OXYGENHUB
◆ main openmsx / src / utils / rapidsax.hh
24.2 KiB · C++ · last changed chore(): initial commit 1 month ago source blame history raw
1 
#ifndef RAPIDSAX_HH
2 
#define RAPIDSAX_HH
3
4 
// This code is _heavily_ based on RapidXml 1.13
5 
//   http://rapidxml.sourceforge.net/
6 
//
7 
// RapidXml is a very fast XML parser.
8 
//   http://xmlbench.sourceforge.net/results/benchmark200910/index.html
9 
// One of the main reasons it can be this fast is that doesn't do any string
10 
// copies. Instead the XML input data is modified in-place (e.g. for stuff like
11 
// < replacements). Though this also means the output produced by the parser
12 
// is tied to the lifetime of the XML input data.
13 
//
14 
// RapidXml produces a DOM-like output. This parser has a SAX-like interface.
15
16 
#include "one_of.hh"
17 
#include "small_compare.hh"
18 
#include "zstring_view.hh"
19
20 
#include <array>
21 
#include <cassert>
22 
#include <cstdint>
23 
#include <string_view>
24
25 
namespace rapidsax {
26
27 
// Parse given XML text and call callback functions in the given handler.
28 
// - XML text must be zero-terminated
29 
// - Handler must implement the methods defined in NullHandler (below). An
30 
//   easy way to do this is to inherit from NullHandler and only reimplement
31 
//   the methods that you need.
32 
// - The behavior of the parser can be fine-tuned with the FLAGS parameter,
33 
//   see below for more details.
34 
// - When a parse error is encounter, an instance of ParseError is thrown.
35 
// - The lifetime of the string_view's in the callback handler is the same as
36 
//   the lifetime of the input XML data (no string copies are made, instead
37 
//   the XML file is modified in-place and references to this data are passed).
38 
template<int FLAGS, typename HANDLER> void parse(HANDLER& handler, char* xml);
39
40 
// When loading an XML file from disk, the buffer needs to be 8 bytes bigger
41 
// than the filesize. The first of these bytes must be filled with zero
42 
// (zero-terminate the xml data). The other bytes are only there to allow to
43 
// read up-to 8 bytes past the end without triggering memory protection errors.
44 
inline constexpr size_t EXTRA_BUFFER_SPACE = 8;
45
46
47 
// Flags that influence parsing behavior. The flags can be OR'ed together.
48
49 
// Should XML entities like &lt; be expanded or not?
50 
inline constexpr int noEntityTranslation = 0x1;
51 
// Should leading and trailing whitespace be trimmed?
52 
inline constexpr int trimWhitespace      = 0x2;
53 
// Should sequences of whitespace characters be replaced with a single
54 
// space character?
55 
inline constexpr int normalizeWhitespace = 0x4;
56 
// Should strings be modified (in-place) with a zero-terminator?
57 
inline constexpr int zeroTerminateStrings = 0x8;
58
59
60 
// Callback handler with all empty implementations (can be used as a base
61 
// class in case you only need to reimplement a few of the methods).
62 
//
63 
// Several methods are overloaded with either 'std::string_view' or
64 
// 'zstring_view' parameter types. Which one gets called depends on whether the
65 
// 'zeroTerminateStrings' flag was passed.
66 
class NullHandler
67 
{
68 
public:
69 
	// Called when an opening XML tag is encountered.
70 
	// 'name' is the name of the XML tag.
71 
	void start(std::string_view /*name*/) {}
72 
	void start(zstring_view /*name*/) {}
73
74 
	// Called when a XML tag is closed.
75 
	// Note: the parser does currently not check whether the name of the
76 
	//       opening nd closing tags matches.
77 
	void stop() {}
78
79 
	// Called when text inside a tag is parsed.
80 
	// XML entities are replaced (optional)
81 
	// Whitespace is (optionally) trimmed or normalized.
82 
	// This method is not called for an empty text string.
