@uniweb/semantic-parser 1.0.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -0,0 +1,330 @@
1
+ /**
2
+ * Transform a sequence into content groups with semantic structure
3
+ * @param {Array} sequence Flat sequence of elements
4
+ * @param {Object} options Parsing options
5
+ * @returns {Object} Content organized into groups with identified main content
6
+ */
7
+ function processGroups(sequence, options = {}) {
8
+ const result = {
9
+ main: null,
10
+ items: [],
11
+ metadata: {
12
+ dividerMode: false,
13
+ groups: 0,
14
+ },
15
+ };
16
+
17
+ if (!sequence.length) return result;
18
+
19
+ const groups = splitBySlices(sequence);
20
+
21
+ // Process each group's structure
22
+ const processedGroups = groups.map((group) => processGroupContent(group));
23
+
24
+ // Special handling for first group in divider mode
25
+ if (result.metadata.dividerMode && groups.startsWithDivider) {
26
+ result.items = processedGroups;
27
+ } else {
28
+ // Organize into main content and items
29
+ const shouldBeMain = identifyMainContent(processedGroups);
30
+ if (shouldBeMain) {
31
+ result.main = processedGroups[0];
32
+ result.items = processedGroups.slice(1);
33
+ } else {
34
+ result.items = processedGroups;
35
+ }
36
+ }
37
+
38
+ // result.metadata.groups = processedGroups.length;
39
+ return result;
40
+ }
41
+
42
+ function splitBySlices(sequence) {
43
+ const groups = [];
44
+ let currentGroup = [];
45
+
46
+ for (let i = 0; i < sequence.length; i++) {
47
+ const element = sequence[i];
48
+
49
+ // 1. Handle Dividers (Explicit Split)
50
+ if (element.type === "divider") {
51
+ // Close current group if it has content
52
+ if (currentGroup.length > 0) {
53
+ groups.push(currentGroup);
54
+ currentGroup = [];
55
+ }
56
+ continue; // Consume the divider (don't add to group)
57
+ }
58
+
59
+ // 2. Handle Headings (Semantic Split)
60
+ if (element.type === "heading") {
61
+ // SPECIAL CASE: Banner Image for the whole content
62
+ // If we are at the second element (index 1), and the first element was a banner image,
63
+ // we do NOT close the group. We let the heading merge with the image.
64
+ const isBannerMerge = i === 1 && isBannerImage(sequence, 0);
65
+
66
+ // A new Heading Group starts a new visual block.
67
+ // If we have gathered content in the current group, close it now.
68
+ if (currentGroup.length > 0 && !isBannerMerge) {
69
+ groups.push(currentGroup);
70
+ currentGroup = [];
71
+ }
72
+
73
+ // Consume the entire semantic heading block (Title + Subtitles)
74
+ // We reuse your smart readHeadingGroup logic here!
75
+ const headingBlock = readHeadingGroup(sequence, i);
76
+ currentGroup.push(...headingBlock);
77
+
78
+ // Advance the index by the number of headings consumed
79
+ // (Loop increments i by 1, so we add length - 1)
80
+ i += headingBlock.length - 1;
81
+ } else {
82
+ // 3. Handle Content (Body)
83
+ // Paragraphs, images, lists, etc. just append to the current slice.
84
+ currentGroup.push(element);
85
+ }
86
+ }
87
+
88
+ // Push the final group if not empty
89
+ if (currentGroup.length > 0) {
90
+ groups.push(currentGroup);
91
+ }
92
+
93
+ return groups;
94
+ }
95
+
96
+ /**
97
+ * Check if this is a pretitle - any heading followed by a more important heading
98
+ * (e.g., H3→H1, H2→H1, H6→H5, etc.)
