1 // Copyright 2007-2009 Wincent Colaiuta. All rights reserved.
3 // Redistribution and use in source and binary forms, with or without
4 // modification, are permitted provided that the following conditions are met:
6 // 1. Redistributions of source code must retain the above copyright notice,
7 // this list of conditions and the following disclaimer.
8 // 2. Redistributions in binary form must reproduce the above copyright notice,
9 // this list of conditions and the following disclaimer in the documentation
10 // and/or other materials provided with the distribution.
12 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
13 // AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
14 // IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
15 // ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
16 // LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
17 // CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
18 // SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
19 // INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
20 // CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
21 // ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
22 // POSSIBILITY OF SUCH DAMAGE.
30 #include "wikitext_ragel.h"
32 #define IN(type) ary_includes(parser->scope, type)
34 // poor man's object orientation in C:
35 // instead of parsing around multiple parameters between functions in the parser
36 // we pack everything into a struct and pass around only a pointer to that
39 str_t *capture; // capturing to link_target, link_text, or NULL (direct to output, not capturing)
40 str_t *output; // for accumulating output to be returned
41 str_t *link_target; // short term "memory" for parsing links
42 str_t *link_text; // short term "memory" for parsing links
44 str_t *tabulation; // caching buffer for emitting indentation
45 ary_t *scope; // stack for tracking scope
46 ary_t *line; // stack for tracking scope as implied by current line
47 ary_t *line_buffer; // stack for tracking raw tokens (not scope) on current line
48 VALUE external_link_class; // CSS class applied to external links
49 VALUE mailto_class; // CSS class applied to email (mailto) links
50 VALUE img_prefix; // path prepended when emitting img tags
51 int base_indent; // controlled by the :indent option to Wikitext::Parser#parse
52 int current_indent; // fluctuates according to currently nested structures
53 int base_heading_level;
56 bool space_to_underscore;
59 const char null_str[] = { 0 };
60 const char escaped_no_wiki_start[] = "<nowiki>";
61 const char escaped_no_wiki_end[] = "</nowiki>";
62 const char literal_strong_em[] = "'''''";
63 const char literal_strong[] = "'''";
64 const char literal_em[] = "''";
65 const char escaped_em_start[] = "<em>";
66 const char escaped_em_end[] = "</em>";
67 const char escaped_strong_start[] = "<strong>";
68 const char escaped_strong_end[] = "</strong>";
69 const char escaped_tt_start[] = "<tt>";
70 const char escaped_tt_end[] = "</tt>";
71 const char literal_h6[] = "======";
72 const char literal_h5[] = "=====";
73 const char literal_h4[] = "====";
74 const char literal_h3[] = "===";
75 const char literal_h2[] = "==";
76 const char literal_h1[] = "=";
77 const char pre_start[] = "<pre>";
78 const char pre_end[] = "</pre>";
79 const char escaped_pre_start[] = "<pre>";
80 const char escaped_pre_end[] = "</pre>";
81 const char blockquote_start[] = "<blockquote>";
82 const char blockquote_end[] = "</blockquote>";
83 const char escaped_blockquote_start[] = "<blockquote>";
84 const char escaped_blockquote_end[] = "</blockquote>";
85 const char strong_em_start[] = "<strong><em>";
86 const char strong_start[] = "<strong>";
87 const char strong_end[] = "</strong>";
88 const char em_start[] = "<em>";
89 const char em_end[] = "</em>";
90 const char tt_start[] = "<tt>";
91 const char tt_end[] = "</tt>";
92 const char ol_start[] = "<ol>";
93 const char ol_end[] = "</ol>";
94 const char ul_start[] = "<ul>";
95 const char ul_end[] = "</ul>";
96 const char li_start[] = "<li>";
97 const char li_end[] = "</li>";
98 const char h6_start[] = "<h6>";
99 const char h6_end[] = "</h6>";
100 const char h5_start[] = "<h5>";
101 const char h5_end[] = "</h5>";
102 const char h4_start[] = "<h4>";
103 const char h4_end[] = "</h4>";
104 const char h3_start[] = "<h3>";
105 const char h3_end[] = "</h3>";
106 const char h2_start[] = "<h2>";
107 const char h2_end[] = "</h2>";
108 const char h1_start[] = "<h1>";
109 const char h1_end[] = "</h1>";
110 const char p_start[] = "<p>";
111 const char p_end[] = "</p>";
112 const char space[] = " ";
113 const char a_start[] = "<a href=\"";
114 const char a_class[] = "\" class=\"";
115 const char a_start_close[] = "\">";
116 const char a_end[] = "</a>";
117 const char link_start[] = "[[";
118 const char link_end[] = "]]";
119 const char separator[] = "|";
120 const char ext_link_start[] = "[";
121 const char backtick[] = "`";
122 const char quote[] = "\"";
123 const char ampersand[] = "&";
124 const char quot_entity[] = """;
125 const char amp_entity[] = "&";
126 const char lt_entity[] = "<";
127 const char gt_entity[] = ">";
128 const char escaped_blockquote[] = "> ";
129 const char ext_link_end[] = "]";
130 const char literal_img_start[] = "{{";
131 const char img_start[] = "<img src=\"";
132 const char img_end[] = "\" />";
133 const char img_alt[] = "\" alt=\"";
135 // Mark the parser struct designated by ptr as a participant in Ruby's
136 // mark-and-sweep garbage collection scheme. A variable named name is placed on
137 // the C stack to prevent the structure from being prematurely collected.
138 #define GC_WRAP_PARSER(ptr, name) volatile VALUE name __attribute__((unused)) = Data_Wrap_Struct(rb_cObject, 0, parser_free, ptr)
140 parser_t *parser_new(void)
142 parser_t *parser = ALLOC_N(parser_t, 1);
143 parser->capture = NULL; // not a real instance, pointer to other member's instance
144 parser->output = str_new();
145 parser->link_target = str_new();
146 parser->link_text = str_new();
147 parser->line_ending = NULL; // caller should set up
148 parser->tabulation = str_new();
149 parser->scope = ary_new();
150 parser->line = ary_new();
151 parser->line_buffer = ary_new();
152 parser->external_link_class = Qnil; // caller should set up
153 parser->mailto_class = Qnil; // caller should set up
154 parser->img_prefix = Qnil; // caller should set up
155 parser->base_indent = 0;
156 parser->current_indent = 0;
157 parser->base_heading_level = 0;
158 parser->pending_crlf = false;
159 parser->autolink = true;
160 parser->space_to_underscore = true;
164 void parser_free(parser_t *parser)
166 // we don't free parser->capture; it's just a redundant pointer
167 if (parser->output) str_free(parser->output);
168 if (parser->link_target) str_free(parser->link_target);
169 if (parser->link_text) str_free(parser->link_text);
170 if (parser->line_ending) str_free(parser->line_ending);
171 if (parser->tabulation) str_free(parser->tabulation);
172 if (parser->scope) ary_free(parser->scope);
173 if (parser->line) ary_free(parser->line);
174 if (parser->line_buffer) ary_free(parser->line_buffer);
178 // for testing and debugging only
179 VALUE Wikitext_parser_tokenize(VALUE self, VALUE string)
183 string = StringValue(string);
184 VALUE tokens = rb_ary_new();
185 char *p = RSTRING_PTR(string);
186 long len = RSTRING_LEN(string);
189 next_token(&token, NULL, p, pe);
190 rb_ary_push(tokens, _Wikitext_token(&token));
191 while (token.type != END_OF_FILE)
193 next_token(&token, &token, NULL, pe);
194 rb_ary_push(tokens, _Wikitext_token(&token));
199 // for benchmarking raw tokenization speed only
200 VALUE Wikitext_parser_benchmarking_tokenize(VALUE self, VALUE string)
204 string = StringValue(string);
205 char *p = RSTRING_PTR(string);
206 long len = RSTRING_LEN(string);
209 next_token(&token, NULL, p, pe);
210 while (token.type != END_OF_FILE)
211 next_token(&token, &token, NULL, pe);
215 VALUE Wikitext_parser_fulltext_tokenize(int argc, VALUE *argv, VALUE self)
218 VALUE string, options;
219 if (rb_scan_args(argc, argv, "11", &string, &options) == 1) // 1 mandatory argument, 1 optional argument
223 string = StringValue(string);
224 VALUE tokens = rb_ary_new();
226 // check instance variables
227 VALUE min = rb_iv_get(self, "@minimum_fulltext_token_length");
229 // process options hash (can override instance variables)
230 if (!NIL_P(options) && TYPE(options) == T_HASH)
232 if (rb_funcall(options, rb_intern("has_key?"), 1, ID2SYM(rb_intern("minimum"))) == Qtrue)
233 min = rb_hash_aref(options, ID2SYM(rb_intern("minimum")));
235 int min_len = NIL_P(min) ? 3 : NUM2INT(min);
240 char *p = RSTRING_PTR(string);
241 long len = RSTRING_LEN(string);
244 token_t *_token = &token;
245 next_token(&token, NULL, p, pe);
246 while (token.type != END_OF_FILE)
253 if (TOKEN_LEN(_token) >= min_len)
254 rb_ary_push(tokens, TOKEN_TEXT(_token));
257 // ignore everything else
260 next_token(&token, &token, NULL, pe);
265 // we downcase "in place", overwriting the original contents of the buffer
266 void _Wikitext_downcase_bang(char *ptr, long len)
268 for (long i = 0; i < len; i++)
270 if (ptr[i] >= 'A' && ptr[i] <= 'Z')
275 // prepare hyperlink and append it to parser->output
276 // if check_autolink is true, checks parser->autolink to decide whether to emit a real hyperlink
277 // or merely the literal link target
278 // if link_text is Qnil, the link_target is re-used for the link text
279 void _Wikitext_append_hyperlink(parser_t *parser, VALUE link_prefix, VALUE link_target, VALUE link_text, VALUE link_class, bool check_autolink)
281 if (check_autolink && !parser->autolink)
282 str_append_string(parser->output, link_target);
285 str_append(parser->output, a_start, sizeof(a_start) - 1); // <a href="
286 if (!NIL_P(link_prefix))
287 str_append_string(parser->output, link_prefix);
288 str_append_string(parser->output, link_target);
290 // special handling for mailto URIs
291 const char *mailto = "mailto:";
292 if (NIL_P(link_prefix) &&
293 RSTRING_LEN(link_target) >= (long)sizeof(mailto) &&
294 strncmp(mailto, RSTRING_PTR(link_target), sizeof(mailto)) == 0)
295 link_class = parser->mailto_class; // use mailto_class from parser
296 if (link_class != Qnil)
298 str_append(parser->output, a_class, sizeof(a_class) - 1); // " class="
299 str_append_string(parser->output, link_class);
301 str_append(parser->output, a_start_close, sizeof(a_start_close) - 1); // ">
302 if (NIL_P(link_text)) // re-use link_target
303 str_append_string(parser->output, link_target);
305 str_append_string(parser->output, link_text);
306 str_append(parser->output, a_end, sizeof(a_end) - 1); // </a>
310 void _Wikitext_append_img(parser_t *parser, char *token_ptr, int token_len)
312 str_append(parser->output, img_start, sizeof(img_start) - 1); // <img src="
313 if (!NIL_P(parser->img_prefix) && *token_ptr != '/') // len always > 0
314 str_append_string(parser->output, parser->img_prefix);
315 str_append(parser->output, token_ptr, token_len);
316 str_append(parser->output, img_alt, sizeof(img_alt) - 1); // " alt="
317 str_append(parser->output, token_ptr, token_len);
318 str_append(parser->output, img_end, sizeof(img_end) - 1); // " />
321 // will emit indentation only if we are about to emit any of:
322 // <blockquote>, <p>, <ul>, <ol>, <li>, <h1> etc, <pre>
323 // each time we enter one of those spans must ++ the indentation level
324 void _Wikitext_indent(parser_t *parser)
326 if (parser->base_indent == -1) // indentation disabled
328 int space_count = parser->current_indent + parser->base_indent;
331 char *old_end, *new_end;
332 if (parser->tabulation->len < space_count)
333 str_grow(parser->tabulation, space_count); // reallocates if necessary
334 old_end = parser->tabulation->ptr + parser->tabulation->len;
335 new_end = parser->tabulation->ptr + space_count;
336 while (old_end < new_end)
338 if (space_count > parser->tabulation->len)
339 parser->tabulation->len = space_count;
340 str_append(parser->output, parser->tabulation->ptr, space_count);
342 parser->current_indent += 2;
345 void _Wikitext_dedent(parser_t *parser, VALUE emit)
347 if (parser->base_indent == -1) // indentation disabled
349 parser->current_indent -= 2;
352 int space_count = parser->current_indent + parser->base_indent;
354 str_append(parser->output, parser->tabulation->ptr, space_count);
357 // Pops a single item off the parser's scope stack.
