stdex
Additional custom or not Standard C++ covered algorithms
Loading...
Searching...
No Matches
sgml.hpp
1/*
2 SPDX-License-Identifier: MIT
3 Copyright © 2023 Amebis
4*/
5
6#pragma once
7
8#include "mapping.hpp"
9#include "sal.hpp"
10#include "sgml_unicode.hpp"
11#include "string.hpp"
12#include <assert.h>
13#include <string>
14
15namespace stdex
16{
18 template <class T>
19 inline const wchar_t* sgml2uni(_In_reads_or_z_(count) const T* entity, _In_ size_t count)
20 {
21 assert(entity && count);
22 assert(count < 2 || entity[0] != '#'); // No numeric entities
23
24 for (size_t i = 0, j = _countof(sgml_unicode); i < j; ) {
25 size_t m = (i + j) / 2;
26 if (sgml_unicode[m].sgml[0] < entity[0])
27 i = m + 1;
28 else if (sgml_unicode[m].sgml[0] > entity[0])
29 j = m;
30 else {
31 auto r = strncmp<char, T>(sgml_unicode[m].sgml + 1, _countof(sgml_unicode[0].sgml) - 1, entity + 1, count - 1);
32 if (r < 0)
33 i = m + 1;
34 else if (r > 0)
35 j = m;
36 else {
37 for (; i < m && strncmp<char, T>(sgml_unicode[m - 1].sgml, _countof(sgml_unicode[0].sgml), entity, count) == 0; m--);
38 return sgml_unicode[m].unicode;
39 }
40 }
41 }
42 return nullptr;
43 }
44
45 template <class T>
46 inline const T* sgmlend(
47 _In_reads_or_z_opt_(count) const T* str,
48 _In_ size_t count)
49 {
50 assert(str || !count);
51 for (size_t i = 0; i < count; i++) {
52 if (str[i] == ';')
53 return str + i;
54 if (!str[i] || str[i] == '&' || isspace(str[i]))
55 break;
56 }
57 return nullptr;
58 }
60
61 constexpr int sgml_full = 0x80000000;
62 constexpr int sgml_quot = 0x00000001;
63 constexpr int sgml_apos = 0x00000002;
64 constexpr int sgml_quot_apos = sgml_quot | sgml_apos;
65 constexpr int sgml_amp = 0x00000004;
66 constexpr int sgml_lt_gt = 0x00000008;
67 constexpr int sgml_bsol = 0x00000010;
68 constexpr int sgml_dollar = 0x00000020;
69 constexpr int sgml_percnt = 0x00000040;
70 constexpr int sgml_commat = 0x00000080;
71 constexpr int sgml_num = 0x00000100;
72 constexpr int sgml_lpar_rpar = 0x00000200;
73 constexpr int sgml_lcub_rcub = 0x00000400;
74 constexpr int sgml_lsqb_rsqb = 0x00000800;
75 constexpr int sgml_sgml = sgml_amp | sgml_lt_gt;
76 constexpr int sgml_ml_attrib = sgml_amp | sgml_quot_apos;
77 constexpr int sgml_c = sgml_amp | sgml_bsol | sgml_quot_apos;
78 // constexpr int sgml_ajt_lemma = sgml_amp | sgml_quot | sgml_dollar | sgml_percnt;
79 // constexpr int sgml_ajt_form = sgml_ajt_lemma;
80 // constexpr int sgml_kolos = sgml_amp | sgml_quot | sgml_dollar | sgml_percnt | sgml_lt_gt | sgml_bsol/* | sgml_commat | sgml_num*/ | sgml_lpar_rpar | sgml_lcub_rcub | sgml_lsqb_rsqb;
81
94 template <class T>
95 inline void sgml2wstr(
96 _Inout_ std::wstring& dst,
97 _In_reads_or_z_opt_(count_src) const T* src, _In_ size_t count_src,
98 _In_ int skip = 0,
99 _In_ const mapping<size_t>& offset = mapping<size_t>(0, 0),
100 _Inout_opt_ mapping_vector<size_t>* map = nullptr)
101 {
102 assert(src || !count_src);
103
104 const bool
105 skip_quot = (skip & sgml_quot) == 0,
106 skip_apos = (skip & sgml_apos) == 0,
107 skip_amp = (skip & sgml_amp) == 0,
108 skip_lt_gt = (skip & sgml_lt_gt) == 0,
109 skip_bsol = (skip & sgml_bsol) == 0,
110 skip_dollar = (skip & sgml_dollar) == 0,
111 skip_percnt = (skip & sgml_percnt) == 0,
112 skip_commat = (skip & sgml_commat) == 0,
113 skip_num = (skip & sgml_num) == 0,
114 skip_lpar_rpar = (skip & sgml_lpar_rpar) == 0,
115 skip_lcub_rcub = (skip & sgml_lcub_rcub) == 0,
116 skip_lsqb_rsqb = (skip & sgml_lsqb_rsqb) == 0;
117
118 count_src = strnlen(src, count_src);
119 dst.reserve(dst.size() + count_src);