83 
	// (Unlike other SAX parsers) the whole text string is always
84 
	// passed in a single chunk (so no need to concatenate this text
85 
	// with previous chunks in the callback).
86 
	void text(std::string_view /*text*/) {}
87 
	void text(zstring_view /*text*/) {}
88
89 
	// Called for each parsed attribute.
90 
	// Attributes can occur inside xml tags or inside XML declarations.
91 
	void attribute(std::string_view /*name*/, std::string_view /*value*/) {}
92 
	void attribute(zstring_view /*name*/, zstring_view /*value*/) {}
93
94 
	// Called for parsed CDATA sections.
95 
	void cdata(std::string_view /*value*/) {}
96 
	void cdata(zstring_view /*value*/) {}
97
98 
	// Called when a XML comment (<!-- ... -->) is parsed.
99 
	void comment(std::string_view /*value*/) {}
100 
	void comment(zstring_view /*value*/) {}
101
102 
	// Called when XML declaration (<?xml .. ?>) is parsed.
103 
	// Inside a XML declaration there can be attributes.
104 
	void declarationStart() {}
105 
	void declAttribute(std::string_view /*name*/, std::string_view /*value*/) {}
106 
	void declAttribute(zstring_view /*name*/, zstring_view /*value*/) {}
107 
	void declarationStop() {}
108
109 
	// Called when the <!DOCTYPE ..> is parsed.
110 
	void doctype(std::string_view /*text*/) {}
111 
	void doctype(zstring_view /*text*/) {}
112
113 
	// Called when XML processing instructions (<? .. ?>) are parsed.
114 
	void procInstr(std::string_view /*target*/, std::string_view /*instr*/) {}
115 
	void procInstr(zstring_view /*target*/, zstring_view /*instr*/) {}
116 
};
117
118
119 
class ParseError
120 
{
121 
public:
122 
	ParseError(const char* what_, char* where_)
123 
		: m_what(what_)
124 
		, m_where(where_)
125 
	{
126 
	}
127
128 
	[[nodiscard]] const char* what() const { return m_what; }
129 
	[[nodiscard]] char* where() const { return m_where; }
130
131 
private:
132 
	const char* m_what;
133 
	char* m_where;
134 
};
135
136
137 
namespace internal {
138
139 
// Character class lookup table
140 
//   bit 0: \0
141 
//   bit 1: \t \r \r space
142 
//   bit 2: <
143 
//   bit 3: &
144 
//   bit 4: '
145 
//   bit 5: "
146 
//   bit 6: / > ?
147 
//   bit 7: ! =
148 
inline constexpr std::array<const uint8_t, 256> lutChar =
149 
{
150 
	// 0   1    2    3    4    5    6    7    8    9    A    B    C    D    E    F
151 
	0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x02,0x02,0x00,0x00,0x02,0x00,0x00, // 0
152 
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, // 1
153 
	0x02,0x80,0x20,0x00,0x00,0x00,0x08,0x10,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x40, // 2
154 
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x04,0x80,0x40,0x40, // 3
155 
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, // 4
156 
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, // 5
157 
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, // 6
158 
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, // 7
159 
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, // 8
160 
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, // 9
161 
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, // A
162 
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, // B
163 
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, // C
164 
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, // D
165 
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, // E
166 
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, // F
167 
};
168