99
+ */
100
+ function isPreTitle(sequence, i) {
101
+ return (
102
+ i + 1 < sequence.length &&
103
+ sequence[i].type === "heading" &&
104
+ sequence[i + 1].type === "heading" &&
105
+ sequence[i].level > sequence[i + 1].level // Smaller heading before larger
106
+ );
107
+ }
108
+
109
+ function isBannerImage(sequence, i) {
110
+ return (
111
+ i === 0 &&
112
+ i + 1 < sequence.length &&
113
+ sequence[i].type === "image" &&
114
+ (sequence[i].role === "banner" || sequence[i + 1].type === "heading")
115
+ );
116
+ }
117
+
118
+ function readHeadingGroup(sequence, startIdx) {
119
+ const elements = [sequence[startIdx]];
120
+
121
+ // Iterate starting from the next element
122
+ for (let i = startIdx + 1; i < sequence.length; i++) {
123
+ const element = sequence[i];
124
+ const previousElement = elements[elements.length - 1];
125
+
126
+ if (element.type !== "heading") {
127
+ break;
128
+ }
129
+
130
+ // Case 1: Strictly Deeper (Standard Subtitle/Deep Header)
131
+ // e.g. H1 -> H2
132
+ if (element.level > previousElement.level) {
133
+ elements.push(element);
134
+ continue;
135
+ }
136
+
137
+ // Case 2: Pretitle Promotion (Small -> Big)
138
+ // Only allowed if we haven't gone deep yet (length is 1)
139
+ // e.g. H2 -> H1
140
+ if (elements.length === 1 && element.level < previousElement.level) {
141
+ elements.push(element);
142
+ continue;
143
+ }
144
+
145
+ // Otherwise (Sibling or New Section), stop.
146
+ break;
147
+ }
148
+ return elements;
149
+ }
150
+
151
+ /**
152
+ * Process a group's content to identify its structure
153
+ */
154
+ function processGroupContent(elements) {
155
+ const header = {
156
+ pretitle: "",
157
+ title: "",
158
+ subtitle: "",
159
+ subtitle2: "",
160
+ alignment: null,
161
+ };
162
+
163
+ const body = {
164
+ imgs: [],
165
+ icons: [],
166
+ videos: [],
167
+ paragraphs: [],
168
+ links: [],
169
+ lists: [],
170
+ buttons: [],
171
+ properties: {},
172
+ propertyBlocks: [],
173
+ cards: [],
174
+ documents: [],
175
+ forms: [],
176
+ quotes: [],
177
+ headings: [],
178
+ };
179
+
180
+ const metadata = {
181
+ level: null,
182
+ contentTypes: new Set(),
183
+ };
184
+
185
+ if (!elements)
186
+ return {
187
+ header,
188
+ body,
189
+ metadata,
190
+ };
191
+
192
+ for (let i = 0; i < elements.length; i++) {
193
+ //We shuold only set pretitle once
194
+ if (isPreTitle(elements, i) && !header.pretitle) {
195
+ header.pretitle = elements[i].text;
196
+ i++; // move to known next heading (H1 or h2)
197
+ }
198
+
199
+ const element = elements[i];
200
+
201
+ if (element.type === "heading") {
202
+ if (element.children && Array.isArray(element.children))
203
+ processInlineElements(element.children, body);
204
+
205
+ //We shuold set the group level to the highest one instead of the first one.
206
+ metadata.level ??= element.level;
207
+
208
+ // Extract alignment from first heading
209
+ if (!header.alignment && element.attrs?.textAlign) {
210
+ header.alignment = element.attrs.textAlign;
211
+ }
212
+ // h3 h2 h1 h1
213
+ // Assign to header fields
214
+ // h3 h2 h3 h4
215
+ if (!header.title) {
216
+ header.title = element.text;
217
+ } else if (!header.subtitle) {
218
+ header.subtitle = element.text;
219
+ } else if (!header.subtitle2) {
220
+ header.subtitle2 = element.text;
221
+ } else {
222
+ // After subtitle2, we're in body - collect heading
223
+ body.headings.push(element.text);
224
+ }
225
+ } else if (element.type === "list") {
226
+ const listItems = element.children;
227
+
228
+ body.lists.push(
229
+ listItems.map((listItem) => processGroupContent(listItem).body)
230
+ );
231
+ } else {
232
+ let preserveProps = {
233
+ ...element.attrs,
234
+ };
235
+
236
+ switch (element.type) {
237
+ case "paragraph":
238
+ if (element.children && Array.isArray(element.children))
239
+ processInlineElements(element.children, body);
240
+
241
+ if (element.text) body.paragraphs.push(element.text);
242
+ break;
243
+
244
+ case "image":
245
+ body.imgs.push(preserveProps);
246
+ break;
247
+
248
+ case "video":
249
+ body.videos.push(preserveProps);
250
+ break;
251
+
252
+ case "link":
253
+ if (element.children && Array.isArray(element.children))
254
+ processInlineElements(element.children, body);
255
+
256
+ body.links.push(preserveProps);
257
+ break;
258
+
259
+ case "icon":
260
+ //Might be string
261
+ body.icons.push(element.attrs);