358 // A corresponding closing tag is written to the target string.
359 // The target string may be the main output buffer, or a substring capturing buffer if a link is being scanned.
360 void _Wikitext_pop_from_stack(parser_t *parser, str_t *target)
362 int top = ary_entry(parser->scope, -1);
366 target = parser->output;
368 // for headings, take base_heading_level into account
369 if (top >= H1_START && top <= H6_START)
371 top += parser->base_heading_level;
372 // no need to check for underflow (base_heading_level is never negative)
381 str_append(target, pre_end, sizeof(pre_end) - 1);
382 str_append_str(target, parser->line_ending);
383 _Wikitext_dedent(parser, Qfalse);
387 case BLOCKQUOTE_START:
388 _Wikitext_dedent(parser, Qtrue);
389 str_append(target, blockquote_end, sizeof(blockquote_end) - 1);
390 str_append_str(target, parser->line_ending);
394 // not a real HTML tag; so nothing to pop
399 str_append(target, strong_end, sizeof(strong_end) - 1);
404 str_append(target, em_end, sizeof(em_end) - 1);
409 str_append(target, tt_end, sizeof(tt_end) - 1);
413 _Wikitext_dedent(parser, Qtrue);
414 str_append(target, ol_end, sizeof(ol_end) - 1);
415 str_append_str(target, parser->line_ending);
419 _Wikitext_dedent(parser, Qtrue);
420 str_append(target, ul_end, sizeof(ul_end) - 1);
421 str_append_str(target, parser->line_ending);
425 // next token to pop will be a LI
426 // LI is an interesting token because sometimes we want it to behave like P (ie. do a non-emitting indent)
427 // and other times we want it to behave like BLOCKQUOTE (ie. when it has a nested list inside)
428 // hence this hack: we do an emitting dedent on behalf of the LI that we know must be coming
429 // and then when we pop the actual LI itself (below) we do the standard non-emitting indent
430 _Wikitext_dedent(parser, Qtrue); // we really only want to emit the spaces
431 parser->current_indent += 2; // we don't want to decrement the actual indent level, so put it back
435 str_append(target, li_end, sizeof(li_end) - 1);
436 str_append_str(target, parser->line_ending);
437 _Wikitext_dedent(parser, Qfalse);
441 str_append(target, h6_end, sizeof(h6_end) - 1);
442 str_append_str(target, parser->line_ending);
443 _Wikitext_dedent(parser, Qfalse);
447 str_append(target, h5_end, sizeof(h5_end) - 1);
448 str_append_str(target, parser->line_ending);
449 _Wikitext_dedent(parser, Qfalse);
453 str_append(target, h4_end, sizeof(h4_end) - 1);
454 str_append_str(target, parser->line_ending);
455 _Wikitext_dedent(parser, Qfalse);
459 str_append(target, h3_end, sizeof(h3_end) - 1);
460 str_append_str(target, parser->line_ending);
461 _Wikitext_dedent(parser, Qfalse);
465 str_append(target, h2_end, sizeof(h2_end) - 1);
466 str_append_str(target, parser->line_ending);
467 _Wikitext_dedent(parser, Qfalse);
471 str_append(target, h1_end, sizeof(h1_end) - 1);
472 str_append_str(target, parser->line_ending);
473 _Wikitext_dedent(parser, Qfalse);
477 // not an HTML tag; so nothing to emit
481 // not an HTML tag; so nothing to emit
485 // not an HTML tag; so nothing to emit
489 // not an HTML tag (only used to separate an external link target from the link text); so nothing to emit
493 // not an HTML tag (only used to separate an external link target from the link text); so nothing to emit
497 str_append(target, p_end, sizeof(p_end) - 1);
498 str_append_str(target, parser->line_ending);
499 _Wikitext_dedent(parser, Qfalse);
507 // should probably raise an exception here
510 ary_pop(parser->scope);
513 // Pops items off the top of parser's scope stack, accumulating closing tags for them into the target string, until item is reached.
514 // If including is true then the item itself is also popped.
515 // The target string may be the main output buffer, or a substring capturing buffer when scanning links.
516 void _Wikitext_pop_from_stack_up_to(parser_t *parser, str_t *target, int item, bool including)
518 int continue_looping = 1;
521 int top = ary_entry(parser->scope, -1);
528 continue_looping = 0;
530 _Wikitext_pop_from_stack(parser, target);
531 } while (continue_looping);
534 void _Wikitext_pop_all_from_stack(parser_t *parser)
536 for (int i = 0, max = parser->scope->count; i < max; i++)
537 _Wikitext_pop_from_stack(parser, NULL);
540 void _Wikitext_start_para_if_necessary(parser_t *parser)
545 // if no block open yet, or top of stack is BLOCKQUOTE/BLOCKQUOTE_START (with nothing in it yet)
546 if (parser->scope->count == 0 ||
547 ary_entry(parser->scope, -1) == BLOCKQUOTE ||
548 ary_entry(parser->scope, -1) == BLOCKQUOTE_START)
550 _Wikitext_indent(parser);
551 str_append(parser->output, p_start, sizeof(p_start) - 1);
552 ary_push(parser->scope, P);
553 ary_push(parser->line, P);
555 else if (parser->pending_crlf)
558 // already in a paragraph block; convert pending CRLF into a space
559 str_append(parser->output, space, sizeof(space) - 1);
561 // PRE blocks can have pending CRLF too (helps us avoid emitting the trailing newline)
562 str_append_str(parser->output, parser->line_ending);
564 parser->pending_crlf = false;
567 void _Wikitext_emit_pending_crlf_if_necessary(parser_t *parser)
569 if (parser->pending_crlf)
571 str_append_str(parser->output, parser->line_ending);
572 parser->pending_crlf = false;
576 // Helper function that pops any excess elements off scope (pushing is already handled in the respective rules).
577 // For example, given input like:
582 // Upon seeing "bar", we want to pop two BLOCKQUOTE elements from the scope.
583 // The reverse case (shown below) is handled from inside the BLOCKQUOTE rule itself:
588 // Things are made slightly more complicated by the fact that there is one block-level tag that can be on the scope
589 // but not on the line scope:
594 // Here on seeing "bar" we have one item on the scope (BLOCKQUOTE_START) which we don't want to pop, but we have nothing
595 // on the line scope.
596 // Luckily, BLOCKQUOTE_START tokens can only appear at the start of the scope array, so we can check for them first before
597 // entering the for loop.
598 void _Wikitext_pop_excess_elements(parser_t *parser)
602 for (int i = parser->scope->count - ary_count(parser->scope, BLOCKQUOTE_START), j = parser->line->count; i > j; i--)
604 // special case for last item on scope
607 // don't auto-pop P if it is only item on scope
608 if (ary_entry(parser->scope, -1) == P)
610 // add P to the line scope to prevent us entering the loop at all next time around
611 ary_push(parser->line, P);
615 _Wikitext_pop_from_stack(parser, NULL);
619 #define INVALID_ENCODING(msg) do { if (dest_ptr) free(dest_ptr); rb_raise(eWikitextParserError, "invalid encoding: " msg); } while(0)
621 // Convert a single UTF-8 codepoint to UTF-32
623 // Expects an input buffer, src, containing a UTF-8 encoded character (which
624 // may be multi-byte). The end of the input buffer, end, is also passed in to
625 // allow the detection of invalidly truncated codepoints. The number of bytes
626 // in the UTF-8 character (between 1 and 4) is returned by reference in
629 // Raises a RangeError if the supplied character is invalid UTF-8 (in which
630 // case it also frees the block of memory indicated by dest_ptr if it is
632 uint32_t _Wikitext_utf8_to_utf32(char *src, char *end, long *width_out, void *dest_ptr)
635 if ((unsigned char)src[0] <= 0x7f)
641 else if ((src[0] & 0xe0) == 0xc0)
643 // byte starts with 110..... : this should be a two-byte sequence
646 INVALID_ENCODING("truncated byte sequence");
647 else if (((unsigned char)src[0] == 0xc0) ||
648 ((unsigned char)src[0] == 0xc1))
649 // overlong encoding: lead byte of 110..... but code point <= 127
650 INVALID_ENCODING("overlong encoding");
651 else if ((src[1] & 0xc0) != 0x80 )
652 // should have second byte starting with 10......
653 INVALID_ENCODING("malformed byte sequence");
656 ((uint32_t)(src[0] & 0x1f)) << 6 |
660 else if ((src[0] & 0xf0) == 0xe0)
662 // byte starts with 1110.... : this should be a three-byte sequence
664 // missing second or third byte
665 INVALID_ENCODING("truncated byte sequence");
666 else if (((src[1] & 0xc0) != 0x80 ) ||
667 ((src[2] & 0xc0) != 0x80 ))
668 // should have second and third bytes starting with 10......
669 INVALID_ENCODING("malformed byte sequence");
672 ((uint32_t)(src[0] & 0x0f)) << 12 |
673 ((uint32_t)(src[1] & 0x3f)) << 6 |
677 else if ((src[0] & 0xf8) == 0xf0)
679 // bytes starts with 11110... : this should be a four-byte sequence
681 // missing second, third, or fourth byte
682 INVALID_ENCODING("truncated byte sequence");
683 else if ((unsigned char)src[0] >= 0xf5 &&
684 (unsigned char)src[0] <= 0xf7)
685 // disallowed by RFC 3629 (codepoints above 0x10ffff)
686 INVALID_ENCODING("overlong encoding");
687 else if (((src[1] & 0xc0) != 0x80 ) ||
688 ((src[2] & 0xc0) != 0x80 ) ||
689 ((src[3] & 0xc0) != 0x80 ))
690 // should have second and third bytes starting with 10......