169 
// Digits (dec and hex, 255 denotes end of numeric character reference)
170 
inline constexpr std::array<const uint8_t, 256> lutDigits =
171 
{
172 
	// 0   1   2   3   4   5   6   7   8   9   A   B   C   D   E   F
173 
	 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,  // 0
174 
	 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,  // 1
175 
	 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,  // 2
176 
	   0,  1,  2,  3,  4,  5,  6,  7,  8,  9,255,255,255,255,255,255,  // 3
177 
	 255, 10, 11, 12, 13, 14, 15,255,255,255,255,255,255,255,255,255,  // 4
178 
	 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,  // 5
179 
	 255, 10, 11, 12, 13, 14, 15,255,255,255,255,255,255,255,255,255,  // 6
180 
	 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,  // 7
181 
	 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,  // 8
182 
	 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,  // 9
183 
	 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,  // A
184 
	 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,  // B
185 
	 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,  // C
186 
	 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,  // D
187 
	 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,  // E
188 
	 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255   // F
189 
};
190
191 
// Detect whitespace character (space \n \r \t)
192 
struct WhitespacePred {
193 
	[[nodiscard]] static bool test(char ch) { return (lutChar[uint8_t(ch)] & 0x02) != 0; }
194 
};
195
196 
// Detect node name character (anything but space \n \r \t / > ? \0)
197 
struct NodeNamePred {
198 
	[[nodiscard]] static bool test(char ch) { return !(lutChar[uint8_t(ch)] & 0x43); }
199 
};
200
201 
// Detect attribute name character (anything but space \n \r \t / < > = ? ! \0)
202 
struct AttributeNamePred {
203 
	[[nodiscard]] static bool test(char ch) { return !(lutChar[uint8_t(ch)] & 0xC7); }
204 
};
205
206 
// Detect text character (PCDATA) (anything but < \0)
207 
struct TextPred {
208 
	[[nodiscard]] static bool test(char ch) { return !(lutChar[uint8_t(ch)] & 0x05); }
209 
};
210
211 
// Detect text character (PCDATA) that does not require processing when ws
212 
// normalization is disabled (anything but < \0 &)
213 
struct TextPureNoWsPred {
214 
	[[nodiscard]] static bool test(char ch) { return !(lutChar[uint8_t(ch)] & 0x0D); }
215 
};
216
217 
// Detect text character (PCDATA) that does not require processing when ws
218 
// normalization is enabled (anything but < \0 & space \n \r \t)
219 
struct TextPureWithWsPred {
220 
	[[nodiscard]] static bool test(char ch) { return !(lutChar[uint8_t(ch)] & 0x0F); }
221 
};
222
223 
// Detect attribute value character, single quote (anything but ' \0)
224 
struct AttPred1 {
225 
	[[nodiscard]] static bool test(char ch) { return !(lutChar[uint8_t(ch)] & 0x11); }
226 
};
227 
// Detect attribute value character, double quote (anything but " \0)
228 
struct AttPred2 {
229 
	[[nodiscard]] static bool test(char ch) { return !(lutChar[uint8_t(ch)] & 0x21); }
230 
};
231
232 
// Detect attribute value character, single quote, that does not require
233 
// processing (anything but ' \0 &)
234 
struct AttPurePred1 {
235 
	[[nodiscard]] static bool test(char ch) { return !(lutChar[uint8_t(ch)] & 0x19); }
236 
};
237 
// Detect attribute value character, double quote, that does not require
238 
// processing (anything but " \0 &)
239 
struct AttPurePred2 {