120 for (size_t i = 0; i < count_src;) {
121 if (src[i] == '&') {
122 auto end = sgmlend(src + i + 1, count_src - i - 1);
123 if (end) {
124 const wchar_t* entity_w;
125 wchar_t chr[3];
126 size_t n = end - src - i - 1;
127 if (n >= 2 && src[i + 1] == '#') {
128 uint32_t unicode;
129 if (src[i + 2] == 'x' || src[i + 2] == 'X')
130 unicode = strtou32(src + i + 3, n - 2, nullptr, 16);
131 else
132 unicode = strtou32(src + i + 2, n - 1, nullptr, 10);
133#ifdef _WIN32
134 if (unicode < 0x10000) {
135 chr[0] = (wchar_t)unicode;
136 chr[1] = 0;
137 }
138 else {
139 ucs4_to_surrogate_pair(chr, unicode);
140 chr[2] = 0;
141 }
142#else
143 chr[0] = (wchar_t)unicode;
144 chr[1] = 0;
145#endif
146 entity_w = chr;
147 }
148 else
149 entity_w = sgml2uni(src + i + 1, n);
150
151 if (entity_w &&
152 (skip_quot || (entity_w[0] != L'"')) &&
153 (skip_apos || (entity_w[0] != L'\'')) &&
154 (skip_amp || (entity_w[0] != L'&')) &&
155 (skip_lt_gt || (entity_w[0] != L'<' && entity_w[0] != L'>')) &&
156 (skip_bsol || (entity_w[0] != L'\\')) &&
157 (skip_dollar || (entity_w[0] != L'$')) &&
158 (skip_percnt || (entity_w[0] != L'%')) &&
159 (skip_commat || (entity_w[0] != L'@')) &&
160 (skip_num || (entity_w[0] != L'#')) &&
161 (skip_lpar_rpar || (entity_w[0] != L'(' && entity_w[0] != L')')) &&
162 (skip_lcub_rcub || (entity_w[0] != L'{' && entity_w[0] != L'}')) &&
163 (skip_lsqb_rsqb || (entity_w[0] != L'[' && entity_w[0] != L']')))
164 {
165 if (map) map->push_back(mapping<size_t>(offset.from + i, offset.to + dst.size()));
166 dst.append(entity_w);
167 i = end - src + 1;
168 if (map) map->push_back(mapping<size_t>(offset.from + i, offset.to + dst.size()));
169 continue;
170 }
171 }
172 }
173 dst.append(1, src[i++]);
174 }
175 }
176
188 template <class T>
189 inline void sgml2wstr(
190 _Inout_ std::wstring& dst,
191 _In_ const std::basic_string<T>& src,
192 _In_ int skip = 0,
193 _In_ const mapping<size_t>& offset = mapping<size_t>(0, 0),
194 _Inout_opt_ mapping_vector<size_t>* map = nullptr)
195 {
196 sgml2wstr(dst, src.data(), src.size(), skip, offset, map);
197 }
198
210 template <class T>
211 inline std::wstring sgml2wstr(
212 _In_reads_or_z_opt_(count_src) const T* src, _In_ size_t count_src,
213 _In_ int skip = 0,
214 _In_ const mapping<size_t>& offset = mapping<size_t>(0, 0),
215 _Inout_opt_ mapping_vector<size_t>* map = nullptr)
216 {
217 std::wstring dst;
218 sgml2wstr(dst, src, count_src, skip, offset, map);
219 return dst;
220 }
221
232 template <class T>
233 inline std::wstring sgml2wstr(
234 _In_ const std::basic_string<T>& src,
235 _In_ int skip = 0,
236 _In_ const mapping<size_t>& offset = mapping<size_t>(0, 0),
237 _Inout_opt_ mapping_vector<size_t>* map = nullptr)
238 {
239 return sgml2wstr(src.c_str(), src.size(), skip, offset, map);
240 }
241
243 inline const char* chr2sgml(_In_reads_or_z_(count) const wchar_t* entity, _In_ size_t count)
244 {
245 assert(entity && count);
246
247 const wchar_t e2 = entity[0];
248 for (size_t i = 0, j = _countof(unicode_sgml); i < j; ) {
249 size_t m = (i + j) / 2;
250 wchar_t e1 = sgml_unicode[unicode_sgml[m]].unicode[0];
251 if (e1 < e2)
252 i = m + 1;
253 else if (e1 > e2)
254 j = m;
255 else {
256 auto r = strncmp(sgml_unicode[unicode_sgml[m]].unicode + 1, _countof(sgml_unicode[0].unicode) - 1, entity + 1, count - 1);
257 if (r < 0)
258 i = m + 1;
259 else if (r > 0)
260 j = m;
261 else {
262 for (; i < m && sgml_unicode[unicode_sgml[m - 1]].unicode[0] == e2 && strncmp(sgml_unicode[unicode_sgml[m - 1]].unicode + 1, _countof(sgml_unicode[0].unicode) - 1, entity + 1, count - 1) == 0; m--);
