035f5e578af48492912c9c8a94017ebf3f44771f
3 * based on parsejavascript.js by Marijn Haverbeke
5 * A parser that can be plugged into the CodeMirror system has to
6 * implement the following interface: It is a function that, when
7 * called with a string stream (stringstream.js) as an argument,
8 * returns a MochiKit-style iterator (object with a 'next' method).
9 * This iterator, when called, consumes some input from the string
10 * stream, and returns a token object. Token objects must have a
11 * 'value' property (the text they represent), a 'style' property (the
12 * CSS style that should be used to colour them). Tokens for newline
13 * characters must also have a 'lexicalContext' property, which has an
14 * 'indentation' method that can be used to determine the proper
15 * indentation level for the next line. This method optionally takes
16 * the first character of the next line as an argument, which it can
17 * use to adjust the indentation level.
19 * So far this should be easy. The hard part is that the iterator
20 * produced by the parse function must also have a 'copy' method. This
21 * method, called without arguments, returns a function representing
22 * the current state of the parser. When this function is later called
23 * with a string stream as its argument, it returns a parser iterator
24 * object that resumes parsing using the old state and the new input
25 * stream. It may assume that only one parser is active at a time, and
26 * clobber the state of the old parser (the implementation below
30 // Parse function for TypoScript. Makes use of the tokenizer from
31 // tokenizetyposcript.js. Note that your parsers do not have to be
32 // this complicated -- if you don't want to recognize local variables,
33 // in many languages it is enough to just look for braces, semicolons,
34 // parentheses, etc, and know when you are inside a string or comment.
35 var parseTypoScript = function() {
36 // Token types that can be considered to be atoms.
37 var atomicTypes
= setObject("atom", "number", "variable", "string", "regexp");
39 // Constructor for the lexical context objects.
40 function TSLexical(indented
, column
, type
, align
, prev
) {
41 // indentation at start of this line
42 this.indented
= indented
;
43 // column at which this scope was opened
45 // type of scope ('vardef', 'stat' (statement), '[', '{', or '(')
47 // '[', '{', or '(' blocks that have any text after their opening
48 // character are said to be 'aligned' -- any lines below are
49 // indented all the way to the opening character.
52 // Parent scope, if any.
55 // My favourite indentation rules.
56 TSLexical
.prototype.indentation = function(firstChar
) {
57 var closing
= firstChar
== this.type
;
59 return this.indented
+ 2;
62 return this.column
- (closing
? 1 : 0);
64 return this.indented
+ (closing
? 0 : 2);
67 // The parser-iterator-producing function itself.
68 return function(input
){
69 // Wrap the input in a token stream
70 var tokens
= tokenizeTypoScript(input
);
71 // The parser state. cc is a stack of actions that have to be
72 // performed to finish the current statement. For example we might
73 // know that we still need to find a closing parenthesis and a
74 // semicolon. Actions at the end of the stack go first. It is
75 // initialized with an infinitely looping action that consumes
77 var cc
= [statements
];
78 // Context contains information about the current local scope, the
79 // variables defined in that, and the scopes above it.
81 // The lexical scope, used mostly for indentation.
82 var lexical
= new TSLexical(-2, 0, "block", false);
83 // Current column, and the indentation at the start of the current
84 // line. Used to create lexical scope objects.
87 // Variables which are used by the mark, cont, and pass functions
88 // below to communicate with the driver loop in the 'next'
92 // The iterator object.
93 var parser
= {next
: next
, copy
: copy
};
96 // Start by performing any 'lexical' actions (adjusting the
97 // lexical variable), or the operations below will be working
98 // with the wrong lexical state.
99 while(cc
[cc
.length
- 1].lex
)
103 var token
= tokens
.next();
104 // Adjust column and indented.
105 if (token
.type
== "whitespace" && column
== 0)
106 indented
= token
.value
.length
;
107 column
+= token
.value
.length
;
108 if (token
.type
== "newline"){
109 indented
= column
= 0;
110 // If the lexical scope's align property is still undefined at
111 // the end of the line, it is an un-aligned scope.
112 if (!("align" in lexical
))
113 lexical
.align
= false;
114 // Newline tokens get a lexical context associated with them,
115 // which is used for indentation.
116 token
.lexicalContext
= lexical
;
118 // No more processing for meaningless tokens.
119 if (token
.type
== "whitespace" || token
.type
== "newline" || token
.type
== "comment")
121 // When a meaningful token is found and the lexical scope's
122 // align is undefined, it is an aligned scope.
123 if (!("align" in lexical
))
124 lexical
.align
= true;
126 // Execute actions until one 'consumes' the token and we can
127 // return it. Marked is used to
129 consume
= marked
= false;
130 // Take and execute the topmost action.
131 cc
.pop()(token
.type
, token
.name
);
133 // Marked is used to change the style of the current token.
135 token
.style
= marked
;
142 // This makes a copy of the parser state. It stores all the
143 // stateful variables in a closure, and returns a function that
144 // will restore them when called with a new input stream. Note
145 // that the cc array has to be copied, because it is contantly
146 // being modified. Lexical objects are not mutated, and context
147 // objects are not mutated in a harmful way, so they can be shared
148 // between runs of the parser.