240 
	[[nodiscard]] static bool test(char ch) { return !(lutChar[uint8_t(ch)] & 0x29); }
241 
};
242
243 
// Insert coded character, using UTF8
244 
inline void insertUTF8char(char*& text, uint32_t code)
245 
{
246 
	if (code < 0x80) { // 1 byte sequence
247 
		text[0] = char(code);
248 
		text += 1;
249 
	} else if (code < 0x800) {// 2 byte sequence
250 
		text[1] = char((code | 0x80) & 0xBF); code >>= 6;
251 
		text[0] = char (code | 0xC0);
252 
		text += 2;
253 
	} else if (code < 0x10000) { // 3 byte sequence
254 
		text[2] = char((code | 0x80) & 0xBF); code >>= 6;
255 
		text[1] = char((code | 0x80) & 0xBF); code >>= 6;
256 
		text[0] = char (code | 0xE0);
257 
		text += 3;
258 
	} else if (code < 0x110000) { // 4 byte sequence
259 
		text[3] = char((code | 0x80) & 0xBF); code >>= 6;
260 
		text[2] = char((code | 0x80) & 0xBF); code >>= 6;
261 
		text[1] = char((code | 0x80) & 0xBF); code >>= 6;
262 
		text[0] = char (code | 0xF0);
263 
		text += 4;
264 
	} else { // Invalid, only codes up to 0x10FFFF are allowed in Unicode
265 
		throw ParseError("invalid numeric character entity", text);
266 
	}
267 
}
268
269 
template<StringLiteral Str> [[nodiscard]] static inline bool next(const char* p)
270 
{
271 
	return small_compare<Str>(p);
272 
}
273
274
275 
// Skip characters until predicate evaluates to true
276 
template<typename StopPred> static inline void skip(char*& text)
277 
{
278 
	char* tmp = text;
279 
	while (StopPred::test(*tmp)) ++tmp;
280 
	text = tmp;
281 
}
282
283 
// Skip characters until predicate evaluates to true while doing the following:
284 
// - replacing XML character entity references with proper characters
285 
//   (&apos; &amp; &quot; &lt; &gt; &#...;)
286 
// - condensing whitespace sequences to single space character
287 
template<typename StopPred, class StopPredPure, int FLAGS>
288 
[[nodiscard]] static inline char* skipAndExpand(char*& text)
289 
{
290 
	// If entity translation, whitespace condense and whitespace
291 
	// trimming is disabled, use plain skip.
292 
	if constexpr ( (FLAGS & noEntityTranslation) &&
293 
	              !(FLAGS & normalizeWhitespace)  &&
294 
	              !(FLAGS & trimWhitespace)) {
295 
		skip<StopPred>(text);
296 
		return text;
297 
	}
298
299 
	// Use simple skip until first modification is detected
300 
	skip<StopPredPure>(text);
301
302 
	// Use translation skip
303 
	char* src = text;
304 
	char* dest = src;
305 
	while (StopPred::test(*src)) {
306 
		// Test if replacement is needed
307 
		if (!(FLAGS & noEntityTranslation) &&
308 
		    (src[0] == '&')) {
309 
			switch (src[1]) {
310 
			case 'a': // &amp; &apos;
311 
				if (next<"amp;">(&src[1])) {
312 
					*dest = '&';
313 
					++dest;
314 
					src += 5;
315 
					continue;
316 
				}
317 
				if (next<"pos;">(&src[2])) {
318 
					*dest = '\'';
319 
					++dest;
320 
					src += 6;
321 
					continue;
322 
				}
323 
				break;
324
325 
			case 'q': // &quot;
326 
				if (next<"uot;">(&src[2])) {
327 
					*dest = '"';
328 
					++dest;
329 
					src += 6;
330 
					continue;
331 
				}
332 
				break;
333
334 
			case 'g': // &gt;
335 
				if (next<"t;">(&src[2])) {
336 
					*dest = '>';
337 
					++dest;
338 
					src += 4;
339 
					continue;
340 
				}
341 
				break;
342
343 
			case 'l': // &lt;
344 
				if (next<"t;">(&src[2])) {
345 
					*dest = '<';
346 
					++dest;
347 
					src += 4;
348 
					continue;
349 
				}
350 
				break;
351
352 
			case '#': // &#...; - assumes ASCII