263 return sgml_unicode[unicode_sgml[m]].sgml;
264 }
265 }
266 }
267 return nullptr;
268 }
270
279 inline void wstr2sgml(
280 _Inout_ std::string& dst,
281 _In_reads_or_z_opt_(count_src) const wchar_t* src,
282 _In_ size_t count_src,
283 _In_ size_t what = 0)
284 {
285 assert(src || !count_src);
286
287 const bool
288 do_ascii = (what & sgml_full) == 0,
289 do_quot = (what & sgml_quot) == 0,
290 do_apos = (what & sgml_apos) == 0,
291 do_lt_gt = (what & sgml_lt_gt) == 0,
292 do_bsol = (what & sgml_bsol) == 0,
293 do_dollar = (what & sgml_dollar) == 0,
294 do_percnt = (what & sgml_percnt) == 0,
295 do_commat = (what & sgml_commat) == 0,
296 do_num = (what & sgml_num) == 0,
297 do_lpar_rpar = (what & sgml_lpar_rpar) == 0,
298 do_lcub_rcub = (what & sgml_lcub_rcub) == 0,
299 do_lsqb_rsqb = (what & sgml_lsqb_rsqb) == 0;
300
301 count_src = wcsnlen(src, count_src);
302 dst.reserve(dst.size() + count_src);
303 for (size_t i = 0; i < count_src;) {
304 size_t n = glyphlen(src + i, count_src - i);
305 if (n == 1 &&
306 do_ascii && (unsigned int)src[i] < 128 &&
307 src[i] != L'&' &&
308 (do_quot || (src[i] != L'"')) &&
309 (do_apos || (src[i] != L'\'')) &&
310 (do_lt_gt || (src[i] != L'<' && src[i] != L'>')) &&
311 (do_bsol || (src[i] != L'\\')) &&
312 (do_dollar || (src[i] != L'$')) &&
313 (do_percnt || (src[i] != L'%')) &&
314 (do_commat || (src[i] != L'@')) &&
315 (do_num || (src[i] != L'#')) &&
316 (do_lpar_rpar || (src[i] != L'(' && src[i] != L')')) &&
317 (do_lcub_rcub || (src[i] != L'{' && src[i] != L'}')) &&
318 (do_lsqb_rsqb || (src[i] != L'[' && src[i] != L']')))
319 {
320 // 7-bit ASCII and no desire to encode it as an SGML entity.
321 dst.append(1, (char)src[i++]);
322 }
323 else {
324 const char* entity = chr2sgml(src + i, n);
325 if (entity) {
326 dst.append(1, '&');
327 dst.append(entity);
328 dst.append(1, ';');
329 i += n;
330 }
331 else if (n == 1) {
332 // Trivial character (1 code unit, 1 glyph), no entity available.
333 if ((unsigned int)src[i] < 128)
334 dst.append(1, (char)src[i++]);
335 else {
336 char tmp[3 + 8 + 1 + 1];
337 snprintf(tmp, _countof(tmp), "&#x%x;", src[i++]);
338 dst.append(tmp);
339 }
340 }
341 else {
342 // Non-trivial character. Decompose.
343 const size_t end = i + n;
344 while (i < end) {
345 if ((entity = chr2sgml(src + i, 1)) != nullptr) {
346 dst.append(1, '&');
347 dst.append(entity);
348 dst.append(1, ';');
349 i++;
350 }
351 else if ((unsigned int)src[i] < 128)
352 dst.append(1, (char)src[i++]);
353 else {
354 uint32_t unicode;
355#ifdef _WIN32
356 if (i + 1 < end && is_surrogate_pair(src + i)) {
357 unicode = surrogate_pair_to_ucs4(src + i);
358 i += 2;
359 }
360 else
361#endif
362 {
363 unicode = src[i++];
364 }
365 char tmp[3 + 8 + 1 + 1];
366 snprintf(tmp, _countof(tmp), "&#x%x;", unicode);
367 dst.append(tmp);
368 }
369 }
370 }
371 }
372 }
373 }
374
382 inline void wstr2sgml(
383 _Inout_ std::string& dst,
384 _In_ const std::wstring& src,
385 _In_ size_t what = 0)
386 {
387 wstr2sgml(dst, src.c_str(), src.size(), what);
388 }
389
399 inline std::string wstr2sgml(
400 _In_reads_or_z_opt_(count_src) const wchar_t* src,
401 _In_ size_t count_src,
402 _In_ size_t what = 0)
403 {
404 std::string dst;
405 wstr2sgml(dst, src, count_src, what);
406 return dst;
407 }
408
417 inline std::string wstr2sgml(
418 _In_ const std::wstring& src,
419 _In_ size_t what = 0)
420 {
421 return wstr2sgml(src.c_str(), src.size(), what);
422 }
423}
Generated on Mon Jul 24 2023 07:46:44 for stdex by doxygen 1.9.7