691 INVALID_ENCODING("malformed byte sequence");
694 ((uint32_t)(src[0] & 0x07)) << 18 |
695 ((uint32_t)(src[1] & 0x3f)) << 12 |
696 ((uint32_t)(src[1] & 0x3f)) << 6 |
701 INVALID_ENCODING("unexpected byte");
705 void _Wikitext_append_entity_from_utf32_char(char *output, uint32_t character)
707 // TODO: consider special casing some entities (ie. quot, amp, lt, gt etc)?
708 char hex_string[8] = { '&', '#', 'x', 0, 0, 0, 0, ';' };
709 char scratch = (character & 0xf000) >> 12;
710 hex_string[3] = (scratch <= 9 ? scratch + 48 : scratch + 87);
711 scratch = (character & 0x0f00) >> 8;
712 hex_string[4] = (scratch <= 9 ? scratch + 48 : scratch + 87);
713 scratch = (character & 0x00f0) >> 4;
714 hex_string[5] = (scratch <= 9 ? scratch + 48 : scratch + 87);
715 scratch = character & 0x000f;
716 hex_string[6] = (scratch <= 9 ? scratch + 48 : scratch + 87);
717 memcpy(output, hex_string, sizeof(hex_string));
720 // trim parser->link_text in place
721 void _Wikitext_parser_trim_link_text(parser_t *parser)
723 char *src = parser->link_text->ptr;
724 char *start = src; // remember this so we can check if we're at the start
726 char *non_space = src; // remember last non-space character output
727 long len = parser->link_text->len;
728 char *end = src + len;
740 if (left != start || non_space + 1 != end)
742 // TODO: could potentially avoid this memmove by extending the str_t struct with an "offset" or "free" member
743 parser->link_text->len = (non_space + 1) - left;
744 memmove(parser->link_text->ptr, left, parser->link_text->len);
748 // - non-printable (non-ASCII) characters converted to numeric entities
749 // - QUOT and AMP characters converted to named entities
750 // - if rollback is true, there is no special treatment of spaces
751 // - if rollback is false, leading and trailing whitespace trimmed
752 void _Wikitext_parser_append_sanitized_link_target(parser_t *parser, str_t *output, bool rollback)
754 char *src = parser->link_target->ptr;
755 char *start = src; // remember this so we can check if we're at the start
756 long len = parser->link_target->len;
757 char *end = src + len;
759 // start with a destination buffer twice the size of the source, will realloc if necessary
760 // slop = (len / 8) * 8 (ie. one in every 8 characters can be converted into an entity, each entity requires 8 bytes)
761 // this efficiently handles the most common case (where the size of the buffer doesn't change much)
762 char *dest = ALLOC_N(char, len * 2);
763 char *dest_ptr = dest; // hang on to this so we can pass it to free() later
764 char *non_space = dest; // remember last non-space character output
767 // need at most 8 characters (8 bytes) to display each character
768 if (dest + 8 > dest_ptr + len) // outgrowing buffer, must reallocate
770 char *old_dest = dest;
771 char *old_dest_ptr = dest_ptr;
772 len = len + (end - src) * 8; // allocate enough for worst case
773 dest = realloc(dest_ptr, len); // will never have to realloc more than once
776 // would have used reallocf, but this has to run on Linux too, not just Darwin
778 rb_raise(rb_eNoMemError, "failed to re-allocate temporary storage (memory allocation error)");
781 dest = dest_ptr + (old_dest - old_dest_ptr);
782 non_space = dest_ptr + (non_space - old_dest_ptr);
787 char quot_entity_literal[] = { '&', 'q', 'u', 'o', 't', ';' }; // no trailing NUL
788 memcpy(dest, quot_entity_literal, sizeof(quot_entity_literal));
789 dest += sizeof(quot_entity_literal);
791 else if (*src == '&')
793 char amp_entity_literal[] = { '&', 'a', 'm', 'p', ';' }; // no trailing NUL
794 memcpy(dest, amp_entity_literal, sizeof(amp_entity_literal));
795 dest += sizeof(amp_entity_literal);
797 else if (*src == '<' || *src == '>')
800 rb_raise(rb_eRangeError, "invalid link text (\"%c\" may not appear in link text)", *src);
802 else if (*src == ' ' && src == start && !rollback)
803 start++; // we eat leading space
804 else if (*src >= 0x20 && *src <= 0x7e) // printable ASCII
809 else // all others: must convert to entities
812 _Wikitext_append_entity_from_utf32_char(dest, _Wikitext_utf8_to_utf32(src, end, &width, dest_ptr));
813 dest += 8; // always 8 characters (�)
823 // trim trailing space if necessary
824 if (!rollback && non_space > dest_ptr && dest != non_space)
825 len = non_space - dest_ptr;
827 len = dest - dest_ptr;
828 str_append(output, dest_ptr, len);
832 VALUE Wikitext_parser_sanitize_link_target(VALUE self, VALUE string)
835 parser.link_target = str_new_from_string(string);
836 GC_WRAP_STR(parser.link_target, link_target_gc);
837 str_t *output = str_new();
838 GC_WRAP_STR(output, output_gc);
839 _Wikitext_parser_append_sanitized_link_target(&parser, output, false);
840 return string_from_str(output);
843 // encodes the input string according to RFCs 2396 and 2718
844 // leading and trailing whitespace trimmed
845 // note that the first character of the target link is not case-sensitive
846 // (this is a recommended application-level constraint; it is not imposed at this level)
847 // this is to allow links like:
848 // ...the [[foo]] is...
849 // to be equivalent to:
850 // thing. [[Foo]] was...
851 static void _Wikitext_parser_encode_link_target(parser_t *parser)
853 VALUE in = string_from_str(parser->link_target);
854 char *input = RSTRING_PTR(in);
855 char *start = input; // remember this so we can check if we're at the start
856 long len = RSTRING_LEN(in);
859 char *end = input + len;
860 static char hex[] = { '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'a', 'b', 'c', 'd', 'e', 'f' };
862 // to avoid most reallocations start with a destination buffer twice the size of the source
863 // this handles the most common case (where most chars are in the ASCII range and don't require more storage, but there are
864 // often quite a few spaces, which are encoded as "%20" and occupy 3 bytes)
865 // the worst case is where _every_ byte must be written out using 3 bytes
866 long dest_len = len * 2;
867 char *dest = ALLOC_N(char, dest_len);
868 char *dest_ptr = dest; // hang on to this so we can pass it to free() later
869 char *non_space = dest; // remember last non-space character output
870 for (; input < end; input++)
872 if ((dest + 3) > (dest_ptr + dest_len)) // worst case: a single character may grow to 3 characters once encoded
874 // outgrowing buffer, must reallocate
875 char *old_dest = dest;
876 char *old_dest_ptr = dest_ptr;
878 dest = realloc(dest_ptr, dest_len);
881 // would have used reallocf, but this has to run on Linux too, not just Darwin
883 rb_raise(rb_eNoMemError, "failed to re-allocate temporary storage (memory allocation error)");
886 dest = dest_ptr + (old_dest - old_dest_ptr);
887 non_space = dest_ptr + (non_space - old_dest_ptr);
890 // pass through unreserved characters
891 if (((*input >= 'a') && (*input <= 'z')) ||
892 ((*input >= 'A') && (*input <= 'Z')) ||
893 ((*input >= '0') && (*input <= '9')) ||
902 else if (*input == ' ' && input == start)
903 start++; // we eat leading space
904 else if (*input == ' ' && parser->space_to_underscore)
906 else // everything else gets URL-encoded
909 *dest++ = hex[(unsigned char)(*input) / 16]; // left
910 *dest++ = hex[(unsigned char)(*input) % 16]; // right
916 // trim trailing space if necessary
917 if (non_space > dest_ptr && dest != non_space)
918 dest_len = non_space - dest_ptr;
920 dest_len = dest - dest_ptr;
922 // TOOD: try to avoid this copy by working directly inside link_target buffer
923 str_clear(parser->link_target);
924 str_append(parser->link_target, dest_ptr, dest_len);
928 VALUE Wikitext_parser_encode_link_target(VALUE self, VALUE in)
931 parser.space_to_underscore = false;
932 parser.link_target = str_new_from_string(in);
933 GC_WRAP_STR(parser.link_target, link_target_gc);
934 _Wikitext_parser_encode_link_target(&parser);
935 return string_from_str(parser.link_target);
938 // this method exposed for testing only
939 VALUE Wikitext_parser_encode_special_link_target(VALUE self, VALUE in)
942 parser.space_to_underscore = false;
943 parser.link_target = str_new_from_string(in);
944 GC_WRAP_STR(parser.link_target, link_target_gc);
945 _Wikitext_parser_encode_link_target(&parser);
946 return string_from_str(parser.link_target);
949 // returns 1 (true) if supplied string is blank (nil, empty, or all whitespace)
950 // returns 0 (false) otherwise
951 bool _Wikitext_blank(str_t *str)
955 for (char *ptr = str->ptr,
956 *end = str->ptr + str->len;
965 void _Wikitext_rollback_failed_internal_link(parser_t *parser)
968 return; // nothing to do!
969 int scope_includes_separator = IN(SEPARATOR);
970 _Wikitext_pop_from_stack_up_to(parser, NULL, LINK_START, true);
971 str_append(parser->output, link_start, sizeof(link_start) - 1);
972 if (parser->link_target->len > 0)
974 _Wikitext_parser_append_sanitized_link_target(parser, parser->output, true);
975 if (scope_includes_separator)
977 str_append(parser->output, separator, sizeof(separator) - 1);
978 if (parser->link_text->len > 0)
979 str_append_str(parser->output, parser->link_text);
982 parser->capture = NULL;
983 str_clear(parser->link_target);
984 str_clear(parser->link_text);
987 void _Wikitext_rollback_failed_external_link(parser_t *parser)
989 if (!IN(EXT_LINK_START))
990 return; // nothing to do!