353 
				if (src[2] == 'x') {
354 
					uint32_t code = 0;
355 
					src += 3; // skip &#x
356 
					while (true) {
357 
						uint8_t digit = lutDigits[uint8_t(*src)];
358 
						if (digit == 0xFF) break;
359 
						code = code * 16 + digit;
360 
						++src;
361 
					}
362 
					insertUTF8char(dest, code);
363 
				} else {
364 
					uint32_t code = 0;
365 
					src += 2; // skip &#
366 
					while (true) {
367 
						   uint8_t digit = lutDigits[uint8_t(*src)];
368 
						   if (digit == 0xFF) break;
369 
						   code = code * 10 + digit;
370 
						   ++src;
371 
					}
372 
					insertUTF8char(dest, code);
373 
				}
374 
				if (*src != ';') {
375 
					throw ParseError("expected ;", src);
376 
				}
377 
				++src;
378 
				continue;
379
380 
			default:
381 
				// Something else, ignore, just copy '&' verbatim
382 
				break;
383 
			}
384 
		}
385
386 
		// Test if condensing is needed
387 
		if ((FLAGS & normalizeWhitespace) &&
388 
		    (WhitespacePred::test(*src))) {
389 
			*dest++ = ' '; // single space in dest
390 
			++src;         // skip first whitespace char
391 
			// Skip remaining whitespace chars
392 
			while (WhitespacePred::test(*src)) ++src;
393 
			continue;
394 
		}
395
396 
		// No replacement, only copy character
397 
		*dest++ = *src++;
398 
	}
399
400 
	// Return new end
401 
	text = src;
402 
	return dest;
403 
}
404
405 
inline void skipBOM(char*& text)
406 
{
407 
	if (next<"\357\273\277">(text)) { // char(0xEF), char(0xBB), char(0xBF)
408 
		text += 3; // skip utf-8 bom
409 
	}
410 
}
411
412
413 
template<int FLAGS, typename HANDLER> class Parser
414 
{
415 
	HANDLER& handler;
416
417 
public:
418 
	Parser(HANDLER& handler_, char* text)
419 
		: handler(handler_)
420 
	{
421 
		skipBOM(text);
422 
		while (true) {
423 
			// Skip whitespace before node
424 
			skip<WhitespacePred>(text);
425 
			if (*text == 0) break;
426
427 
			if (*text != '<') {
428 
				throw ParseError("expected <", text);
429 
			}
430 
			++text; // skip '<'
431 
			parseNode(text);
432 
		}
433 
	}
434
435 
private:
436 
	// Parse XML declaration (<?xml...)
437 
	void parseDeclaration(char*& text)
438 
	{
439 
		handler.declarationStart();
440 
		skip<WhitespacePred>(text); // skip ws before attributes or ?>
441 
		parseAttributes(text, true);
442 
		handler.declarationStop();
443
444 
		// skip ?>
445 
		if (!next<"?>">(text)) {
446 
			throw ParseError("expected ?>", text);
447 
		}
448 
		text += 2;
449 
	}
450
451 
	// Parse XML comment (<!--...)
452 
	void parseComment(char*& text)
453 
	{
454 
		// Skip until end of comment
455 
		const char* value = text; // remember value start
456 
		while (!next<"-->">(text)) {
457 
			if (text[0] == 0) {
458 
				throw ParseError("unexpected end of data", text);
459 
			}
460 
			++text;
461 
		}
462 
		if constexpr (FLAGS & zeroTerminateStrings) {
463 
			*text = '\0';
464 
			handler.comment(zstring_view(value, text - value));
465 
		} else {
466 
			handler.comment(std::string_view(value, text - value));
467 
		}
468 
		text += 3; // skip '-->'
469 
	}
470
471 
	void parseDoctype(char*& text)
472 
	{
473 
		const char* value = text; // remember value start
474
475 
		// skip to >
476 
		while (*text != '>') {
477 
			switch (*text) {
478 
			case '[': {
479 
				// If '[' encountered, scan for matching ending
480 
				// ']' using naive algorithm with depth. This
481 
				// works for all W3C test files except for 2
482 
				// most wicked.
