// Copyright 2008 the V8 project authors. All rights reserved. // Redistribution and use in source and binary forms, with or without // modification, are permitted provided that the following conditions are // met: // // * Redistributions of source code must retain the above copyright // notice, this list of conditions and the following disclaimer. // * Redistributions in binary form must reproduce the above // copyright notice, this list of conditions and the following // disclaimer in the documentation and/or other materials provided // with the distribution. // * Neither the name of Google Inc. nor the names of its // contributors may be used to endorse or promote products derived // from this software without specific prior written permission. // // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. function testEscape(str, regex) { assertEquals("foo:bar:baz", str.split(regex).join(":")); } testEscape("foo\nbar\nbaz", /\n/); testEscape("foo bar baz", /\s/); testEscape("foo\tbar\tbaz", /\s/); testEscape("foo-bar-baz", /\u002D/); // Test containing null char in regexp. var s = '[' + String.fromCharCode(0) + ']'; var re = new RegExp(s); assertEquals(s.match(re).length, 1); assertEquals(s.match(re)[0], String.fromCharCode(0)); // Test strings containing all line separators s = 'aA\nbB\rcC\r\ndD\u2028eE\u2029fF'; re = /^./gm; // any non-newline character at the beginning of a line var result = s.match(re); assertEquals(result.length, 6); assertEquals(result[0], 'a'); assertEquals(result[1], 'b'); assertEquals(result[2], 'c'); assertEquals(result[3], 'd'); assertEquals(result[4], 'e'); assertEquals(result[5], 'f'); re = /.$/gm; // any non-newline character at the end of a line result = s.match(re); assertEquals(result.length, 6); assertEquals(result[0], 'A'); assertEquals(result[1], 'B'); assertEquals(result[2], 'C'); assertEquals(result[3], 'D'); assertEquals(result[4], 'E'); assertEquals(result[5], 'F'); re = /^[^]/gm; // *any* character at the beginning of a line result = s.match(re); assertEquals(result.length, 7); assertEquals(result[0], 'a'); assertEquals(result[1], 'b'); assertEquals(result[2], 'c'); assertEquals(result[3], '\n'); assertEquals(result[4], 'd'); assertEquals(result[5], 'e'); assertEquals(result[6], 'f'); re = /[^]$/gm; // *any* character at the end of a line result = s.match(re); assertEquals(result.length, 7); assertEquals(result[0], 'A'); assertEquals(result[1], 'B'); assertEquals(result[2], 'C'); assertEquals(result[3], '\r'); assertEquals(result[4], 'D'); assertEquals(result[5], 'E'); assertEquals(result[6], 'F'); // Some tests from the Mozilla tests, where our behavior differs from // SpiderMonkey. // From ecma_3/RegExp/regress-334158.js assertTrue(/\ca/.test( "\x01" )); assertFalse(/\ca/.test( "\\ca" )); // Passes in KJS, fails in IrregularExpressions. // See http://code.google.com/p/v8/issues/detail?id=152 //assertTrue(/\c[a/]/.test( "\x1ba/]" )); // Test \c in character class re = /^[\cM]$/; assertTrue(re.test("\r")); assertFalse(re.test("M")); assertFalse(re.test("c")); assertFalse(re.test("\\")); assertFalse(re.test("\x03")); // I.e., read as \cc re = /^[\c]]$/; assertTrue(re.test("c]")); assertFalse(re.test("\\]")); assertFalse(re.test("\x1d")); // ']' & 0x1f assertFalse(re.test("\\]")); assertFalse(re.test("\x03]")); // I.e., read as \cc // Test that we handle \s and \S correctly inside some bizarre // character classes. re = /[\s-:]/; assertTrue(re.test('-')); assertTrue(re.test(':')); assertTrue(re.test(' ')); assertTrue(re.test('\t')); assertTrue(re.test('\n')); assertFalse(re.test('a')); assertFalse(re.test('Z')); re = /[\S-:]/; assertTrue(re.test('-')); assertTrue(re.test(':')); assertFalse(re.test(' ')); assertFalse(re.test('\t')); assertFalse(re.test('\n')); assertTrue(re.test('a')); assertTrue(re.test('Z')); re = /[^\s-:]/; assertFalse(re.test('-')); assertFalse(re.test(':')); assertFalse(re.test(' ')); assertFalse(re.test('\t')); assertFalse(re.test('\n')); assertTrue(re.test('a')); assertTrue(re.test('Z')); re = /[^\S-:]/; assertFalse(re.test('-')); assertFalse(re.test(':')); assertTrue(re.test(' ')); assertTrue(re.test('\t')); assertTrue(re.test('\n')); assertFalse(re.test('a')); assertFalse(re.test('Z')); re = /[\s]/; assertFalse(re.test('-')); assertFalse(re.test(':')); assertTrue(re.test(' ')); assertTrue(re.test('\t')); assertTrue(re.test('\n')); assertFalse(re.test('a')); assertFalse(re.test('Z')); re = /[^\s]/; assertTrue(re.test('-')); assertTrue(re.test(':')); assertFalse(re.test(' ')); assertFalse(re.test('\t')); assertFalse(re.test('\n')); assertTrue(re.test('a')); assertTrue(re.test('Z')); re = /[\S]/; assertTrue(re.test('-')); assertTrue(re.test(':')); assertFalse(re.test(' ')); assertFalse(re.test('\t')); assertFalse(re.test('\n')); assertTrue(re.test('a')); assertTrue(re.test('Z')); re = /[^\S]/; assertFalse(re.test('-')); assertFalse(re.test(':')); assertTrue(re.test(' ')); assertTrue(re.test('\t')); assertTrue(re.test('\n')); assertFalse(re.test('a')); assertFalse(re.test('Z')); re = /[\s\S]/; assertTrue(re.test('-')); assertTrue(re.test(':')); assertTrue(re.test(' ')); assertTrue(re.test('\t')); assertTrue(re.test('\n')); assertTrue(re.test('a')); assertTrue(re.test('Z')); re = /[^\s\S]/; assertFalse(re.test('-')); assertFalse(re.test(':')); assertFalse(re.test(' ')); assertFalse(re.test('\t')); assertFalse(re.test('\n')); assertFalse(re.test('a')); assertFalse(re.test('Z')); // Test beginning and end of line assertions with or without the // multiline flag. re = /^\d+/; assertFalse(re.test("asdf\n123")); re = /^\d+/m; assertTrue(re.test("asdf\n123")); re = /\d+$/; assertFalse(re.test("123\nasdf")); re = /\d+$/m; assertTrue(re.test("123\nasdf")); // Test that empty matches are handled correctly for multiline global // regexps. re = /^(.*)/mg; assertEquals(3, "a\n\rb".match(re).length); assertEquals("*a\n*b\r*c\n*\r*d\r*\n*e", "a\nb\rc\n\rd\r\ne".replace(re, "*$1")); // Test that empty matches advance one character re = new RegExp("", "g"); assertEquals("xAx", "A".replace(re, "x")); assertEquals(3, String.fromCharCode(161).replace(re, "x").length); // Test that we match the KJS behavior with regard to undefined constructor // arguments: re = new RegExp(); // KJS actually shows this as '//'. Here we match the Firefox behavior (ie, // giving a syntactically legal regexp literal). assertEquals('/(?:)/', re.toString()); re = new RegExp(void 0); assertEquals('/(?:)/', re.toString()); re.compile(); assertEquals('/(?:)/', re.toString()); re.compile(void 0); assertEquals('/undefined/', re.toString()); // Check for lazy RegExp literal creation function lazyLiteral(doit) { if (doit) return "".replace(/foo(/gi, ""); return true; } assertTrue(lazyLiteral(false)); assertThrows("lazyLiteral(true)"); // Check $01 and $10 re = new RegExp("(.)(.)(.)(.)(.)(.)(.)(.)(.)