262
+ break;
263
+
264
+ case "button":
265
+ body.buttons.push({
266
+ attrs: element.attrs,
267
+ content: element.text,
268
+ });
269
+ break;
270
+
271
+ case "blockquote":
272
+ // Process blockquote content recursively
273
+ const quoteContent = processGroupContent(element.children);
274
+ body.quotes.push(quoteContent.body);
275
+ break;
276
+
277
+ case "codeBlock":
278
+ const codeData = element.text;
279
+ body.properties = codeData; // Last one
280
+ body.propertyBlocks.push(codeData); // All of them
281
+ break;
282
+
283
+ case "form":
284
+ body.forms.push(element.data || element.attrs);
285
+ break;
286
+
287
+ case "card-group":
288
+ body.cards.push(...element.cards);
289
+ break;
290
+
291
+ case "document-group":
292
+ body.documents.push(...element.documents);
293
+ break;
294
+ }
295
+ }
296
+ }
297
+
298
+ return {
299
+ header,
300
+ body,
301
+ metadata,
302
+ };
303
+ }
304
+
305
+ /**
306
+ * Determine if the first group should be treated as main content
307
+ */
308
+ function identifyMainContent(groups) {
309
+ if (groups.length === 0) return false;
310
+
311
+ // Single group is main content
312
+ if (groups.length === 1) return true;
313
+
314
+ // First group should be more important (lower level) than second to be main
315
+ const first = groups[0].metadata.level;
316
+ const second = groups[1].metadata.level;
317
+
318
+ return first ? !second || first < second : false;
319
+ }
320
+
321
+ function processInlineElements(children, body) {
322
+ children.forEach((item) => {
323
+ //Handle icons only for now
324
+ if (item.type === "icon") {
325
+ body.icons.push(item.attrs);
326
+ }
327
+ });
328
+ }
329
+
330
+ export { processGroups };
@@ -0,0 +1,379 @@
1
+ /**
2
+ * Transform a sequence into content groups with semantic structure
3
+ * @param {Array} sequence Flat sequence of elements
4
+ * @param {Object} options Parsing options
5
+ * @returns {Object} Content organized into groups with identified main content
6
+ */
7
+ function processGroups(sequence, options = {}) {
8
+ const result = {
9
+ main: null,
10
+ items: [],
11
+ metadata: {
12
+ dividerMode: false,
13
+ groups: 0,
14
+ },
15
+ };
16
+
17
+ if (!sequence.length) return result;
18
+
19
+ // Check if using divider mode
20
+ result.metadata.dividerMode = sequence.some((el) => el.type === "divider");
21
+
22
+ // Split sequence into raw groups
23
+ const groups = result.metadata.dividerMode
24
+ ? splitByDividers(sequence)
25
+ : splitByHeadings(sequence, options);
26
+
27
+ // Process each group's structure
28
+ const processedGroups = groups.map((group) => processGroupContent(group));
29
+
30
+ // Special handling for first group in divider mode
31
+ if (result.metadata.dividerMode && groups.startsWithDivider) {
32
+ result.items = processedGroups;
33
+ } else {
34
+ // Organize into main content and items
35
+ const shouldBeMain = identifyMainContent(processedGroups);
36
+ if (shouldBeMain) {
37
+ result.main = processedGroups[0];
38
+ result.items = processedGroups.slice(1);
39
+ } else {
40
+ result.items = processedGroups;
41
+ }
42
+ }
43
+
44
+ // result.metadata.groups = processedGroups.length;
45
+ return result;
46
+ }
47
+
48
+ /**
49
+ * Split sequence into groups using dividers
50
+ */
51
+ function splitByDividers(sequence) {
52
+ const groups = [];
53
+ let currentGroup = [];
54
+ let startsWithDivider = false;
55
+
56
+ // Check if content effectively starts with divider (ignoring whitespace etc)
57
+ for (let i = 0; i < sequence.length; i++) {
58
+ const element = sequence[i];
59
+
60
+ if (element.type === "divider") {
61
+ if (currentGroup.length === 0 && groups.length === 0) {
62
+ startsWithDivider = true;
63
+ } else if (currentGroup.length > 0) {
64
+ groups.push(currentGroup);
65
+ currentGroup = [];
66
+ }
67
+ } else {
68
+ currentGroup.push(element);
69
+ }
70
+ }
71
+
72
+ if (currentGroup.length > 0) {
73
+ groups.push(currentGroup);
74
+ }
75
+
76
+ groups.startsWithDivider = startsWithDivider;
77
+ return groups;
78
+ }
79
+
80
+ /**
81
+ * Split sequence into groups using heading patterns
82
+ */
83
+ function splitByHeadings(sequence, options = {}) {
84
+ const groups = [];
85
+ let currentGroup = [];
86
+ let isPreOpened = false;
87
+
88
+ // Consider if current group is pre opened (only has banner or pretitle)
89
+ // before starting a new group.