483 
				++text; // skip '['
484 
				int depth = 1;
485 
				while (depth > 0) {
486 
					switch (*text) {
487 
						case '[': ++depth; break;
488 
						case ']': --depth; break;
489 
						case 0: throw ParseError(
490 
							"unexpected end of data", text);
491 
					}
492 
					++text;
493 
				}
494 
				break;
495 
			}
496 
			case '\0':
497 
				throw ParseError("unexpected end of data", text);
498
499 
			default:
500 
				++text;
501 
			}
502 
		}
503
504 
		if constexpr (FLAGS & zeroTerminateStrings) {
505 
			*text = '\0';
506 
			handler.doctype(zstring_view(value, text - value));
507 
		} else {
508 
			handler.doctype(std::string_view(value, text - value));
509 
		}
510 
		text += 1; // skip '>'
511 
	}
512
513 
	void parsePI(char*& text)
514 
	{
515 
		// Extract PI target name
516 
		const char* name = text;
517 
		skip<NodeNamePred>(text);
518 
		char* nameEnd = text;
519 
		if (name == nameEnd) {
520 
			throw ParseError("expected PI target", text);
521 
		}
522
523 
		// Skip whitespace between pi target and pi
524 
		skip<WhitespacePred>(text);
525
526 
		// Skip to '?>'
527 
		const char* value = text; // Remember start of pi
528 
		while (!next<"?>">(text)) {
529 
			if (*text == 0) {
530 
				throw ParseError("unexpected end of data", text);
531 
			}
532 
			++text;
533 
		}
534 
		// Set pi value (verbatim, no entity expansion or ws normalization)
535 
		if constexpr (FLAGS & zeroTerminateStrings) {
536 
			*nameEnd = '\0';
537 
			*text = '\0';
538 
			handler.procInstr(zstring_view(name,  nameEnd - name),
539 
			                  zstring_view(value, text - value));
540 
		} else {
541 
			handler.procInstr(std::string_view(name,  nameEnd - name),
542 
			                  std::string_view(value, text - value));
543 
		}
544 
		text += 2; // skip '?>'
545 
	}
546
547 
	void parseText(char*& text, char* contentsStart)
548 
	{
549 
		// Backup to contents start if whitespace trimming is disabled
550 
		if constexpr (!(FLAGS & trimWhitespace)) {
551 
			text = contentsStart;
552 
		}
553 
		// Skip until end of data
554 
		const char* value = text;
555 
		char* end = (FLAGS & normalizeWhitespace)
556 
			? skipAndExpand<TextPred, TextPureWithWsPred, FLAGS>(text)
557 
			: skipAndExpand<TextPred, TextPureNoWsPred  , FLAGS>(text);
558
559 
		// Trim trailing whitespace; leading was already trimmed by
560 
		// whitespace skip after >
561 
		if constexpr ((FLAGS & trimWhitespace) != 0) {
562 
			if constexpr (FLAGS & normalizeWhitespace) {
563 
				// Whitespace is already condensed to single
564 
				// space characters by skipping function, so
565 
				// just trim 1 char off the end.
566 
				if (end[-1] == ' ') {
567 
					--end;
568 
				}
569 
			} else {
570 
				// Backup until non-whitespace character is found
571 
				while (WhitespacePred::test(end[-1])) {
572 
					--end;
573 
				}
574 
			}
575 
		}
576
577 
		// check next char before calling handler.text()
578 
		if (*text == '\0') {
579 
			throw ParseError("unexpected end of data", text);
580 
		} else {
581 
			assert(*text == '<');
582 
		}
583
584 
		// Handle text, but only if non-empty.