992 // store a couple of values before popping
993 int scope_includes_space = IN(SPACE);
994 VALUE link_class = IN(PATH) ? Qnil : parser->external_link_class;
995 _Wikitext_pop_from_stack_up_to(parser, NULL, EXT_LINK_START, true);
997 str_append(parser->output, ext_link_start, sizeof(ext_link_start) - 1);
998 if (parser->link_target->len > 0)
1000 _Wikitext_append_hyperlink(parser, Qnil, string_from_str(parser->link_target), Qnil, link_class, true);
1001 if (scope_includes_space)
1003 str_append(parser->output, space, sizeof(space) - 1);
1004 if (parser->link_text->len > 0)
1005 str_append_str(parser->output, parser->link_text);
1008 parser->capture = NULL;
1009 str_clear(parser->link_target);
1010 str_clear(parser->link_text);
1013 void _Wikitext_rollback_failed_link(parser_t *parser)
1015 _Wikitext_rollback_failed_internal_link(parser);
1016 _Wikitext_rollback_failed_external_link(parser);
1019 VALUE Wikitext_parser_initialize(int argc, VALUE *argv, VALUE self)
1021 // process arguments
1023 if (rb_scan_args(argc, argv, "01", &options) == 0) // 0 mandatory arguments, 1 optional argument
1027 VALUE autolink = Qtrue;
1028 VALUE line_ending = rb_str_new2("\n");
1029 VALUE external_link_class = rb_str_new2("external");
1030 VALUE mailto_class = rb_str_new2("mailto");
1031 VALUE internal_link_prefix = rb_str_new2("/wiki/");
1032 VALUE img_prefix = rb_str_new2("/images/");
1033 VALUE space_to_underscore = Qtrue;
1034 VALUE minimum_fulltext_token_length = INT2NUM(3);
1035 VALUE base_heading_level = INT2NUM(0);
1037 // process options hash (override defaults)
1038 if (!NIL_P(options) && TYPE(options) == T_HASH)
1040 #define OVERRIDE_IF_SET(name) rb_funcall(options, rb_intern("has_key?"), 1, ID2SYM(rb_intern(#name))) == Qtrue ? \
1041 rb_hash_aref(options, ID2SYM(rb_intern(#name))) : name
1042 autolink = OVERRIDE_IF_SET(autolink);
1043 line_ending = OVERRIDE_IF_SET(line_ending);
1044 external_link_class = OVERRIDE_IF_SET(external_link_class);
1045 mailto_class = OVERRIDE_IF_SET(mailto_class);
1046 internal_link_prefix = OVERRIDE_IF_SET(internal_link_prefix);
1047 img_prefix = OVERRIDE_IF_SET(img_prefix);
1048 space_to_underscore = OVERRIDE_IF_SET(space_to_underscore);
1049 minimum_fulltext_token_length = OVERRIDE_IF_SET(minimum_fulltext_token_length);
1050 base_heading_level = OVERRIDE_IF_SET(base_heading_level);
1053 // no need to call super here; rb_call_super()
1054 rb_iv_set(self, "@autolink", autolink);
1055 rb_iv_set(self, "@line_ending", line_ending);
1056 rb_iv_set(self, "@external_link_class", external_link_class);
1057 rb_iv_set(self, "@mailto_class", mailto_class);
1058 rb_iv_set(self, "@internal_link_prefix", internal_link_prefix);
1059 rb_iv_set(self, "@img_prefix", img_prefix);
1060 rb_iv_set(self, "@space_to_underscore", space_to_underscore);
1061 rb_iv_set(self, "@minimum_fulltext_token_length", minimum_fulltext_token_length);
1062 rb_iv_set(self, "@base_heading_level", base_heading_level);
1066 VALUE Wikitext_parser_profiling_parse(VALUE self, VALUE string)
1068 for (int i = 0; i < 100000; i++)
1069 Wikitext_parser_parse(1, &string, self);
1073 VALUE Wikitext_parser_parse(int argc, VALUE *argv, VALUE self)
1075 // process arguments
1076 VALUE string, options;
1077 if (rb_scan_args(argc, argv, "11", &string, &options) == 1) // 1 mandatory argument, 1 optional argument
1081 string = StringValue(string);
1083 // process options hash
1084 int base_indent = 0;
1085 int base_heading_level = NUM2INT(rb_iv_get(self, "@base_heading_level"));
1086 if (!NIL_P(options) && TYPE(options) == T_HASH)
1088 // :indent => 0 (or more)
1089 if (rb_funcall(options, rb_intern("has_key?"), 1, ID2SYM(rb_intern("indent"))) == Qtrue)
1091 VALUE indent = rb_hash_aref(options, ID2SYM(rb_intern("indent")));
1092 if (indent == Qfalse)
1093 base_indent = -1; // indentation disabled
1096 base_indent = NUM2INT(indent);
1097 if (base_indent < 0)
1102 // :base_heading_level => 0/1/2/3/4/5/6
1103 if (rb_funcall(options, rb_intern("has_key?"), 1, ID2SYM(rb_intern("base_heading_level"))) == Qtrue)
1104 base_heading_level = NUM2INT(rb_hash_aref(options, ID2SYM(rb_intern("base_heading_level"))));
1107 // normalize, regardless of whether this came from instance variable or override
1108 if (base_heading_level < 0)
1109 base_heading_level = 0;
1110 if (base_heading_level > 6)
1111 base_heading_level = 6;
1114 char *p = RSTRING_PTR(string);
1115 long len = RSTRING_LEN(string);
1118 // the eventual return value
1119 VALUE out = rb_str_new2("");
1121 // access these once per parse
1122 VALUE line_ending = rb_iv_get(self, "@line_ending");
1123 line_ending = StringValue(line_ending);
1124 VALUE link_class = rb_iv_get(self, "@external_link_class");
1125 link_class = NIL_P(link_class) ? Qnil : StringValue(link_class);
1126 VALUE mailto_class = rb_iv_get(self, "@mailto_class");
1127 mailto_class = NIL_P(mailto_class) ? Qnil : StringValue(mailto_class);
1128 VALUE prefix = rb_iv_get(self, "@internal_link_prefix");
1130 // set up parser struct to make passing parameters a little easier
1131 parser_t *parser = parser_new();
1132 GC_WRAP_PARSER(parser, parser_gc);
1133 parser->external_link_class = link_class;
1134 parser->mailto_class = mailto_class;
1135 parser->img_prefix = rb_iv_get(self, "@img_prefix");
1136 parser->autolink = rb_iv_get(self, "@autolink") == Qtrue ? true : false;
1137 parser->space_to_underscore = rb_iv_get(self, "@space_to_underscore") == Qtrue ? true : false;
1138 parser->line_ending = str_new_from_string(line_ending);
1139 parser->base_indent = base_indent;
1140 parser->base_heading_level = base_heading_level;
1142 // this simple looping design leads to a single enormous function,
1143 // but it's faster than doing actual recursive descent and also secure in the face of
1144 // malicious input that seeks to overflow the stack
1145 // (with "<blockquote><blockquote><blockquote>..." times by 10,000, for example)
1146 // given that we expect to deal with a lot of malformed input, a recursive descent design is less appropriate
1147 // than a straightforward looping translator like this one anyway
1149 _token.type = NO_TOKEN;
1150 token_t *token = NULL;
1153 // note that whenever we grab a token we push it into the line buffer
1154 // this provides us with context-sensitive "memory" of what's been seen so far on this line
1155 #define NEXT_TOKEN() token = &_token, next_token(token, token, NULL, pe), ary_push(parser->line_buffer, token->type)
1157 // check to see if we have a token hanging around from a previous iteration of this loop
1160 if (_token.type == NO_TOKEN)
1162 // first time here (haven't started scanning yet)
1164 next_token(token, NULL, p, pe);
1165 ary_push(parser->line_buffer, token->type);
1171 int type = token->type;
1173 // can't declare new variables inside a switch statement, so predeclare them here
1174 long remove_strong = -1;
1175 long remove_em = -1;
1177 // general purpose counters, flags and pointers
1181 str_t *output = NULL;
1183 // The following giant switch statement contains cases for all the possible token types.
1184 // In the most basic sense we are emitting the HTML that corresponds to each token,
1185 // but some tokens require context information in order to decide what to output.
1186 // For example, does the STRONG token (''') translate to <strong> or </strong>?
1187 // So when looking at any given token we have three state-maintaining variables which gives us a notion of "where we are":
1189 // - the "scope" stack (indicates what HTML DOM structures we are currently nested inside, similar to a CSS selector)
1190 // - the line buffer (records tokens seen so far on the current line)
1191 // - the line "scope" stack (indicates what the scope should be based only on what is visible on the line so far)
1193 // Although this is fairly complicated, there is one key simplifying factor:
1194 // The translator continuously performs auto-correction, and this means that we always have a guarantee that the
1195 // scope stack (up to the current token) is valid; our translator can take this as a given.
1196 // Auto-correction basically consists of inserting missing tokens (preventing subsquent HTML from being messed up),
1197 // or converting illegal (unexpected) tokens to their plain text equivalents (providing visual feedback to Wikitext author).
