You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

258 lines
9.5 KiB

"use strict";
var __importDefault = (this && this.__importDefault) || function (mod) {
return (mod && mod.__esModule) ? mod : { "default": mod };
};
var __importStar = (this && this.__importStar) || function (mod) {
if (mod && mod.__esModule) return mod;
var result = {};
if (mod != null) for (var k in mod) if (Object.hasOwnProperty.call(mod, k)) result[k] = mod[k];
result["default"] = mod;
return result;
};
Object.defineProperty(exports, "__esModule", { value: true });
var assert_1 = __importDefault(require("assert"));
var types = __importStar(require("ast-types"));
var b = types.builders;
var isObject = types.builtInTypes.object;
var isArray = types.builtInTypes.array;
var options_1 = require("./options");
var lines_1 = require("./lines");
var comments_1 = require("./comments");
var util = __importStar(require("./util"));
function parse(source, options) {
options = options_1.normalize(options);
var lines = lines_1.fromString(source, options);
var sourceWithoutTabs = lines.toString({
tabWidth: options.tabWidth,
reuseWhitespace: false,
useTabs: false
});
var comments = [];
var ast = options.parser.parse(sourceWithoutTabs, {
jsx: true,
loc: true,
locations: true,
range: options.range,
comment: true,
onComment: comments,
tolerant: util.getOption(options, "tolerant", true),
ecmaVersion: 6,
sourceType: util.getOption(options, "sourceType", "module")
});
// Use ast.tokens if possible, and otherwise fall back to the Esprima
// tokenizer. All the preconfigured ../parsers/* expose ast.tokens
// automatically, but custom parsers might need additional configuration
// to avoid this fallback.
var tokens = Array.isArray(ast.tokens)
? ast.tokens
: require("esprima").tokenize(sourceWithoutTabs, {
loc: true
});
// We will reattach the tokens array to the file object below.
delete ast.tokens;
// Make sure every token has a token.value string.
tokens.forEach(function (token) {
if (typeof token.value !== "string") {
token.value = lines.sliceString(token.loc.start, token.loc.end);
}
});
if (Array.isArray(ast.comments)) {
comments = ast.comments;
delete ast.comments;
}
if (ast.loc) {
// If the source was empty, some parsers give loc.{start,end}.line
// values of 0, instead of the minimum of 1.
util.fixFaultyLocations(ast, lines);
}
else {
ast.loc = {
start: lines.firstPos(),
end: lines.lastPos()
};
}
ast.loc.lines = lines;
ast.loc.indent = 0;
var file;
var program;
if (ast.type === "Program") {
program = ast;
// In order to ensure we reprint leading and trailing program
// comments, wrap the original Program node with a File node. Only
// ESTree parsers (Acorn and Esprima) return a Program as the root AST
// node. Most other (Babylon-like) parsers return a File.
file = b.file(ast, options.sourceFileName || null);
file.loc = {
start: lines.firstPos(),
end: lines.lastPos(),
lines: lines,
indent: 0
};
}
else if (ast.type === "File") {
file = ast;
program = file.program;
}
// Expose file.tokens unless the caller passed false for options.tokens.
if (options.tokens) {
file.tokens = tokens;
}
// Expand the Program's .loc to include all comments (not just those
// attached to the Program node, as its children may have comments as
// well), since sometimes program.loc.{start,end} will coincide with the
// .loc.{start,end} of the first and last *statements*, mistakenly
// excluding comments that fall outside that region.
var trueProgramLoc = util.getTrueLoc({
type: program.type,
loc: program.loc,
body: [],
comments: comments
}, lines);
program.loc.start = trueProgramLoc.start;
program.loc.end = trueProgramLoc.end;
// Passing file.program here instead of just file means that initial
// comments will be attached to program.body[0] instead of program.
comments_1.attach(comments, program.body.length ? file.program : file, lines);
// Return a copy of the original AST so that any changes made may be
// compared to the original.