585 
		auto len = end - value;
586 
		if (len) {
587 
			if constexpr (FLAGS & zeroTerminateStrings) {
588 
				*end = '\0';
589 
				handler.text(zstring_view(value, len));
590 
			} else {
591 
				handler.text(std::string_view(value, len));
592 
			}
593 
		}
594 
	}
595
596 
	void parseCdata(char*& text)
597 
	{
598 
		// Skip until end of cdata
599 
		const char* value = text;
600 
		while (!next<"]]>">(text)) {
601 
			if (text[0] == 0) {
602 
				throw ParseError("unexpected end of data", text);
603 
			}
604 
			++text;
605 
		}
606 
		if constexpr (FLAGS & zeroTerminateStrings) {
607 
			*text = '\0';
608 
			handler.cdata(zstring_view(value, text - value));
609 
		} else {
610 
			handler.cdata(std::string_view(value, text - value));
611 
		}
612 
		text += 3; // skip ]]>
613 
	}
614
615 
	void parseElement(char*& text)
616 
	{
617 
		// Extract element name
618 
		const char* name = text;
619 
		skip<NodeNamePred>(text);
620 
		char* nameEnd = text;
621 
		if (name == nameEnd) {
622 
			throw ParseError("expected element name", text);
623 
		}
624 
		char savedChar = *nameEnd;
625 
		skip<WhitespacePred>(text); // skip ws before attributes or >
626 
		if constexpr (FLAGS & zeroTerminateStrings) {
627 
			*nameEnd = '\0';
628 
			handler.start(zstring_view(name, nameEnd - name));
629 
		} else {
630 
			handler.start(std::string_view(name, nameEnd - name));
631 
		}
632
633 
		parseAttributes(text, false);
634
635 
		// Determine ending type
636 
		char endChar = ((FLAGS & zeroTerminateStrings) && (text == nameEnd))
637 
		             ? savedChar : *text;
638 
		if (endChar == '>') {
639 
			++text;
640 
			parseNodeContents(text);
641 
		} else if (endChar == '/') {
642 
			handler.stop();
643 
			++text;
644 
			if (*text != '>') {
645 
				throw ParseError("expected >", text);
646 
			}
647 
			++text;
648 
		} else {
649 
			throw ParseError("expected >", text);
650 
		}
651 
	}
652
653 
	// Determine node type, and parse it
654 
	void parseNode(char*& text)
655 
	{
656 
		switch (text[0]) {
657 
		case '?': // <?...
658 
			++text; // skip ?
659 
			// Note: this doesn't detect mixed case (xMl), does
660 
			// that matter?
661 
			if ((next<"xml">(text) || next<"XML">(text)) &&
662 
			    WhitespacePred::test(text[3])) {
663 
				// '<?xml ' - xml declaration
664 
				text += 4; // skip 'xml '
665 
				parseDeclaration(text);
666 
			} else {
667 
				parsePI(text);
668 
			}
669 
			break;
670
671 
		case '!': // <!...
672 
			// Parse proper subset of <! node
673 
			switch (text[1]) {
674 
			case '-': // <!-
675 
				if (text[2] == '-') {
676 
					// '<!--' - xml comment
677 
					text += 3; // skip '!--'
678 
					parseComment(text);
679 
					return;
680 
				}
681 
				break;
682
683 
			case '[': // <![
684 
				if (next<"CDATA[">(&text[2])) {
685 
					// '<![CDATA[' - cdata
686 
					text += 8; // skip '![CDATA['
687 
					parseCdata(text);
688 
					return;
689 
				}
690 
				break;
691
692 
			case 'D': // <!D
693 
				if (next<"OCTYPE">(&text[2]) &&
694 
				    WhitespacePred::test(text[8])) {
695 
					// '<!DOCTYPE ' - doctype
696 
					text += 9; // skip '!DOCTYPE '
697 
					parseDoctype(text);
698 
					return;
699 
				}
700 
				break;
701 
			}
702 
			// Attempt to skip other, unrecognized types starting with <!
703 
			++text; // skip !
704 
			while (*text != '>') {
705 
				if (*text == 0) {
706 
					throw ParseError(
707 
						"unexpected end of data", text);
708 
				}
709 
				++text;
710 
			}
711 
			++text; // skip '>'
712 
			break;
713
714 
		default: // <...
715 
			parseElement(text);
716 
			break;
717 
		}
718 
	}
719
720 
	// Parse contents of the node - children, data etc.
721 
	void parseNodeContents(char*& text)
722 
	{
723 
		while (true) {
724 
			char* contentsStart = text; // start before ws is skipped
725 
			skip<WhitespacePred>(text); // Skip ws between > and contents
726
727 
			switch (*text) {
728 
			case '<': // Node closing or child node
729 
afterText:		// After parseText() jump here instead of continuing
730 
			// the loop, because skipping whitespace is unnecessary.