1201 if (IN(NO_WIKI_START) || IN(PRE_START))
1203 str_append(parser->output, space, sizeof(space) - 1);
1206 else if (IN(BLOCKQUOTE_START))
1208 // this kind of nesting not allowed (to avoid user confusion)
1209 _Wikitext_pop_excess_elements(parser);
1210 _Wikitext_start_para_if_necessary(parser);
1211 output = parser->capture ? parser->capture : parser->output;
1212 str_append(output, space, sizeof(space) - 1);
1216 // count number of BLOCKQUOTE tokens in line buffer and in scope stack
1217 ary_push(parser->line, PRE);
1218 i = ary_count(parser->line, BLOCKQUOTE);
1219 j = ary_count(parser->scope, BLOCKQUOTE);
1222 // must pop (reduce nesting level)
1223 for (i = j - i; i > 0; i--)
1224 _Wikitext_pop_from_stack_up_to(parser, NULL, BLOCKQUOTE, true);
1229 parser->pending_crlf = false;
1230 _Wikitext_pop_from_stack_up_to(parser, NULL, BLOCKQUOTE, false);
1231 _Wikitext_indent(parser);
1232 str_append(parser->output, pre_start, sizeof(pre_start) - 1);
1233 ary_push(parser->scope, PRE);
1238 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1240 _Wikitext_emit_pending_crlf_if_necessary(parser);
1241 str_append(parser->output, escaped_pre_start, sizeof(escaped_pre_start) - 1);
1243 else if (IN(BLOCKQUOTE_START))
1245 _Wikitext_rollback_failed_link(parser); // if any
1246 _Wikitext_pop_from_stack_up_to(parser, NULL, BLOCKQUOTE_START, false);
1247 _Wikitext_indent(parser);
1248 str_append(parser->output, pre_start, sizeof(pre_start) - 1);
1249 ary_push(parser->scope, PRE_START);
1250 ary_push(parser->line, PRE_START);
1252 else if (IN(BLOCKQUOTE))
1254 if (token->column_start == 1) // only allowed in first column
1256 _Wikitext_rollback_failed_link(parser); // if any
1257 _Wikitext_pop_all_from_stack(parser);
1258 _Wikitext_indent(parser);
1259 str_append(parser->output, pre_start, sizeof(pre_start) - 1);
1260 ary_push(parser->scope, PRE_START);
1261 ary_push(parser->line, PRE_START);
1263 else // PRE_START illegal here
1265 output = parser->capture ? parser->capture : parser->output;
1266 _Wikitext_pop_excess_elements(parser);
1267 _Wikitext_start_para_if_necessary(parser);
1268 str_append(output, escaped_pre_start, sizeof(escaped_pre_start) - 1);
1273 _Wikitext_rollback_failed_link(parser); // if any
1274 _Wikitext_pop_from_stack_up_to(parser, NULL, P, true);
1275 _Wikitext_indent(parser);
1276 str_append(parser->output, pre_start, sizeof(pre_start) - 1);
1277 ary_push(parser->scope, PRE_START);
1278 ary_push(parser->line, PRE_START);
1283 if (IN(NO_WIKI_START) || IN(PRE))
1285 _Wikitext_emit_pending_crlf_if_necessary(parser);
1286 str_append(parser->output, escaped_pre_end, sizeof(escaped_pre_end) - 1);
1291 _Wikitext_pop_from_stack_up_to(parser, parser->output, PRE_START, true);
1294 output = parser->capture ? parser->capture : parser->output;
1295 _Wikitext_pop_excess_elements(parser);
1296 _Wikitext_start_para_if_necessary(parser);
1297 str_append(output, escaped_pre_end, sizeof(escaped_pre_end) - 1);
1303 if (IN(NO_WIKI_START) || IN(PRE_START))
1304 // no need to check for <pre>; can never appear inside it
1305 str_append(parser->output, escaped_blockquote, TOKEN_LEN(token) + 3); // will either emit ">" or "> "
1306 else if (IN(BLOCKQUOTE_START))
1308 // this kind of nesting not allowed (to avoid user confusion)
1309 _Wikitext_pop_excess_elements(parser);
1310 _Wikitext_start_para_if_necessary(parser);
1311 output = parser->capture ? parser->capture : parser->output;
1312 str_append(output, escaped_blockquote, TOKEN_LEN(token) + 3); // will either emit ">" or "> "
1317 ary_push(parser->line, BLOCKQUOTE);
1319 // count number of BLOCKQUOTE tokens in line buffer and in scope stack
1320 i = ary_count(parser->line, BLOCKQUOTE);
1321 j = ary_count(parser->scope, BLOCKQUOTE);
1323 // given that BLOCKQUOTE tokens can be nested, peek ahead and see if there are any more which might affect the decision to push or pop
1324 while (NEXT_TOKEN(), (token->type == BLOCKQUOTE))
1326 ary_push(parser->line, BLOCKQUOTE);
1330 // now decide whether to push, pop or do nothing
1333 // must push (increase nesting level)
1334 _Wikitext_pop_from_stack_up_to(parser, NULL, BLOCKQUOTE, false);
1335 for (i = i - j; i > 0; i--)
1337 _Wikitext_indent(parser);
1338 str_append(parser->output, blockquote_start, sizeof(blockquote_start) - 1);
1339 str_append_str(parser->output, parser->line_ending);
1340 ary_push(parser->scope, BLOCKQUOTE);
1345 // must pop (reduce nesting level)
1346 for (i = j - i; i > 0; i--)
1347 _Wikitext_pop_from_stack_up_to(parser, NULL, BLOCKQUOTE, true);
1350 // jump to top of the loop to process token we scanned during lookahead
1355 case BLOCKQUOTE_START:
1356 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1358 _Wikitext_emit_pending_crlf_if_necessary(parser);
1359 str_append(parser->output, escaped_blockquote_start, sizeof(escaped_blockquote_start) - 1);
1361 else if (IN(BLOCKQUOTE_START))
1363 // nesting is fine here
1364 _Wikitext_rollback_failed_link(parser); // if any
1365 _Wikitext_pop_from_stack_up_to(parser, NULL, BLOCKQUOTE_START, false);
1366 _Wikitext_indent(parser);
1367 str_append(parser->output, blockquote_start, sizeof(blockquote_start) - 1);
1368 str_append_str(parser->output, parser->line_ending);
1369 ary_push(parser->scope, BLOCKQUOTE_START);
1370 ary_push(parser->line, BLOCKQUOTE_START);
1372 else if (IN(BLOCKQUOTE))
1374 if (token->column_start == 1) // only allowed in first column
1376 _Wikitext_rollback_failed_link(parser); // if any
1377 _Wikitext_pop_all_from_stack(parser);
1378 _Wikitext_indent(parser);
1379 str_append(parser->output, blockquote_start, sizeof(blockquote_start) - 1);
1380 str_append_str(parser->output, parser->line_ending);
1381 ary_push(parser->scope, BLOCKQUOTE_START);
1382 ary_push(parser->line, BLOCKQUOTE_START);
1384 else // BLOCKQUOTE_START illegal here
1386 output = parser->capture ? parser->capture : parser->output;
1387 _Wikitext_pop_excess_elements(parser);
1388 _Wikitext_start_para_if_necessary(parser);
1389 str_append(output, escaped_blockquote_start, sizeof(escaped_blockquote_start) - 1);
1394 // would be nice to eliminate the repetition here but it's probably the clearest way
1395 _Wikitext_rollback_failed_link(parser); // if any
1396 _Wikitext_pop_from_stack_up_to(parser, NULL, P, true);
1397 _Wikitext_indent(parser);
1398 str_append(parser->output, blockquote_start, sizeof(blockquote_start) - 1);
1399 str_append_str(parser->output, parser->line_ending);
1400 ary_push(parser->scope, BLOCKQUOTE_START);
1401 ary_push(parser->line, BLOCKQUOTE_START);
1405 case BLOCKQUOTE_END:
1406 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1408 _Wikitext_emit_pending_crlf_if_necessary(parser);
1409 str_append(parser->output, escaped_blockquote_end, sizeof(escaped_blockquote_end) - 1);
1413 if (IN(BLOCKQUOTE_START))
1414 _Wikitext_pop_from_stack_up_to(parser, parser->output, BLOCKQUOTE_START, true);
1417 output = parser->capture ? parser->capture : parser->output;
1418 _Wikitext_pop_excess_elements(parser);
1419 _Wikitext_start_para_if_necessary(parser);
1420 str_append(output, escaped_blockquote_end, sizeof(escaped_blockquote_end) - 1);
1426 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1428 _Wikitext_emit_pending_crlf_if_necessary(parser);
1429 str_append(parser->output, escaped_no_wiki_start, sizeof(escaped_no_wiki_start) - 1);
1433 _Wikitext_pop_excess_elements(parser);
1434 _Wikitext_start_para_if_necessary(parser);
1435 ary_push(parser->scope, NO_WIKI_START);
1436 ary_push(parser->line, NO_WIKI_START);
1441 if (IN(NO_WIKI_START))
1442 // <nowiki> should always only ever be the last item in the stack, but use the helper routine just in case
1443 _Wikitext_pop_from_stack_up_to(parser, NULL, NO_WIKI_START, true);
1446 _Wikitext_pop_excess_elements(parser);
1447 _Wikitext_start_para_if_necessary(parser);
1448 str_append(parser->output, escaped_no_wiki_end, sizeof(escaped_no_wiki_end) - 1);
1453 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1455 _Wikitext_emit_pending_crlf_if_necessary(parser);
1456 str_append(parser->output, literal_strong_em, sizeof(literal_strong_em) - 1);
1460 output = parser->capture ? parser->capture : parser->output;
1461 _Wikitext_pop_excess_elements(parser);
1463 // if you've seen STRONG/STRONG_START or EM/EM_START, must close them in the reverse order that you saw them!
1464 // otherwise, must open them
1467 j = parser->scope->count;
1468 for (j = j - 1; j >= 0; j--)
1470 int val = ary_entry(parser->scope, j);
1471 if (val == STRONG || val == STRONG_START)
1473 str_append(output, strong_end, sizeof(strong_end) - 1);
1476 else if (val == EM || val == EM_START)
1478 str_append(output, em_end, sizeof(em_end) - 1);
1483 if (remove_strong > remove_em) // must remove strong first
1485 ary_pop(parser->scope);
1487 ary_pop(parser->scope);
1488 else // there was no em to remove!, so consider this an opening em tag
1490 str_append(output, em_start, sizeof(em_start) - 1);
1491 ary_push(parser->scope, EM);
1492 ary_push(parser->line, EM);
1495 else if (remove_em > remove_strong) // must remove em first
1497 ary_pop(parser->scope);
1498 if (remove_strong > -1)
1499 ary_pop(parser->scope);
1500 else // there was no strong to remove!, so consider this an opening strong tag
1502 str_append(output, strong_start, sizeof(strong_start) - 1);
1503 ary_push(parser->scope, STRONG);
1504 ary_push(parser->line, STRONG);
1507 else // no strong or em to remove, so this must be a new opening of both
1509 _Wikitext_start_para_if_necessary(parser);
1510 str_append(output, strong_em_start, sizeof(strong_em_start) - 1);
1511 ary_push(parser->scope, STRONG);
1512 ary_push(parser->line, STRONG);
1513 ary_push(parser->scope, EM);
1514 ary_push(parser->line, EM);
1519 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1521 _Wikitext_emit_pending_crlf_if_necessary(parser);
1522 str_append(parser->output, literal_strong, sizeof(literal_strong) - 1);
1526 output = parser->capture ? parser->capture : parser->output;
1527 if (IN(STRONG_START))
1528 // already in span started with <strong>, no choice but to emit this literally
1529 str_append(output, literal_strong, sizeof(literal_strong) - 1);
1530 else if (IN(STRONG))
1531 // STRONG already seen, this is a closing tag
1532 _Wikitext_pop_from_stack_up_to(parser, output, STRONG, true);
1535 // this is a new opening
1536 _Wikitext_pop_excess_elements(parser);
1537 _Wikitext_start_para_if_necessary(parser);
1538 str_append(output, strong_start, sizeof(strong_start) - 1);
1539 ary_push(parser->scope, STRONG);
1540 ary_push(parser->line, STRONG);
1546 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1548 _Wikitext_emit_pending_crlf_if_necessary(parser);
1549 str_append(parser->output, escaped_strong_start, sizeof(escaped_strong_start) - 1);
1553 output = parser->capture ? parser->capture : parser->output;
1554 if (IN(STRONG_START) || IN(STRONG))
1555 str_append(output, escaped_strong_start, sizeof(escaped_strong_start) - 1);
1558 _Wikitext_pop_excess_elements(parser);