(.)"); assertEquals("t", "123456789t".replace(re, "$10"), "$10"); assertEquals("15", "123456789t".replace(re, "$15"), "$10"); assertEquals("1", "123456789t".replace(re, "$01"), "$01"); assertEquals("$001", "123456789t".replace(re, "$001"), "$001"); re = new RegExp("foo(.)"); assertEquals("bar$0", "foox".replace(re, "bar$0"), "$0"); assertEquals("bar$00", "foox".replace(re, "bar$00"), "$00"); assertEquals("bar$000", "foox".replace(re, "bar$000"), "$000"); assertEquals("barx", "foox".replace(re, "bar$01"), "$01 2"); assertEquals("barx5", "foox".replace(re, "bar$15"), "$15"); assertFalse(/()foo$\1/.test("football"), "football1"); assertFalse(/foo$(?=ball)/.test("football"), "football2"); assertFalse(/foo$(?!bar)/.test("football"), "football3"); assertTrue(/()foo$\1/.test("foo"), "football4"); assertTrue(/foo$(?=(ball)?)/.test("foo"), "football5"); assertTrue(/()foo$(?!bar)/.test("foo"), "football6"); assertFalse(/(x?)foo$\1/.test("football"), "football7"); assertFalse(/foo$(?=ball)/.test("football"), "football8"); assertFalse(/foo$(?!bar)/.test("football"), "football9"); assertTrue(/(x?)foo$\1/.test("foo"), "football10"); assertTrue(/foo$(?=(ball)?)/.test("foo"), "football11"); assertTrue(/foo$(?!bar)/.test("foo"), "football12"); // Check that the back reference has two successors. See // BackReferenceNode::PropagateForward. assertFalse(/f(o)\b\1/.test('foo')); assertTrue(/f(o)\B\1/.test('foo')); // Back-reference, ignore case: // ASCII assertEquals("xaAx,a", String(/x(a)\1x/i.exec("xaAx")), "backref-ASCII"); assertFalse(/x(...)\1/i.test("xaaaaa"), "backref-ASCII-short"); assertTrue(/x((?:))\1\1x/i.test("xx"), "backref-ASCII-empty"); assertTrue(/x(?:...|(...))\1x/i.test("xabcx"), "backref-ASCII-uncaptured"); assertTrue(/x(?:...|(...))\1x/i.test("xabcABCx"), "backref-ASCII-backtrack"); assertEquals("xaBcAbCABCx,aBc", String(/x(...)\1\1x/i.exec("xaBcAbCABCx")), "backref-ASCII-twice"); for (var i = 0; i < 128; i++) { var testName = "backref-ASCII-char-" + i + "," + (i^0x20); var test = /^(.)\1$/i.test(String.fromCharCode(i, i ^ 0x20)) var c = String.fromCharCode(i); if (('A' <= c && c <= 'Z') || ('a' <= c && c <= 'z')) { assertTrue(test, testName); } else { assertFalse(test, testName); } } assertFalse(/f(o)$\1/.test('foo'), "backref detects at_end"); // Check decimal escapes doesn't overflow. // (Note: \214 is interpreted as octal). assertEquals(/\2147483648/.exec("\x8c7483648"), ["\x8c7483648"], "Overflow decimal escape"); // Check numbers in quantifiers doesn't overflow and doesn't throw on // too large numbers. assertFalse(/a{111111111111111111111111111111111111111111111}/.test('b'), "overlarge1"); assertFalse(/a{999999999999999999999999999999999999999999999}/.test('b'), "overlarge2"); assertFalse(/a{1,111111111111111111111111111111111111111111111}/.test('b'), "overlarge3"); assertFalse(/a{1,999999999999999999999999999999999999999999999}/.test('b'), "overlarge4"); assertFalse(/a{2147483648}/.test('b'), "overlarge5"); assertFalse(/a{21474836471}/.test('b'), "overlarge6"); assertFalse(/a{1,2147483648}/.test('b'), "overlarge7"); assertFalse(/a{1,21474836471}/.test('b'), "overlarge8"); assertFalse(/a{2147483648,2147483648}/.test('b'), "overlarge9"); assertFalse(/a{21474836471,21474836471}/.test('b'), "overlarge10"); assertFalse(/a{2147483647}/.test('b'), "overlarge11"); assertFalse(/a{1,2147483647}/.test('b'), "overlarge12"); assertTrue(/a{1,2147483647}/.test('a'), "overlarge13"); assertFalse(/a{2147483647,2147483647}/.test('a'), "overlarge14"); // Check that we don't read past the end of the string. assertFalse(/f/.test('b')); assertFalse(/[abc]f/.test('x')); assertFalse(/[abc]f/.test('xa')); assertFalse(/[abc]= 