731 
				if (text[1] == '/') {
732 
					// Node closing
733 
					text += 2; // skip '</'
734 
					skip<NodeNamePred>(text);
735 
					// TODO validate closing tag??
736 
					handler.stop();
737 
					// Skip remaining whitespace after node name
738 
					skip<WhitespacePred>(text);
739 
					if (*text != '>') {
740 
						throw ParseError("expected >", text);
741 
					}
742 
					++text; // skip '>'
743 
					return;
744 
				} else {
745 
					// Child node
746 
					++text; // skip '<'
747 
					parseNode(text);
748 
				}
749 
				break;
750
751 
			case '\0':
752 
				throw ParseError("unexpected end of data", text);
753
754 
			default:
755 
				parseText(text, contentsStart);
756 
				goto afterText;
757 
			}
758 
		}
759 
	}
760
761 
	// Parse XML attributes of the node
762 
	void parseAttributes(char*& text, bool declaration)
763 
	{
764 
		// For all attributes
765 
		while (AttributeNamePred::test(*text)) {
766 
			// Extract attribute name
767 
			char* name = text;
768 
			++text; // Skip first character of attribute name
769 
			skip<AttributeNamePred>(text);
770 
			char* nameEnd = text;
771 
			if (name == nameEnd) {
772 
				throw ParseError("expected attribute name", name);
773 
			}
774
775 
			skip<WhitespacePred>(text); // skip ws after name
776 
			if (*text != '=') {
777 
				throw ParseError("expected =", text);
778 
			}
779 
			++text; // skip =
780 
			skip<WhitespacePred>(text); // skip ws after =
781
782 
			// Skip quote and remember if it was ' or "
783 
			char quote = *text;
784 
			if (quote != one_of('\'', '"')) {
785 
				throw ParseError("expected ' or \"", text);
786 
			}
787 
			++text;
788
789 
			// Extract attribute value and expand char refs in it
790 
			// No whitespace normalization in attributes
791 
			constexpr int FLAGS2 = FLAGS & ~normalizeWhitespace;
792 
			const char* value = text;
793 
			char* valueEnd = (quote == '\'')
794 
				? skipAndExpand<AttPred1, AttPurePred1, FLAGS2>(text)
795 
				: skipAndExpand<AttPred2, AttPurePred2, FLAGS2>(text);
796 
			// Make sure that end quote is present
797 
			// check before calling handler.xxx()
798 
			if (*text != quote) {
799 
				throw ParseError("expected ' or \"", text);
800 
			}
801 
			++text; // skip quote
802
803 
			if constexpr (FLAGS & zeroTerminateStrings) {
804 
				*nameEnd = '\0';
805 
				*valueEnd = '\0';
806 
				if (!declaration) {
807 
					handler.attribute(zstring_view(name, nameEnd - name),
808 
					                  zstring_view(value, valueEnd - value));
809 
				} else {
810 
					handler.declAttribute(zstring_view(name, nameEnd - name),
811 
					                      zstring_view(value, valueEnd - value));
812 
				}
813 
			} else {
814 
				if (!declaration) {
815 
					handler.attribute(std::string_view(name, nameEnd - name),
816 
					                  std::string_view(value, valueEnd - value));
817 
				} else {
818 
					handler.declAttribute(std::string_view(name, nameEnd - name),
819 
					                      std::string_view(value, valueEnd - value));
820 
				}
821 
			}
822
823 
			skip<WhitespacePred>(text); // skip ws after value
824 
		}
825 
	}
826 
};
827
828 
} // namespace internal
829
830 
template<int FLAGS, typename HANDLER>
831 
inline void parse(HANDLER& handler, char* xml)
832 
{
833 
	internal::Parser<FLAGS, HANDLER> parser(handler, xml);
834 
}
835
836 
} // namespace rapidsax
837
838 
#endif