90
+ const startGroup = (preOpen) => {
91
+ if (currentGroup.length && !isPreOpened) {
92
+ groups.push(currentGroup);
93
+ currentGroup = [];
94
+ }
95
+ isPreOpened = preOpen;
96
+ };
97
+
98
+ for (let i = 0; i < sequence.length; i++) {
99
+ // Only allow a banner for the first group
100
+ if (!groups.length && isBannerImage(sequence, i)) {
101
+ startGroup(true); // pre open a new group
102
+ currentGroup.push(sequence[i]);
103
+ i++; // move to known next element (it will be a heading)
104
+ }
105
+
106
+ // Handle special pretitle case before consuming all consecutive
107
+ // headings with increasing levels
108
+ if (isPreTitle(sequence, i)) {
109
+ startGroup(true); // pre open a new group
110
+ currentGroup.push(sequence[i]);
111
+ i++; // move to known next element (it will be a heading)
112
+ }
113
+
114
+ const element = sequence[i];
115
+
116
+ if (element.type === "heading") {
117
+ const headings = readHeadingGroup(sequence, i);
118
+ startGroup(false);
119
+
120
+ // Add headings to the current group
121
+ currentGroup.push(...headings);
122
+ i += headings.length - 1; // skip all the added headings
123
+ } else {
124
+ currentGroup.push(element);
125
+ }
126
+ }
127
+
128
+ if (currentGroup.length > 0) {
129
+ groups.push(currentGroup);
130
+ }
131
+
132
+ return groups;
133
+ }
134
+
135
+ /**
136
+ * Check if this is a pretitle - any heading followed by a more important heading
137
+ * (e.g., H3→H1, H2→H1, H6→H5, etc.)
138
+ */
139
+ function isPreTitle(sequence, i) {
140
+ return (
141
+ i + 1 < sequence.length &&
142
+ sequence[i].type === "heading" &&
143
+ sequence[i + 1].type === "heading" &&
144
+ sequence[i].level > sequence[i + 1].level // Smaller heading before larger
145
+ );
146
+ }
147
+
148
+ function isBannerImage(sequence, i) {
149
+ return (
150
+ i + 1 < sequence.length &&
151
+ sequence[i].type === "image" &&
152
+ (sequence[i].role === "banner" || sequence[i + 1].type === "heading")
153
+ );
154
+ }
155
+
156
+ /**
157
+ * Eagerly consume all consecutive headings with increasing levels
158
+ * and return them as an array.
159
+ */
160
+ function readHeadingGroup(sequence, i) {
161
+ const elements = [sequence[i]];
162
+ for (i++; i < sequence.length; i++) {
163
+ const element = sequence[i];
164
+
165
+ if (
166
+ element.type === "heading" &&
167
+ element.level > sequence[i - 1].level
168
+ ) {
169
+ elements.push(element);
170
+ } else {
171
+ break;
172
+ }
173
+ }
174
+ return elements;
175
+ }
176
+
177
+ /**
178
+ * Process a group's content to identify its structure
179
+ */
180
+ function processGroupContent(elements) {
181
+ const header = {
182
+ pretitle: "",
183
+ title: "",
184
+ subtitle: "",
185
+ subtitle2: "",
186
+ alignment: null,
187
+ };
188
+ let banner = null;
189
+ const body = {
190
+ imgs: [],
191
+ icons: [],
192
+ videos: [],
193
+ paragraphs: [],
194
+ links: [],
195
+ lists: [],
196
+ buttons: [],
197
+ properties: {},
198
+ propertyBlocks: [],
199
+ cards: [],
200
+ documents: [],
201
+ forms: [],
202
+ quotes: [],
203
+ headings: [],
204
+ };
205
+
206
+ const metadata = {
207
+ level: null,
208
+ contentTypes: new Set(),
209
+ };
210
+
211
+ let inBody = false; // Track when we've finished header section
212
+
213
+ for (let i = 0; i < elements.length; i++) {
214
+ if (isPreTitle(elements, i)) {
215
+ header.pretitle = elements[i].content;
216
+ i++; // move to known next heading (H1 or h2)
217
+ }
218
+
219
+ if (isBannerImage(elements, i)) {
220
+ banner = {
221
+ url: elements[i].src,
222
+ caption: elements[i].caption,
223
+ alt: elements[i].alt,
224
+ };
225
+ i++;
226
+ }
227
+
228
+ const element = elements[i];
229
+
230
+ if (element.type === "heading") {
231