1559 _Wikitext_start_para_if_necessary(parser);
1560 str_append(output, strong_start, sizeof(strong_start) - 1);
1561 ary_push(parser->scope, STRONG_START);
1562 ary_push(parser->line, STRONG_START);
1568 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1570 _Wikitext_emit_pending_crlf_if_necessary(parser);
1571 str_append(parser->output, escaped_strong_end, sizeof(escaped_strong_end) - 1);
1575 output = parser->capture ? parser->capture : parser->output;
1576 if (IN(STRONG_START))
1577 _Wikitext_pop_from_stack_up_to(parser, output, STRONG_START, true);
1580 // no STRONG_START in scope, so must interpret the STRONG_END without any special meaning
1581 _Wikitext_pop_excess_elements(parser);
1582 _Wikitext_start_para_if_necessary(parser);
1583 str_append(output, escaped_strong_end, sizeof(escaped_strong_end) - 1);
1589 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1591 _Wikitext_emit_pending_crlf_if_necessary(parser);
1592 str_append(parser->output, literal_em, sizeof(literal_em) - 1);
1596 output = parser->capture ? parser->capture : parser->output;
1598 // already in span started with <em>, no choice but to emit this literally
1599 str_append(output, literal_em, sizeof(literal_em) - 1);
1601 // EM already seen, this is a closing tag
1602 _Wikitext_pop_from_stack_up_to(parser, output, EM, true);
1605 // this is a new opening
1606 _Wikitext_pop_excess_elements(parser);
1607 _Wikitext_start_para_if_necessary(parser);
1608 str_append(output, em_start, sizeof(em_start) - 1);
1609 ary_push(parser->scope, EM);
1610 ary_push(parser->line, EM);
1616 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1618 _Wikitext_emit_pending_crlf_if_necessary(parser);
1619 str_append(parser->output, escaped_em_start, sizeof(escaped_em_start) - 1);
1623 output = parser->capture ? parser->capture : parser->output;
1624 if (IN(EM_START) || IN(EM))
1625 str_append(output, escaped_em_start, sizeof(escaped_em_start) - 1);
1628 _Wikitext_pop_excess_elements(parser);
1629 _Wikitext_start_para_if_necessary(parser);
1630 str_append(output, em_start, sizeof(em_start) - 1);
1631 ary_push(parser->scope, EM_START);
1632 ary_push(parser->line, EM_START);
1638 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1640 _Wikitext_emit_pending_crlf_if_necessary(parser);
1641 str_append(parser->output, escaped_em_end, sizeof(escaped_em_end) - 1);
1645 output = parser->capture ? parser->capture : parser->output;
1647 _Wikitext_pop_from_stack_up_to(parser, output, EM_START, true);
1650 // no EM_START in scope, so must interpret the TT_END without any special meaning
1651 _Wikitext_pop_excess_elements(parser);
1652 _Wikitext_start_para_if_necessary(parser);
1653 str_append(output, escaped_em_end, sizeof(escaped_em_end) - 1);
1659 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1661 _Wikitext_emit_pending_crlf_if_necessary(parser);
1662 str_append(parser->output, backtick, sizeof(backtick) - 1);
1666 output = parser->capture ? parser->capture : parser->output;
1668 // already in span started with <tt>, no choice but to emit this literally
1669 str_append(output, backtick, sizeof(backtick) - 1);
1671 // TT (`) already seen, this is a closing tag
1672 _Wikitext_pop_from_stack_up_to(parser, output, TT, true);
1675 // this is a new opening
1676 _Wikitext_pop_excess_elements(parser);
1677 _Wikitext_start_para_if_necessary(parser);
1678 str_append(output, tt_start, sizeof(tt_start) - 1);
1679 ary_push(parser->scope, TT);
1680 ary_push(parser->line, TT);
1686 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1688 _Wikitext_emit_pending_crlf_if_necessary(parser);
1689 str_append(parser->output, escaped_tt_start, sizeof(escaped_tt_start) - 1);
1693 output = parser->capture ? parser->capture : parser->output;
1694 if (IN(TT_START) || IN(TT))
1695 str_append(output, escaped_tt_start, sizeof(escaped_tt_start) - 1);
1698 _Wikitext_pop_excess_elements(parser);
1699 _Wikitext_start_para_if_necessary(parser);
1700 str_append(output, tt_start, sizeof(tt_start) - 1);
1701 ary_push(parser->scope, TT_START);
1702 ary_push(parser->line, TT_START);
1708 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1710 _Wikitext_emit_pending_crlf_if_necessary(parser);
1711 str_append(parser->output, escaped_tt_end, sizeof(escaped_tt_end) - 1);
1715 output = parser->capture ? parser->capture : parser->output;
1717 _Wikitext_pop_from_stack_up_to(parser, output, TT_START, true);
1720 // no TT_START in scope, so must interpret the TT_END without any special meaning
1721 _Wikitext_pop_excess_elements(parser);
1722 _Wikitext_start_para_if_necessary(parser);
1723 str_append(output, escaped_tt_end, sizeof(escaped_tt_end) - 1);
1730 if (IN(NO_WIKI_START) || IN(PRE_START))
1732 // no need to check for PRE; can never appear inside it
1733 str_append(parser->output, token->start, TOKEN_LEN(token));
1737 // count number of tokens in line and scope stacks
1738 int bq_count = ary_count(parser->scope, BLOCKQUOTE_START);
1739 i = parser->line->count - ary_count(parser->line, BLOCKQUOTE_START);
1740 j = parser->scope->count - bq_count;
1743 // list tokens can be nested so look ahead for any more which might affect the decision to push or pop
1747 if (type == OL || type == UL)
1750 if (i - k >= 2) // already seen at least one OL or UL
1752 ary_push(parser->line, NESTED_LIST); // which means this is a nested list
1757 ary_push(parser->line, type);
1758 ary_push(parser->line, LI);
1760 // want to compare line with scope but can only do so if scope has enough items on it
1763 if (ary_entry(parser->scope, i + bq_count - 2) == type &&
1764 ary_entry(parser->scope, i + bq_count - 1) == LI)
1766 // line and scope match at this point: do nothing yet
1770 // item just pushed onto line does not match corresponding slot of scope!
1771 for (; j >= i - 2; j--)
1772 // must pop back before emitting
1773 _Wikitext_pop_from_stack(parser, NULL);
1775 // will emit UL or OL, then LI
1779 else // line stack size now exceeds scope stack size: must increase nesting level
1780 break; // will emit UL or OL, then LI
1784 // not a OL or UL token!
1786 // must close existing LI and re-open new one
1787 _Wikitext_pop_from_stack(parser, NULL);
1790 // item just pushed onto line does not match corresponding slot of scope!
1792 // must pop back before emitting
1793 _Wikitext_pop_from_stack(parser, NULL);
1801 if (type == OL || type == UL)
1803 // if LI is at the top of a stack this is the start of a nested list
1804 if (j > 0 && ary_entry(parser->scope, -1) == LI)
1806 // so we should precede it with a CRLF, and indicate that it's a nested list
1807 str_append(parser->output, parser->line_ending->ptr, parser->line_ending->len);
1808 ary_push(parser->scope, NESTED_LIST);
1812 // this is a new list
1813 if (IN(BLOCKQUOTE_START))
1814 _Wikitext_pop_from_stack_up_to(parser, NULL, BLOCKQUOTE_START, false);
1816 _Wikitext_pop_from_stack_up_to(parser, NULL, BLOCKQUOTE, false);
1820 _Wikitext_indent(parser);
1822 str_append(parser->output, ol_start, sizeof(ol_start) - 1);
1823 else if (type == UL)
1824 str_append(parser->output, ul_start, sizeof(ul_start) - 1);
1825 ary_push(parser->scope, type);
1826 str_append(parser->output, parser->line_ending->ptr, parser->line_ending->len);
1828 else if (type == SPACE)
1829 // silently throw away the optional SPACE token after final list marker
1832 _Wikitext_indent(parser);
1833 str_append(parser->output, li_start, sizeof(li_start) - 1);
1834 ary_push(parser->scope, LI);
1836 // any subsequent UL or OL tokens on this line are syntax errors and must be emitted literally
1837 if (type == OL || type == UL)
1840 while (k++, NEXT_TOKEN(), (type = token->type))
1842 if (type == OL || type == UL)
1843 str_append(parser->output, token->start, TOKEN_LEN(token));
1844 else if (type == SPACE && k == 1)
1846 // silently throw away the optional SPACE token after final list marker
1855 // jump to top of the loop to process token we scanned during lookahead
1864 if (IN(NO_WIKI_START) || IN(PRE_START))
1866 // no need to check for PRE; can never appear inside it
1867 str_append(parser->output, token->start, TOKEN_LEN(token));
1871 // pop up to but not including the last BLOCKQUOTE on the scope stack
1872 if (IN(BLOCKQUOTE_START))
1873 _Wikitext_pop_from_stack_up_to(parser, NULL, BLOCKQUOTE_START, false);
1875 _Wikitext_pop_from_stack_up_to(parser, NULL, BLOCKQUOTE, false);
1877 // count number of BLOCKQUOTE tokens in line buffer and in scope stack
1878 ary_push(parser->line, type);
1879 i = ary_count(parser->line, BLOCKQUOTE);
1880 j = ary_count(parser->scope, BLOCKQUOTE);
1882 // decide whether we need to pop off excess BLOCKQUOTE tokens (will never need to push; that is handled above in the BLOCKQUOTE case itself)
1885 // must pop (reduce nesting level)
1886 for (i = j - i; i > 0; i--)
1887 _Wikitext_pop_from_stack_up_to(parser, NULL, BLOCKQUOTE, true);
1890 // discard any whitespace here (so that "== foo ==" will be translated to "<h2>foo</h2>" rather than "<h2> foo </h2")
1891 while (NEXT_TOKEN(), (token->type == SPACE))
1894 ary_push(parser->scope, type);
1895 _Wikitext_indent(parser);
1897 // take base_heading_level into account
1898 type += base_heading_level;
1899 if (type > H6_START) // no need to check for underflow (base_heading_level never negative)
1902 // rather than repeat all that code for each kind of heading, share it and use a conditional here
1903 if (type == H6_START)
1904 str_append(parser->output, h6_start, sizeof(h6_start) - 1);
1905 else if (type == H5_START)
1906 str_append(parser->output, h5_start, sizeof(h5_start) - 1);
1907 else if (type == H4_START)
1908 str_append(parser->output, h4_start, sizeof(h4_start) - 1);
1909 else if (type == H3_START)
1910 str_append(parser->output, h3_start, sizeof(h3_start) - 1);
1911 else if (type == H2_START)
1912 str_append(parser->output, h2_start, sizeof(h2_start) - 1);
1913 else if (type == H1_START)
1914 str_append(parser->output, h1_start, sizeof(h1_start) - 1);
1916 // jump to top of the loop to process token we scanned during lookahead
1920 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1922 _Wikitext_emit_pending_crlf_if_necessary(parser);
1923 str_append(parser->output, literal_h6, sizeof(literal_h6) - 1);
1927 _Wikitext_rollback_failed_external_link(parser); // if any
1930 // literal output only if not in h6 scope (we stay silent in that case)
1931 _Wikitext_start_para_if_necessary(parser);
1932 str_append(parser->output, literal_h6, sizeof(literal_h6) - 1);
1938 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1940 _Wikitext_emit_pending_crlf_if_necessary(parser);
1941 str_append(parser->output, literal_h5, sizeof(literal_h5) - 1);
1945 _Wikitext_rollback_failed_external_link(parser); // if any
1948 // literal output only if not in h5 scope (we stay silent in that case)
1949 _Wikitext_start_para_if_necessary(parser);
1950 str_append(parser->output, literal_h5, sizeof(literal_h5) - 1);
1956 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1958 _Wikitext_emit_pending_crlf_if_necessary(parser);