150 var _context
= context
, _lexical
= lexical
, _cc
= cc
.concat([]), _regexp
= tokens
.regexp
, _comment
= tokens
.inComment
;
152 return function(input
){
155 cc
= _cc
.concat([]); // copies the array
156 column
= indented
= 0;
157 tokens
= tokenizeTypoScript(input
);
158 tokens
.regexp
= _regexp
;
159 tokens
.inComment
= _comment
;
164 // Helper function for pushing a number of actions onto the cc
165 // stack in reverse order.
167 for (var i
= fs
.length
- 1; i
>= 0; i
--)
170 // cont and pass are used by the action functions to add other
171 // actions to the stack. cont will cause the current token to be
172 // consumed, pass will leave it for the next action.
181 // Used to change the style of the current token.
182 function mark(style
){
186 // Push a new scope. Will automatically link the the current
188 function pushcontext(){
189 context
= {prev
: context
, vars
: {"this": true, "arguments": true}};
191 // Pop off the current scope.
192 function popcontext(){
193 context
= context
.prev
;
195 // Register a variable in the current scope.
196 function register(varname
){
199 context
.vars
[varname
] = true;
204 // Push a new lexical context of the given type.
205 function pushlex(type
){
206 var result = function(){
207 lexical
= new TSLexical(indented
, column
, type
, null, lexical
)
212 // Pop off the current lexical context.
214 lexical
= lexical
.prev
;
217 // The 'lex' flag on these actions is used by the 'next' function
218 // to know they can (and have to) be ran before moving on to the
221 // Creates an action that discards tokens until it finds one of
223 function expect(wanted
){
224 return function(type
){
225 if (type
== wanted
) cont();
226 else cont(arguments
.callee
);
230 // Looks for a statement, and then calls itself.
231 function statements(type
){
232 return pass(statement
, statements
);
234 // Dispatches various types of statements based on the type of the
236 function statement(type
){
237 if (type
== "{") cont(pushlex("{"), block
, poplex
);
238 // else if (type == "[") cont(pushlex("]"), condition, poplex);
242 // Dispatch expression types.
243 function expression(type
){
244 if (atomicTypes
.hasOwnProperty(type
)) cont(maybeoperator
);
245 else if (type
== "function") cont(functiondef
);
246 else if (type
== "keyword c") cont(expression
);
247 else if (type
== "(") cont(pushlex(")"), expression
, expect(")"), poplex
);
248 else if (type
== "operator") cont(expression
);
249 else if (type
== "[") cont(pushlex("]"), commasep(expression
), expect("]"), poplex
);
250 else if (type
== "{") cont(pushlex("}"), commasep(objprop
), expect("}"), poplex
);
252 // Called for places where operators, function calls, or
253 // subscripts are valid. Will skip on to the next action if none
255 function maybeoperator(type
){
256 if (type
== "operator") cont(expression
);
257 else if (type
== "(") cont(pushlex(")"), expression
, commasep(expression
), expect(")"), poplex
);
258 else if (type
== ".") cont(property
, maybeoperator
);
259 else if (type
== "[") cont(pushlex("]"), expression
, expect("]"), poplex
);
261 // When a statement starts with a variable name, it might be a
262 // label. If no colon follows, it's a regular statement.
263 function maybelabel(type
){
264 if (type
== ":") cont(poplex
, statement
);
265 else pass(maybeoperator
, expect(";"), poplex
);
267 // Property names need to have their style adjusted -- the
268 // tokenizer think they are variables.
269 function property(type
){
270 if (type
== "variable") {mark("property"); cont();}
272 // This parses a property and its value in an object literal.
273 function objprop(type
){
274 if (type
== "variable") mark("property");
275 if (atomicTypes
.hasOwnProperty(type
)) cont(expect(":"), expression
);
277 // Parses a comma-separated list of the things that are recognized
278 // by the 'what' argument.
279 function commasep(what
){
280 function proceed(type
) {
281 if (type
== ",") cont(what
, proceed
);
288 // Look for statements until a closing brace is found.
289 function block(type
){
290 if (type
== "}") cont();
291 else pass(statement
, block
);
294 // Look for statements until a closing brace is found.
295 function condition(type
){
296 if (type
== "]") cont();
297 else pass(statement
, block
);
301 // Variable definitions are split into two actions -- 1 looks for
302 // a name or the end of the definition, 2 looks for an '=' sign or
304 function vardef1(type
, value
){
305 if (type
== "variable"){register(value
); cont(vardef2
);}
308 function vardef2(type
){
309 if (type
== "operator") cont(expression
, vardef2
);
310 else if (type
== ",") cont(vardef1
);
313 function forspec1(type
, value
){
314 if (type
== "var") cont(vardef1
, forspec2
);
315 else cont(expression
, forspec2
);
317 function forspec2(type
){
318 if (type
== ",") cont(forspec1
);
319 if (type
== ";") cont(expression
, expect(";"), expression
);
321 // A function definition creates a new context, and the variables
322 // in its argument list have to be added to this context.
323 function functiondef(type
, value
){
324 if (type
== "variable"){register(value
); cont(functiondef
);}
325 else if (type
== "(") cont(pushcontext
, commasep(funarg
), expect(")"), statement
, popcontext
);
327 function funarg(type
, value
){
328 if (type
== "variable"){register(value
); cont();}