plexpy/lib/html5lib/tests/testdata/tokenizer/entities.test

{"tests": [

{"description": "Undefined named entity in a double-quoted attribute value ending in semicolon and whose name starts with a known entity name.",
"input":"<h a=\"&noti;\">",
"output": [["StartTag", "h", {"a": "&noti;"}]]},

{"description": "Entity name requiring semicolon instead followed by the equals sign in a double-quoted attribute value.",
"input":"<h a=\"&lang=\">",
"output": [["StartTag", "h", {"a": "&lang="}]]},

{"description": "Valid entity name followed by the equals sign in a double-quoted attribute value.",
"input":"<h a=\"&not=\">",
"output": [["StartTag", "h", {"a": "&not="}]]},

{"description": "Undefined named entity in a single-quoted attribute value ending in semicolon and whose name starts with a known entity name.",
"input":"<h a='&noti;'>",
"output": [["StartTag", "h", {"a": "&noti;"}]]},

{"description": "Entity name requiring semicolon instead followed by the equals sign in a single-quoted attribute value.",
"input":"<h a='&lang='>",
"output": [["StartTag", "h", {"a": "&lang="}]]},

{"description": "Valid entity name followed by the equals sign in a single-quoted attribute value.",
"input":"<h a='&not='>",
"output": [["StartTag", "h", {"a": "&not="}]]},

{"description": "Undefined named entity in an unquoted attribute value ending in semicolon and whose name starts with a known entity name.",
"input":"<h a=&noti;>",
"output": [["StartTag", "h", {"a": "&noti;"}]]},

{"description": "Entity name requiring semicolon instead followed by the equals sign in an unquoted attribute value.",
"input":"<h a=&lang=>",
"output": [["StartTag", "h", {"a": "&lang="}]],
"errors":[
    { "code": "unexpected-character-in-unquoted-attribute-value", "line": 1, "col": 11 }
]},

{"description": "Valid entity name followed by the equals sign in an unquoted attribute value.",
"input":"<h a=&not=>",
"output": [["StartTag", "h", {"a": "&not="}]],
"errors":[
    { "code": "unexpected-character-in-unquoted-attribute-value", "line": 1, "col": 10 }
]},

{"description": "Ambiguous ampersand.",
"input":"&rrrraannddom;",
"output": [["Character", "&rrrraannddom;"]],
"errors":[
    { "code": "unknown-named-character-reference", "line": 1, "col": 14 }
]},

{"description": "Semicolonless named entity 'not' followed by 'i;' in body",
"input":"&noti;",
"output": [["Character", "\u00ACi;"]],
"errors":[
    { "code": "missing-semicolon-after-character-reference", "line": 1, "col": 5 }
]},

{"description": "Very long undefined named entity in body",
"input":"&ammmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmp;",
"output": [["Character", "&ammmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmp;"]],
"errors":[
    { "code": "unknown-named-character-reference", "line": 1, "col": 950 }
]},

{"description": "CR as numeric entity",
"input":"&#013;",
"output": [["Character", "\r"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 7 }
]},

