
gen
public abstract void gen(ActionElement action)
Generate code for the given grammar element.
action
- The {...} action to generate
gen
public abstract void gen(AlternativeBlock blk)
Generate code for the given grammar element.
blk
- The "x|y|z|..." block to generate
gen
public abstract void gen(BlockEndElement end)
Generate code for the given grammar element.
end
- The block-end element to generate. Block-end
elements are synthesized by the grammar parser to represent
the end of a block.
gen
public abstract void gen(CharLiteralElement atom)
Generate code for the given grammar element.
atom
- The character literal reference to generate
gen
public abstract void gen(CharRangeElement r)
Generate code for the given grammar element.
r
- The character-range reference to generate
gen
public abstract void gen(LexerGrammar g)
throws IOException
Generate the code for a parser
gen
public abstract void gen(OneOrMoreBlock blk)
Generate code for the given grammar element.
blk
- The (...)+ block to generate
gen
public abstract void gen(ParserGrammar g)
throws IOException
Generate the code for a parser
gen
public abstract void gen(RuleRefElement rr)
Generate code for the given grammar element.
rr
- The rule-reference to generate
gen
public abstract void gen(StringLiteralElement atom)
Generate code for the given grammar element.
atom
- The string-literal reference to generate
gen
public abstract void gen(TokenRangeElement r)
Generate code for the given grammar element.
r
- The token-range reference to generate
gen
public abstract void gen(TokenRefElement atom)
Generate code for the given grammar element.
atom
- The token-reference to generate
gen
public abstract void gen(TreeElement t)
Generate code for the given grammar element.
gen
public abstract void gen(TreeWalkerGrammar g)
throws IOException
Generate the code for a parser
gen
public abstract void gen(WildcardElement wc)
Generate code for the given grammar element.
wc
- The wildcard element to generate
gen
public abstract void gen(ZeroOrMoreBlock blk)
Generate code for the given grammar element.
blk
- The (...)* block to generate
genTokenInterchange
protected void genTokenInterchange(TokenManager tm)
throws IOException
Generate the token types as a text file for persistence across shared lexer/parser
getASTCreateString
public abstract String getASTCreateString(GrammarAtom atom,
String str)
Get a string for an expression to generate creating of an AST node
str
- The text of the arguments to the AST construction
getASTCreateString
public abstract String getASTCreateString(Vector v)
Get a string for an expression to generate creation of an AST subtree.
v
- A Vector of String, where each element is an expression in the target language yielding an AST node.
getBitsetName
protected String getBitsetName(int index)
Given the index of a bitset in the bitset list, generate a unique name.
Specific code-generators may want to override this
if the language does not allow '_' or numerals in identifiers.
index
- The index of the bitset in the bitset list.
getFIRSTBitSet
public String getFIRSTBitSet(String ruleName,
int k)
getFOLLOWBitSet
public String getFOLLOWBitSet(String ruleName,
int k)
mapTreeId
public abstract String mapTreeId(String id,
ActionTransInfo tInfo)
Map an identifier to it's corresponding tree-node variable.
This is context-sensitive, depending on the rule and alternative
being generated
id
- The identifier name to map
- The mapped id (which may be the same as the input), or null if the mapping is invalid due to duplicates
markBitsetForGen
protected int markBitsetForGen(BitSet p)
Add a bitset to the list of bitsets to be generated.
if the bitset is already in the list, ignore the request.
Always adds the bitset to the end of the list, so the
caller can rely on the position of bitsets in the list.
The returned position can be used to format the bitset
name, since it is invariant.
p
- Bit set to mark for code generation
- The position of the bitset in the list.
print
protected void print(String s)
Output tab indent followed by a String, to the currentOutput stream.
Ignored if string is null.
s
- The string to output.
printAction
protected void printAction(String s)
Print an action with leading tabs, attempting to
preserve the current indentation level for multi-line actions
Ignored if string is null.
s
- The action string to output
printTabs
protected void printTabs()
Output the current tab indentation. This outputs the number of tabs
indicated by the "tabs" variable to the currentOutput stream.
println
protected void println(String s)
Output tab indent followed by a String followed by newline,
to the currentOutput stream. Ignored if string is null.
processActionForSpecialSymbols
protected abstract String processActionForSpecialSymbols(String actionStr,
int line,
RuleBlock currentRule,
ActionTransInfo tInfo)
Lexically process $ and # references within the action.
This will replace #id and #(...) with the appropriate
function calls and/or variables etc...
processStringForASTConstructor
public String processStringForASTConstructor(String str)
Process a string for an simple expression for use in xx/action.g
it is used to cast simple tokens/references to the right type for
the generated language.
removeAssignmentFromDeclaration
protected String removeAssignmentFromDeclaration(String d)
Remove the assignment portion of a declaration, if any.
- the declaration without any assignment portion
reverseLexerRuleName
public static String reverseLexerRuleName(String id)
setGrammar
protected void setGrammar(Grammar g)
Set a grammar for the code generator to use
setTool
public void setTool(Tool tool)