0, "overflow"); } // Test that compile works on modified objects var re = /re+/; assertEquals("re+", re.source); assertFalse(re.global); assertFalse(re.ignoreCase); assertFalse(re.multiline); assertEquals(0, re.lastIndex); re.compile("ro+", "gim"); assertEquals("ro+", re.source); assertTrue(re.global); assertTrue(re.ignoreCase); assertTrue(re.multiline); assertEquals(0, re.lastIndex); re.lastIndex = 42; re.someOtherProperty = 42; re.someDeletableProperty = 42; re[37] = 37; re[42] = 42; re.compile("ra+", "i"); assertEquals("ra+", re.source); assertFalse(re.global); assertTrue(re.ignoreCase); assertFalse(re.multiline); assertEquals(0, re.lastIndex); assertEquals(42, re.someOtherProperty); assertEquals(42, re.someDeletableProperty); assertEquals(37, re[37]); assertEquals(42, re[42]); re.lastIndex = -1; re.someOtherProperty = 37; re[42] = 37; assertTrue(delete re[37]); assertTrue(delete re.someDeletableProperty); re.compile("ri+", "gm"); assertEquals("ri+", re.source); assertTrue(re.global); assertFalse(re.ignoreCase); assertTrue(re.multiline); assertEquals(0, re.lastIndex); assertEquals(37, re.someOtherProperty); assertEquals(37, re[42]); // Test boundary-checks. function assertRegExpTest(re, input, test) { assertEquals(test, re.test(input), "test:" + re + ":" + input); } assertRegExpTest(/b\b/, "b", true); assertRegExpTest(/b\b$/, "b", true); assertRegExpTest(/\bb/, "b", true); assertRegExpTest(/^\bb/, "b", true); assertRegExpTest(/,\b/, ",", false); assertRegExpTest(/,\b$/, ",", false); assertRegExpTest(/\b,/, ",", false); assertRegExpTest(/^\b,/, ",", false); assertRegExpTest(/b\B/, "b", false); assertRegExpTest(/b\B$/, "b", false); assertRegExpTest(/\Bb/, "b", false); assertRegExpTest(/^\Bb/, "b", false); assertRegExpTest(/,\B/, ",", true); assertRegExpTest(/,\B$/, ",", true); assertRegExpTest(/\B,/, ",", true); assertRegExpTest(/^\B,/, ",", true); assertRegExpTest(/b\b/, "b,", true); assertRegExpTest(/b\b/, "ba", false); assertRegExpTest(/b\B/, "b,", false); assertRegExpTest(/b\B/, "ba", true); assertRegExpTest(/b\Bb/, "bb", true); assertRegExpTest(/b\bb/, "bb", false); assertRegExpTest(/b\b[,b]/, "bb", false); assertRegExpTest(/b\B[,b]/, "bb", true); assertRegExpTest(/b\b[,b]/, "b,", true); assertRegExpTest(/b\B[,b]/, "b,", false); assertRegExpTest(/[,b]\bb/, "bb", false); assertRegExpTest(/[,b]\Bb/, "bb", true); assertRegExpTest(/[,b]\bb/, ",b", true); assertRegExpTest(/[,b]\Bb/, ",b", false); assertRegExpTest(/[,b]\b[,b]/, "bb", false); assertRegExpTest(/[,b]\B[,b]/, "bb", true); assertRegExpTest(/[,b]\b[,b]/, ",b", true); assertRegExpTest(/[,b]\B[,b]/, ",b", false); assertRegExpTest(/[,b]\b[,b]/, "b,", true); assertRegExpTest(/[,b]\B[,b]/, "b,", false); // Test that caching of result doesn't share result objects. // More iterations increases the chance of hitting a GC. for (var i = 0; i < 100; i++) { var re = /x(y)z/; var res = re.exec("axyzb"); assertTrue(!!res); assertEquals(2, res.length); assertEquals("xyz", res[0]); assertEquals("y", res[1]); assertEquals(1, res.index); assertEquals("axyzb", res.input); assertEquals(undefined, res.foobar); res.foobar = "Arglebargle"; res[3] = "Glopglyf"; assertEquals("Arglebargle", res.foobar); } // Test that we perform the spec required conversions in the correct order. var log; var string = "the string"; var fakeLastIndex = { valueOf: function() { log.push("li"); return 0; } }; var fakeString = { toString: function() { log.push("ts"); return string; }, length: 0 }; var re = /str/; log = []; re.lastIndex = fakeLastIndex; var result = re.exec(fakeString); assertEquals(["str"], result); assertEquals(["ts", "li"], log); // Again, to check if caching interferes. log = []; re.lastIndex = fakeLastIndex; result = re.exec(fakeString); assertEquals(["str"], result); assertEquals(["ts", "li"], log); // And one more time, just to be certain. log = []; re.lastIndex = fakeLastIndex; result = re.exec(fakeString); assertEquals(["str"], result); assertEquals(["ts", "li"], log); // Now with a global regexp, where lastIndex is actually used. re = /str/g; log = []; re.lastIndex = fakeLastIndex; var result = re.exec(fakeString); assertEquals(["str"], result); assertEquals(["ts", "li"], log); // Again, to check if caching interferes. log = []; re.lastIndex = fakeLastIndex; result = re.exec(fakeString); assertEquals(["str"], result); assertEquals(["ts", "li"], log); // And one more time, just to be certain. log = []; re.lastIndex = fakeLastIndex; result = re.exec(fakeString); assertEquals(["str"], result); assertEquals(["ts", "li"], log); // Check that properties of RegExp have the correct permissions. var re = /x/g; var desc = Object.getOwnPropertyDescriptor(re, "global"); assertEquals(true, desc.value); assertEquals(false, desc.configurable); assertEquals(false, desc.enumerable); assertEquals(false, desc.writable); desc = Object.getOwnPropertyDescriptor(re, "multiline"); assertEquals(false, desc.value); assertEquals(false, desc.configurable); assertEquals(false, desc.enumerable); assertEquals(false, desc.writable); desc = Object.getOwnPropertyDescriptor(re, "ignoreCase"); assertEquals(false, desc.value); assertEquals(false, desc.configurable); assertEquals(false, desc.enumerable); assertEquals(false, desc.writable); desc = Object.getOwnPropertyDescriptor(re, "lastIndex"); assertEquals(0, desc.value); assertEquals(false, desc.configurable); assertEquals(false, desc.enumerable); assertEquals(true, desc.writable); // Check that end-anchored regexps are optimized correctly. var re = /(?:a|bc)g$/; assertTrue(re.test("ag")); assertTrue(re.test("bcg")); assertTrue(re.test("abcg")); assertTrue(re.test("zimbag")); assertTrue(re.test("zimbcg")); assertFalse(re.test("g")); assertFalse(re.test("")); // Global regexp (non-zero start). var re = /(?:a|bc)g$/g; assertTrue(re.test("ag")); re.lastIndex = 1; // Near start of string. assertTrue(re.test("zimbag")); re.lastIndex = 5; // Near end of string. assertFalse(re.test("zimbag")); re.lastIndex = 4; assertTrue(re.test("zimbag")); // Anchored at both ends. var re = /^(?:a|bc)g$/g; assertTrue(re.test("ag")); re.lastIndex = 1; assertFalse(re.test("ag")); re.lastIndex = 1; assertFalse(re.test("zag")); // Long max_length of RegExp. var re = /VeryLongRegExp!{1,1000}$/; assertTrue(re.test("BahoolaVeryLongRegExp!!!!!!")); assertFalse(re.test("VeryLongRegExp")); assertFalse(re.test("!")); // End anchor inside disjunction. var re = /(?:a$|bc$)/; assertTrue(re.test("a")); assertTrue(re.test("bc")); assertTrue(re.test("abc")); assertTrue(re.test("zimzamzumba")); assertTrue(re.test("zimzamzumbc")); assertFalse(re.test("c")); assertFalse(re.test("")); // Only partially anchored. var re = /(?:a|bc$)/; assertTrue(re.test("a")); assertTrue(re.test("bc")); assertEquals(["a"], re.exec("abc")); assertEquals(4, re.exec("zimzamzumba").index); assertEquals(["bc"], re.exec("zimzomzumbc")); assertFalse(re.test("c")); assertFalse(re.test(""));