{"description": "CR as hexadecimal numeric entity",
"input":"&#x00D;",
"output": [["Character", "\r"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 EURO SIGN numeric entity.",
"input":"&#0128;",
"output": [["Character", "\u20AC"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 REPLACEMENT CHAR numeric entity.",
"input":"&#0129;",
"output": [["Character", "\u0081"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 SINGLE LOW-9 QUOTATION MARK numeric entity.",
"input":"&#0130;",
"output": [["Character", "\u201A"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LATIN SMALL LETTER F WITH HOOK numeric entity.",
"input":"&#0131;",
"output": [["Character", "\u0192"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 DOUBLE LOW-9 QUOTATION MARK numeric entity.",
"input":"&#0132;",
"output": [["Character", "\u201E"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 HORIZONTAL ELLIPSIS numeric entity.",
"input":"&#0133;",
"output": [["Character", "\u2026"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 DAGGER numeric entity.",
"input":"&#0134;",
"output": [["Character", "\u2020"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 DOUBLE DAGGER numeric entity.",
"input":"&#0135;",
"output": [["Character", "\u2021"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 MODIFIER LETTER CIRCUMFLEX ACCENT numeric entity.",
"input":"&#0136;",
"output": [["Character", "\u02C6"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 PER MILLE SIGN numeric entity.",
"input":"&#0137;",
"output": [["Character", "\u2030"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LATIN CAPITAL LETTER S WITH CARON numeric entity.",
"input":"&#0138;",
"output": [["Character", "\u0160"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 SINGLE LEFT-POINTING ANGLE QUOTATION MARK numeric entity.",
"input":"&#0139;",
"output": [["Character", "\u2039"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LATIN CAPITAL LIGATURE OE numeric entity.",
"input":"&#0140;",
"output": [["Character", "\u0152"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 REPLACEMENT CHAR numeric entity.",
"input":"&#0141;",
"output": [["Character", "\u008D"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LATIN CAPITAL LETTER Z WITH CARON numeric entity.",
"input":"&#0142;",
"output": [["Character", "\u017D"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 REPLACEMENT CHAR numeric entity.",
"input":"&#0143;",
"output": [["Character", "\u008F"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 REPLACEMENT CHAR numeric entity.",
"input":"&#0144;",
"output": [["Character", "\u0090"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LEFT SINGLE QUOTATION MARK numeric entity.",
"input":"&#0145;",
"output": [["Character", "\u2018"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 RIGHT SINGLE QUOTATION MARK numeric entity.",
"input":"&#0146;",
"output": [["Character", "\u2019"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LEFT DOUBLE QUOTATION MARK numeric entity.",
"input":"&#0147;",
"output": [["Character", "\u201C"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 RIGHT DOUBLE QUOTATION MARK numeric entity.",
"input":"&#0148;",
"output": [["Character", "\u201D"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 BULLET numeric entity.",
"input":"&#0149;",
"output": [["Character", "\u2022"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 EN DASH numeric entity.",
"input":"&#0150;",
"output": [["Character", "\u2013"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 EM DASH numeric entity.",
"input":"&#0151;",
"output": [["Character", "\u2014"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 SMALL TILDE numeric entity.",
"input":"&#0152;",
"output": [["Character", "\u02DC"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 TRADE MARK SIGN numeric entity.",
"input":"&#0153;",
"output": [["Character", "\u2122"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LATIN SMALL LETTER S WITH CARON numeric entity.",
"input":"&#0154;",
"output": [["Character", "\u0161"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 SINGLE RIGHT-POINTING ANGLE QUOTATION MARK numeric entity.",
"input":"&#0155;",
"output": [["Character", "\u203A"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LATIN SMALL LIGATURE OE numeric entity.",
"input":"&#0156;",
"output": [["Character", "\u0153"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 REPLACEMENT CHAR numeric entity.",
"input":"&#0157;",
"output": [["Character", "\u009D"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 EURO SIGN hexadecimal numeric entity.",
"input":"&#x080;",
"output": [["Character", "\u20AC"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 REPLACEMENT CHAR hexadecimal numeric entity.",
"input":"&#x081;",
"output": [["Character", "\u0081"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 SINGLE LOW-9 QUOTATION MARK hexadecimal numeric entity.",
"input":"&#x082;",
"output": [["Character", "\u201A"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LATIN SMALL LETTER F WITH HOOK hexadecimal numeric entity.",
"input":"&#x083;",
"output": [["Character", "\u0192"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 DOUBLE LOW-9 QUOTATION MARK hexadecimal numeric entity.",
"input":"&#x084;",