+ metadata.level ??= element.level;
232
+
233
+ // Extract alignment from first heading
234
+ if (!header.alignment && element.attrs?.textAlign) {
235
+ header.alignment = element.attrs.textAlign;
236
+ }
237
+
238
+ // Assign to header fields
239
+ if (!header.title) {
240
+ header.title = element.content;
241
+ } else if (!header.subtitle) {
242
+ header.subtitle = element.content;
243
+ } else if (!header.subtitle2) {
244
+ header.subtitle2 = element.content;
245
+ } else {
246
+ // After subtitle2, we're in body - collect heading
247
+ inBody = true;
248
+ body.headings.push(element.content);
249
+ }
250
+ } else if (element.type === "list") {
251
+ inBody = true;
252
+ body.lists.push(processListContent(element));
253
+ } else {
254
+ inBody = true;
255
+
256
+ switch (element.type) {
257
+ case "paragraph":
258
+ body.paragraphs.push(element.content);
259
+ break;
260
+
261
+ case "image":
262
+ body.imgs.push({
263
+ url: element.src,
264
+ caption: element.caption,
265
+ alt: element.alt,
266
+ });
267
+ break;
268
+
269
+ case "link":
270
+ body.links.push({
271
+ href: element.content.href,
272
+ label: element.content.label,
273
+ });
274
+ break;
275
+
276
+ case "styledLink":
277
+ // Styled link (multi-part with same href)
278
+ body.links.push({
279
+ href: element.href,
280
+ label: element.content,
281
+ target: element.target,
282
+ });
283
+ break;
284
+
285
+ case "icon":
286
+ body.icons.push(element.svg);
287
+ break;
288
+
289
+ case "button":
290
+ body.buttons.push(element);
291
+ break;
292
+
293
+ case "video":
294
+ body.videos.push({
295
+ src: element.src,
296
+ caption: element.caption,
297
+ alt: element.alt,
298
+ });
299
+ break;
300
+
301
+ case "blockquote":
302
+ // Process blockquote content recursively
303
+ const quoteContent = processGroupContent(
304
+ element.content,
305
+ options
306
+ );
307
+ body.quotes.push(quoteContent.body);
308
+ break;
309
+
310
+ case "codeBlock":
311
+ // Use parsed JSON if available, otherwise use text content
312
+ const codeData =
313
+ element.parsed !== null
314
+ ? element.parsed
315
+ : element.content;
316
+ body.properties = codeData; // Last one
317
+ body.propertyBlocks.push(codeData); // All of them
318
+ break;
319
+
320
+ case "card-group":
321
+ body.cards.push(...element.cards);
322
+ break;
323
+
324
+ case "document-group":
325
+ body.documents.push(...element.documents);
326
+ break;
327
+
328
+ case "form":
329
+ body.forms.push(element.data || element.attrs);
330
+ break;
331
+ }
332
+ }
333
+ }
334
+
335
+ return {
336
+ header,
337
+ body,
338
+ banner,
339
+ metadata,
340
+ };
341
+ }
342
+
343
+ function processListContent(list) {
344
+ const { items } = list;
345
+
346
+ return items.map((item) => {
347
+ const { items: nestedList, content: listContent } = item;
348
+
349
+ const parsedContent = processGroupContent(listContent).body;
350
+
351
+ if (nestedList.length) {
352
+ const parsedNestedList = nestedList.map(
353
+ (nestedItem) => processGroupContent(nestedItem.content).body
354
+ );
355
+
356
+ parsedContent.lists = [parsedNestedList];
357
+ }
358
+
359
+ return parsedContent;
360
+ });
361
+ }
362
+
363
+ /**
364
+ * Determine if the first group should be treated as main content
365
+ */
366
+ function identifyMainContent(groups) {
367
+ if (groups.length === 0) return false;
368
+
369
+ // Single group is main content
370
+ if (groups.length === 1) return true;
371
+
372
+ // First group should be more important (lower level) than second to be main
373
+ const first = groups[0].metadata.level;
374
+ const second = groups[1].metadata.level;
375
+
376
+ return first ? !second || first < second : false;
377
+ }
378
+
379
+ export { processGroups };