1959 str_append(parser->output, literal_h4, sizeof(literal_h4) - 1);
1963 _Wikitext_rollback_failed_external_link(parser); // if any
1966 // literal output only if not in h4 scope (we stay silent in that case)
1967 _Wikitext_start_para_if_necessary(parser);
1968 str_append(parser->output, literal_h4, sizeof(literal_h4) - 1);
1974 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1976 _Wikitext_emit_pending_crlf_if_necessary(parser);
1977 str_append(parser->output, literal_h3, sizeof(literal_h3) - 1);
1981 _Wikitext_rollback_failed_external_link(parser); // if any
1984 // literal output only if not in h3 scope (we stay silent in that case)
1985 _Wikitext_start_para_if_necessary(parser);
1986 str_append(parser->output, literal_h3, sizeof(literal_h3) - 1);
1992 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
1994 _Wikitext_emit_pending_crlf_if_necessary(parser);
1995 str_append(parser->output, literal_h2, sizeof(literal_h2) - 1);
1999 _Wikitext_rollback_failed_external_link(parser); // if any
2002 // literal output only if not in h2 scope (we stay silent in that case)
2003 _Wikitext_start_para_if_necessary(parser);
2004 str_append(parser->output, literal_h2, sizeof(literal_h2) - 1);
2010 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
2012 _Wikitext_emit_pending_crlf_if_necessary(parser);
2013 str_append(parser->output, literal_h1, sizeof(literal_h1) - 1);
2017 _Wikitext_rollback_failed_external_link(parser); // if any
2020 // literal output only if not in h1 scope (we stay silent in that case)
2021 _Wikitext_start_para_if_necessary(parser);
2022 str_append(parser->output, literal_h1, sizeof(literal_h1) - 1);
2028 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
2030 _Wikitext_emit_pending_crlf_if_necessary(parser);
2031 str_append(parser->output, token->start, TOKEN_LEN(token));
2035 _Wikitext_pop_excess_elements(parser);
2036 _Wikitext_start_para_if_necessary(parser);
2037 _Wikitext_append_hyperlink(parser, rb_str_new2("mailto:"), TOKEN_TEXT(token), Qnil, mailto_class, true);
2042 if (IN(NO_WIKI_START))
2043 // user can temporarily suppress autolinking by using <nowiki></nowiki>
2044 // note that unlike MediaWiki, we do allow autolinking inside PRE blocks
2045 str_append(parser->output, token->start, TOKEN_LEN(token));
2046 else if (IN(LINK_START))
2048 // if the URI were allowed it would have been handled already in LINK_START
2049 _Wikitext_rollback_failed_internal_link(parser);
2050 _Wikitext_append_hyperlink(parser, Qnil, TOKEN_TEXT(token), Qnil, parser->external_link_class, true);
2052 else if (IN(EXT_LINK_START))
2054 if (parser->link_target->len == 0)
2056 // this must be our link target: look ahead to make sure we see the space we're expecting to see
2057 i = TOKEN_TEXT(token);
2059 if (token->type == SPACE)
2061 ary_push(parser->scope, SPACE);
2062 str_append_string(parser->link_target, i);
2063 str_clear(parser->link_text);
2064 parser->capture = parser->link_text;
2065 token = NULL; // silently consume space
2069 // didn't see the space! this must be an error
2070 _Wikitext_pop_from_stack(parser, NULL);
2071 _Wikitext_pop_excess_elements(parser);
2072 _Wikitext_start_para_if_necessary(parser);
2073 str_append(parser->output, ext_link_start, sizeof(ext_link_start) - 1);
2074 _Wikitext_append_hyperlink(parser, Qnil, i, Qnil, parser->external_link_class, true);
2078 str_append(parser->link_text, token->start, TOKEN_LEN(token));
2082 _Wikitext_pop_excess_elements(parser);
2083 _Wikitext_start_para_if_necessary(parser);
2084 _Wikitext_append_hyperlink(parser, Qnil, TOKEN_TEXT(token), Qnil, parser->external_link_class, true);
2089 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
2090 str_append(parser->output, token->start, TOKEN_LEN(token));
2091 else if (IN(EXT_LINK_START))
2093 if (parser->link_target->len == 0)
2095 // this must be our link target: look ahead to make sure we see the space we're expecting to see
2096 i = TOKEN_TEXT(token);
2098 if (token->type == SPACE)
2100 ary_push(parser->scope, PATH);
2101 ary_push(parser->scope, SPACE);
2102 str_append_string(parser->link_target, i);
2103 str_clear(parser->link_text);
2104 parser->capture = parser->link_text;
2105 token = NULL; // silently consume space
2109 // didn't see the space! this must be an error
2110 _Wikitext_pop_from_stack(parser, NULL);
2111 _Wikitext_pop_excess_elements(parser);
2112 _Wikitext_start_para_if_necessary(parser);
2113 str_append(parser->output, ext_link_start, sizeof(ext_link_start) - 1);
2114 str_append_string(parser->output, i);
2118 str_append(parser->link_text, token->start, TOKEN_LEN(token));
2122 output = parser->capture ? parser->capture : parser->output;
2123 _Wikitext_pop_excess_elements(parser);
2124 _Wikitext_start_para_if_necessary(parser);
2125 str_append(output, token->start, TOKEN_LEN(token));
2129 // internal links (links to other wiki articles) look like this:
2130 // [[another article]] (would point at, for example, "/wiki/another_article")
2131 // [[the other article|the link text we'll use for it]]
2132 // [[the other article | the link text we'll use for it]]
2133 // MediaWiki has strict requirements about what it will accept as a link target:
2134 // all wikitext markup is disallowed:
2135 // example [[foo ''bar'' baz]]
2136 // renders [[foo <em>bar</em> baz]] (ie. not a link)
2137 // example [[foo <em>bar</em> baz]]
2138 // renders [[foo <em>bar</em> baz]] (ie. not a link)
2139 // example [[foo <nowiki>''</nowiki> baz]]
2140 // renders [[foo '' baz]] (ie. not a link)
2141 // example [[foo <bar> baz]]
2142 // renders [[foo <bar> baz]] (ie. not a link)
2143 // HTML entities and non-ASCII, however, make it through:
2144 // example [[foo €]]
2145 // renders <a href="/wiki/Foo_%E2%82%AC">foo €</a>
2146 // example [[foo €]]
2147 // renders <a href="/wiki/Foo_%E2%82%AC">foo €</a>
2148 // we'll impose similar restrictions here for the link target; allowed tokens will be:
2149 // SPACE, SPECIAL_URI_CHARS, PRINTABLE, PATH, ALNUM, DEFAULT, QUOT and AMP
2150 // everything else will be rejected
2152 output = parser->capture ? parser->capture : parser->output;
2153 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
2155 _Wikitext_emit_pending_crlf_if_necessary(parser);
2156 str_append(output, link_start, sizeof(link_start) - 1);
2158 else if (IN(EXT_LINK_START))
2159 // already in external link scope! (and in fact, must be capturing link_text right now)
2160 str_append(output, link_start, sizeof(link_start) - 1);
2161 else if (IN(LINK_START))
2163 // already in internal link scope! this is a syntax error
2164 _Wikitext_rollback_failed_internal_link(parser);
2165 str_append(parser->output, link_start, sizeof(link_start) - 1);
2167 else if (IN(SEPARATOR))
2169 // scanning internal link text
2171 else // not in internal link scope yet
2173 // will either emit a link, or the rollback of a failed link, so start the para now
2174 _Wikitext_pop_excess_elements(parser);
2175 _Wikitext_start_para_if_necessary(parser);
2176 ary_push(parser->scope, LINK_START);
2178 // look ahead and try to gobble up link target
2179 while (NEXT_TOKEN(), (type = token->type))
2181 if (type == SPACE ||
2182 type == SPECIAL_URI_CHARS ||
2184 type == PRINTABLE ||
2188 type == QUOT_ENTITY ||
2190 type == AMP_ENTITY ||
2191 type == IMG_START ||
2193 type == LEFT_CURLY ||
2194 type == RIGHT_CURLY)
2196 // accumulate these tokens into link_target
2197 if (parser->link_target->len == 0)
2199 str_clear(parser->link_target);
2200 parser->capture = parser->link_target;
2202 if (type == QUOT_ENTITY)
2203 // don't insert the entity, insert the literal quote
2204 str_append(parser->link_target, quote, sizeof(quote) - 1);
2205 else if (type == AMP_ENTITY)
2206 // don't insert the entity, insert the literal ampersand
2207 str_append(parser->link_target, ampersand, sizeof(ampersand) - 1);
2209 str_append(parser->link_target, token->start, TOKEN_LEN(token));
2211 else if (type == LINK_END)
2213 if (parser->link_target->len == 0) // bail for inputs like "[[]]"
2214 _Wikitext_rollback_failed_internal_link(parser);
2215 break; // jump back to top of loop (will handle this in LINK_END case below)
2217 else if (type == SEPARATOR)
2219 if (parser->link_target->len == 0) // bail for inputs like "[[|"
2220 _Wikitext_rollback_failed_internal_link(parser);
2223 ary_push(parser->scope, SEPARATOR);
2224 str_clear(parser->link_text);
2225 parser->capture = parser->link_text;
2230 else // unexpected token (syntax error)
2232 _Wikitext_rollback_failed_internal_link(parser);
2233 break; // jump back to top of loop to handle unexpected token
2237 // jump to top of the loop to process token we scanned during lookahead (if any)
2243 output = parser->capture ? parser->capture : parser->output;
2244 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
2246 _Wikitext_emit_pending_crlf_if_necessary(parser);
2247 str_append(output, link_end, sizeof(link_end) - 1);
2249 else if (IN(EXT_LINK_START))
2250 // already in external link scope! (and in fact, must be capturing link_text right now)
2251 str_append(output, link_end, sizeof(link_end) - 1);
2252 else if (IN(LINK_START)) // in internal link scope!
2254 if (_Wikitext_blank(parser->link_target))
2256 // special case for inputs like "[[ ]]"
2257 _Wikitext_rollback_failed_internal_link(parser);
2258 str_append(parser->output, link_end, sizeof(link_end) - 1);
2261 if (parser->link_text->len == 0 ||
2262 _Wikitext_blank(parser->link_text))
2264 // use link target as link text
2265 str_clear(parser->link_text);
2266 _Wikitext_parser_append_sanitized_link_target(parser, parser->link_text, false);
2269 _Wikitext_parser_trim_link_text(parser);
2270 _Wikitext_parser_encode_link_target(parser);
2271 _Wikitext_pop_from_stack_up_to(parser, output, LINK_START, true);
2272 parser->capture = NULL;
2273 _Wikitext_append_hyperlink(parser, prefix, string_from_str(parser->link_target), string_from_str(parser->link_text), Qnil, false);
2274 str_clear(parser->link_target);
2275 str_clear(parser->link_text);
2277 else // wasn't in internal link scope
2279 _Wikitext_pop_excess_elements(parser);
2280 _Wikitext_start_para_if_necessary(parser);
2281 str_append(output, link_end, sizeof(link_end) - 1);
2285 // external links look like this:
2286 // [http://google.com/ the link text]
2287 // [/other/page/on/site see this page]
2288 // strings in square brackets which don't match this syntax get passed through literally; eg:
2289 // he was very angery [sic] about the turn of events
2290 case EXT_LINK_START:
2291 output = parser->capture ? parser->capture : parser->output;
2292 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
2294 _Wikitext_emit_pending_crlf_if_necessary(parser);
2295 str_append(output, ext_link_start, sizeof(ext_link_start) - 1);
2297 else if (IN(EXT_LINK_START))
2298 // already in external link scope! (and in fact, must be capturing link_text right now)
2299 str_append(output, ext_link_start, sizeof(ext_link_start) - 1);
2300 else if (IN(LINK_START))
2302 // already in internal link scope!