return new TreeCopier(lines, tokens).copy(file);
}
exports.parse = parse;
;
var TreeCopier = function TreeCopier(lines, tokens) {
assert_1.default.ok(this instanceof TreeCopier);
this.lines = lines;
this.tokens = tokens;
this.startTokenIndex = 0;
this.endTokenIndex = tokens.length;
this.indent = 0;
this.seen = new Map;
};
var TCp = TreeCopier.prototype;
TCp.copy = function (node) {
if (this.seen.has(node)) {
return this.seen.get(node);
}
if (isArray.check(node)) {
var copy = new Array(node.length);
this.seen.set(node, copy);
node.forEach(function (item, i) {
copy[i] = this.copy(item);
}, this);
return copy;
}
if (!isObject.check(node)) {
return node;
}
util.fixFaultyLocations(node, this.lines);
var copy = Object.create(Object.getPrototypeOf(node), {
original: {
value: node,
configurable: false,
enumerable: false,
writable: true
}
});
this.seen.set(node, copy);
var loc = node.loc;
var oldIndent = this.indent;
var newIndent = oldIndent;
var oldStartTokenIndex = this.startTokenIndex;
var oldEndTokenIndex = this.endTokenIndex;
if (loc) {
// When node is a comment, we set node.loc.indent to
// node.loc.start.column so that, when/if we print the comment by
// itself, we can strip that much whitespace from the left margin of
// the comment. This only really matters for multiline Block comments,
// but it doesn't hurt for Line comments.
if (node.type === "Block" || node.type === "Line" ||
node.type === "CommentBlock" || node.type === "CommentLine" ||
this.lines.isPrecededOnlyByWhitespace(loc.start)) {
newIndent = this.indent = loc.start.column;
}
// Every node.loc has a reference to the original source lines as well
// as a complete list of source tokens.
loc.lines = this.lines;
loc.tokens = this.tokens;
loc.indent = newIndent;
// Set loc.start.token and loc.end.token such that
// loc.tokens.slice(loc.start.token, loc.end.token) returns a list of
// all the tokens that make up this node.
this.findTokenRange(loc);
}
var keys = Object.keys(node);
var keyCount = keys.length;
for (var i = 0; i < keyCount; ++i) {
var key = keys[i];
if (key === "loc") {
copy[key] = node[key];
}
else if (key === "tokens" &&
node.type === "File") {
// Preserve file.tokens (uncopied) in case client code cares about
// it, even though Recast ignores it when reprinting.
copy[key] = node[key];
}
else {
copy[key] = this.copy(node[key]);
}
}
this.indent = oldIndent;
this.startTokenIndex = oldStartTokenIndex;
this.endTokenIndex = oldEndTokenIndex;
return copy;
};
// If we didn't have any idea where in loc.tokens to look for tokens
// contained by this loc, a binary search would be appropriate, but
// because we maintain this.startTokenIndex and this.endTokenIndex as we
// traverse the AST, we only need to make small (linear) adjustments to
// those indexes with each recursive iteration.
TCp.findTokenRange = function (loc) {
// In the unlikely event that loc.tokens[this.startTokenIndex] starts
// *after* loc.start, we need to rewind this.startTokenIndex first.
while (this.startTokenIndex > 0) {
var token = loc.tokens[this.startTokenIndex];
if (util.comparePos(loc.start, token.loc.start) < 0) {
--this.startTokenIndex;
}
else
break;
}
// In the unlikely event that loc.tokens[this.endTokenIndex - 1] ends
// *before* loc.end, we need to fast-forward this.endTokenIndex first.
while (this.endTokenIndex < loc.tokens.length) {
var token = loc.tokens[this.endTokenIndex];
if (util.comparePos(token.loc.end, loc.end) < 0) {
++this.endTokenIndex;
}
else
break;
}
// Increment this.startTokenIndex until we've found the first token
// contained by this node.
while (this.startTokenIndex < this.endTokenIndex) {
var token = loc.tokens[this.startTokenIndex];
if (util.comparePos(token.loc.start, loc.start) < 0) {
++this.startTokenIndex;
}
else
break;
}
// Index into loc.tokens of the first token within this node.
loc.start.token = this.startTokenIndex;
// Decrement this.endTokenIndex until we've found the first token after
// this node (not contained by the node).
while (this.endTokenIndex > this.startTokenIndex) {
var token = loc.tokens[this.endTokenIndex - 1];
if (util.comparePos(loc.end, token.loc.end) < 0) {
--this.endTokenIndex;
}
else
break;
}
// Index into loc.tokens of the first token *after* this node.
// If loc.start.token === loc.end.token, the node contains no tokens,
// and the index is that of the next token following this node.
loc.end.token = this.endTokenIndex;
};