"output": [["Character", "\u201E"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 HORIZONTAL ELLIPSIS hexadecimal numeric entity.",
"input":"&#x085;",
"output": [["Character", "\u2026"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 DAGGER hexadecimal numeric entity.",
"input":"&#x086;",
"output": [["Character", "\u2020"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 DOUBLE DAGGER hexadecimal numeric entity.",
"input":"&#x087;",
"output": [["Character", "\u2021"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 MODIFIER LETTER CIRCUMFLEX ACCENT hexadecimal numeric entity.",
"input":"&#x088;",
"output": [["Character", "\u02C6"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 PER MILLE SIGN hexadecimal numeric entity.",
"input":"&#x089;",
"output": [["Character", "\u2030"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LATIN CAPITAL LETTER S WITH CARON hexadecimal numeric entity.",
"input":"&#x08A;",
"output": [["Character", "\u0160"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 SINGLE LEFT-POINTING ANGLE QUOTATION MARK hexadecimal numeric entity.",
"input":"&#x08B;",
"output": [["Character", "\u2039"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LATIN CAPITAL LIGATURE OE hexadecimal numeric entity.",
"input":"&#x08C;",
"output": [["Character", "\u0152"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 REPLACEMENT CHAR hexadecimal numeric entity.",
"input":"&#x08D;",
"output": [["Character", "\u008D"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LATIN CAPITAL LETTER Z WITH CARON hexadecimal numeric entity.",
"input":"&#x08E;",
"output": [["Character", "\u017D"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 REPLACEMENT CHAR hexadecimal numeric entity.",
"input":"&#x08F;",
"output": [["Character", "\u008F"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 REPLACEMENT CHAR hexadecimal numeric entity.",
"input":"&#x090;",
"output": [["Character", "\u0090"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LEFT SINGLE QUOTATION MARK hexadecimal numeric entity.",
"input":"&#x091;",
"output": [["Character", "\u2018"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 RIGHT SINGLE QUOTATION MARK hexadecimal numeric entity.",
"input":"&#x092;",
"output": [["Character", "\u2019"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LEFT DOUBLE QUOTATION MARK hexadecimal numeric entity.",
"input":"&#x093;",
"output": [["Character", "\u201C"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 RIGHT DOUBLE QUOTATION MARK hexadecimal numeric entity.",
"input":"&#x094;",
"output": [["Character", "\u201D"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 BULLET hexadecimal numeric entity.",
"input":"&#x095;",
"output": [["Character", "\u2022"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 EN DASH hexadecimal numeric entity.",
"input":"&#x096;",
"output": [["Character", "\u2013"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 EM DASH hexadecimal numeric entity.",
"input":"&#x097;",
"output": [["Character", "\u2014"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 SMALL TILDE hexadecimal numeric entity.",
"input":"&#x098;",
"output": [["Character", "\u02DC"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 TRADE MARK SIGN hexadecimal numeric entity.",
"input":"&#x099;",
"output": [["Character", "\u2122"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LATIN SMALL LETTER S WITH CARON hexadecimal numeric entity.",
"input":"&#x09A;",
"output": [["Character", "\u0161"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 SINGLE RIGHT-POINTING ANGLE QUOTATION MARK hexadecimal numeric entity.",
"input":"&#x09B;",
"output": [["Character", "\u203A"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LATIN SMALL LIGATURE OE hexadecimal numeric entity.",
"input":"&#x09C;",
"output": [["Character", "\u0153"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 REPLACEMENT CHAR hexadecimal numeric entity.",
"input":"&#x09D;",
"output": [["Character", "\u009D"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LATIN SMALL LETTER Z WITH CARON hexadecimal numeric entity.",
"input":"&#x09E;",
"output": [["Character", "\u017E"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Windows-1252 LATIN CAPITAL LETTER Y WITH DIAERESIS hexadecimal numeric entity.",
"input":"&#x09F;",
"output": [["Character", "\u0178"]],
"errors":[
    { "code": "control-character-reference", "line": 1, "col": 8 }
]},

{"description": "Decimal numeric entity followed by hex character a.",
"input":"&#97a",
"output": [["Character", "aa"]],
"errors":[
    { "code": "missing-semicolon-after-character-reference", "line": 1, "col": 5 }
]},

{"description": "Decimal numeric entity followed by hex character A.",
"input":"&#97A",
"output": [["Character", "aA"]],
"errors":[
    { "code": "missing-semicolon-after-character-reference", "line": 1, "col": 5 }
]},

{"description": "Decimal numeric entity followed by hex character f.",
"input":"&#97f",
"output": [["Character", "af"]],
"errors":[
    { "code": "missing-semicolon-after-character-reference", "line": 1, "col": 5 }
]},

{"description": "Decimal numeric entity followed by hex character A.",
"input":"&#97F",
"output": [["Character", "aF"]],
"errors":[
    { "code": "missing-semicolon-after-character-reference", "line": 1, "col": 5 }
]}

]}