2303 if (parser->link_target->len == 0)
2304 // this must be the first character of our link target
2305 str_append(parser->link_target, ext_link_start, sizeof(ext_link_start) - 1);
2307 // link target has already been scanned
2308 str_append(parser->link_text, ext_link_start, sizeof(ext_link_start) - 1);
2310 // TODO: possibly roll this condition into the above
2311 // add to existing link target
2312 str_append(parser->link_target, ext_link_start, sizeof(ext_link_start) - 1);
2314 else // not in external link scope yet
2316 // will either emit a link, or the rollback of a failed link, so start the para now
2317 _Wikitext_pop_excess_elements(parser);
2318 _Wikitext_start_para_if_necessary(parser);
2320 // look ahead: expect an absolute URI (with protocol) or "relative" (path) URI
2322 if (token->type == URI || token->type == PATH)
2323 ary_push(parser->scope, EXT_LINK_START); // so far so good, jump back to the top of the loop
2325 // only get here if there was a syntax error (missing URI)
2326 str_append(parser->output, ext_link_start, sizeof(ext_link_start) - 1);
2327 continue; // jump back to top of loop to handle token (either URI or whatever it is)
2332 output = parser->capture ? parser->capture : parser->output;
2333 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
2335 _Wikitext_emit_pending_crlf_if_necessary(parser);
2336 str_append(output, ext_link_end, sizeof(ext_link_end) - 1);
2338 else if (IN(EXT_LINK_START))
2340 if (parser->link_text->len == 0)
2341 // syntax error: external link with no link text
2342 _Wikitext_rollback_failed_external_link(parser);
2346 j = IN(PATH) ? Qnil : parser->external_link_class;
2347 _Wikitext_pop_from_stack_up_to(parser, output, EXT_LINK_START, true);
2348 parser->capture = NULL;
2349 _Wikitext_append_hyperlink(parser, Qnil, string_from_str(parser->link_target), string_from_str(parser->link_text), j, false);
2351 str_clear(parser->link_target);
2352 str_clear(parser->link_text);
2356 _Wikitext_pop_excess_elements(parser);
2357 _Wikitext_start_para_if_necessary(parser);
2358 str_append(parser->output, ext_link_end, sizeof(ext_link_end) - 1);
2363 output = parser->capture ? parser->capture : parser->output;
2364 _Wikitext_pop_excess_elements(parser);
2365 _Wikitext_start_para_if_necessary(parser);
2366 str_append(output, separator, sizeof(separator) - 1);
2370 output = parser->capture ? parser->capture : parser->output;
2371 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
2373 _Wikitext_emit_pending_crlf_if_necessary(parser);
2374 str_append(output, token->start, TOKEN_LEN(token));
2378 // peek ahead to see next token
2379 char *token_ptr = token->start;
2380 int token_len = TOKEN_LEN(token);
2383 if (((type == H6_END) && IN(H6_START)) ||
2384 ((type == H5_END) && IN(H5_START)) ||
2385 ((type == H4_END) && IN(H4_START)) ||
2386 ((type == H3_END) && IN(H3_START)) ||
2387 ((type == H2_END) && IN(H2_START)) ||
2388 ((type == H1_END) && IN(H1_START)))
2390 // will suppress emission of space (discard) if next token is a H6_END, H5_END etc and we are in the corresponding scope
2395 _Wikitext_pop_excess_elements(parser);
2396 _Wikitext_start_para_if_necessary(parser);
2397 str_append(output, token_ptr, token_len);
2400 // jump to top of the loop to process token we scanned during lookahead
2408 case DECIMAL_ENTITY:
2409 // pass these through unaltered as they are case sensitive
2410 output = parser->capture ? parser->capture : parser->output;
2411 _Wikitext_pop_excess_elements(parser);
2412 _Wikitext_start_para_if_necessary(parser);
2413 str_append(output, token->start, TOKEN_LEN(token));
2417 // normalize hex entities (downcase them)
2418 output = parser->capture ? parser->capture : parser->output;
2419 _Wikitext_pop_excess_elements(parser);
2420 _Wikitext_start_para_if_necessary(parser);
2421 str_append(output, token->start, TOKEN_LEN(token));
2422 _Wikitext_downcase_bang(output->ptr + output->len - TOKEN_LEN(token), TOKEN_LEN(token));
2426 output = parser->capture ? parser->capture : parser->output;
2427 _Wikitext_pop_excess_elements(parser);
2428 _Wikitext_start_para_if_necessary(parser);
2429 str_append(output, quot_entity, sizeof(quot_entity) - 1);
2433 output = parser->capture ? parser->capture : parser->output;
2434 _Wikitext_pop_excess_elements(parser);
2435 _Wikitext_start_para_if_necessary(parser);
2436 str_append(output, amp_entity, sizeof(amp_entity) - 1);
2440 output = parser->capture ? parser->capture : parser->output;
2441 _Wikitext_pop_excess_elements(parser);
2442 _Wikitext_start_para_if_necessary(parser);
2443 str_append(output, lt_entity, sizeof(lt_entity) - 1);
2447 output = parser->capture ? parser->capture : parser->output;
2448 _Wikitext_pop_excess_elements(parser);
2449 _Wikitext_start_para_if_necessary(parser);
2450 str_append(output, gt_entity, sizeof(gt_entity) - 1);
2454 if (IN(NO_WIKI_START) || IN(PRE) || IN(PRE_START))
2456 _Wikitext_emit_pending_crlf_if_necessary(parser);
2457 str_append(parser->output, token->start, TOKEN_LEN(token));
2459 else if (parser->capture)
2460 str_append(parser->capture, token->start, TOKEN_LEN(token));
2463 // not currently capturing: will be emitting something on success or failure, so get ready
2464 _Wikitext_pop_excess_elements(parser);
2465 _Wikitext_start_para_if_necessary(parser);
2467 // scan ahead consuming PATH, PRINTABLE, ALNUM and SPECIAL_URI_CHARS tokens
2468 // will cheat here and abuse the link_target capture buffer to accumulate text
2469 while (NEXT_TOKEN(), (type = token->type))
2471 if (type == PATH || type == PRINTABLE || type == ALNUM || type == SPECIAL_URI_CHARS)
2472 str_append(parser->link_target, token->start, TOKEN_LEN(token));
2473 else if (type == IMG_END && parser->link_target->len > 0)
2476 _Wikitext_append_img(parser, parser->link_target->ptr, parser->link_target->len);
2480 else // unexpected token or zero-length target (syntax error)
2483 str_append(parser->output, literal_img_start, sizeof(literal_img_start) - 1);
2484 if (parser->link_target->len > 0)
2485 str_append(parser->output, parser->link_target->ptr, parser->link_target->len);
2490 // jump to top of the loop to process token we scanned during lookahead
2491 str_clear(parser->link_target);
2497 i = parser->pending_crlf;
2498 parser->pending_crlf = false;
2499 _Wikitext_rollback_failed_link(parser); // if any
2500 if (IN(NO_WIKI_START) || IN(PRE_START))
2502 ary_clear(parser->line_buffer);
2503 str_append_str(parser->output, parser->line_ending);
2508 // beware when BLOCKQUOTE on line buffer (not line stack!) prior to CRLF, that must be end of PRE block
2509 if (ary_entry(parser->line_buffer, -2) == BLOCKQUOTE)
2510 // don't emit in this case
2511 _Wikitext_pop_from_stack_up_to(parser, parser->output, PRE, true);
2514 if (ary_entry(parser->line_buffer, -2) == PRE)
2516 // only thing on line is the PRE: emit pending line ending (if we had one)
2518 str_append_str(parser->output, parser->line_ending);
2521 // clear these _before_ calling NEXT_TOKEN (NEXT_TOKEN adds to the line_buffer)
2522 ary_clear(parser->line);
2523 ary_clear(parser->line_buffer);
2525 // peek ahead to see if this is definitely the end of the PRE block
2528 if (type != BLOCKQUOTE && type != PRE)
2529 // this is definitely the end of the block, so don't emit
2530 _Wikitext_pop_from_stack_up_to(parser, parser->output, PRE, true);
2532 // potentially will emit
2533 parser->pending_crlf = true;
2535 continue; // jump back to top of loop to handle token grabbed via lookahead
2540 parser->pending_crlf = true;
2542 // count number of BLOCKQUOTE tokens in line buffer (can be zero) and pop back to that level
2543 // as a side effect, this handles any open span-level elements and unclosed blocks
2544 // (with special handling for P blocks and LI elements)
2545 i = ary_count(parser->line, BLOCKQUOTE) + ary_count(parser->scope, BLOCKQUOTE_START);
2546 for (j = parser->scope->count; j > i; j--)
2548 if (parser->scope->count > 0 && ary_entry(parser->scope, -1) == LI)
2550 parser->pending_crlf = false;
2554 // special handling on last iteration through the loop if the top item on the scope is a P block
2555 if ((j - i == 1) && ary_entry(parser->scope, -1) == P)
2557 // if nothing or BLOCKQUOTE on line buffer (not line stack!) prior to CRLF, this must be a paragraph break
2558 // (note that we have to make sure we're not inside a BLOCKQUOTE_START block
2559 // because in those blocks BLOCKQUOTE tokens have no special meaning)
2560 if (NO_ITEM(ary_entry(parser->line_buffer, -2)) ||
2561 (ary_entry(parser->line_buffer, -2) == BLOCKQUOTE && !IN(BLOCKQUOTE_START)))
2563 parser->pending_crlf = false;
2565 // not a paragraph break!
2568 _Wikitext_pop_from_stack(parser, NULL);
2572 // delete the entire contents of the line scope stack and buffer
2573 ary_clear(parser->line);
2574 ary_clear(parser->line_buffer);
2577 case SPECIAL_URI_CHARS:
2583 output = parser->capture ? parser->capture : parser->output;
2584 _Wikitext_pop_excess_elements(parser);
2585 _Wikitext_start_para_if_necessary(parser);
2586 str_append(output, token->start, TOKEN_LEN(token));
2590 output = parser->capture ? parser->capture : parser->output;
2591 _Wikitext_pop_excess_elements(parser);
2592 _Wikitext_start_para_if_necessary(parser);
2593 //str_append_string(output, _Wikitext_utf32_char_to_entity(token->code_point)); // convert to entity
2594 // TODO: replace this with a different append function
2595 str_grow(output, output->len + 8);
2596 _Wikitext_append_entity_from_utf32_char(output->ptr + output->len, token->code_point);
2597 output->len = output->len + 8;
2601 // special case for input like " foo\n " (see pre_spec.rb)
2603 ary_entry(parser->line_buffer, -2) == PRE &&
2604 parser->pending_crlf)
2605 str_append(parser->output, parser->line_ending->ptr, parser->line_ending->len);
2607 // close any open scopes on hitting EOF
2608 _Wikitext_rollback_failed_link(parser); // if any
2609 _Wikitext_pop_all_from_stack(parser);
2610 goto return_output; // break not enough here (want to break out of outer while loop, not inner switch statement)
2616 // reset current token; forcing lexer to return another token at the top of the loop
2620 // nasty hack to avoid re-allocating our return value
2621 str_append(parser->output, null_str, 1); // null-terminate
2623 // won't work for Ruby 1.9
2624 free(RSTRING(out)->ptr);
2625 RSTRING(out)->ptr = parser->output->ptr;
2626 RSTRING(out)->len = parser->output->len - 1; // don't count null termination
2627 parser->output->ptr = NULL; // don't double-free