1 | #include "duckdb/function/cast/vector_cast_helpers.hpp" |
2 | |
3 | namespace duckdb { |
4 | |
5 | // ------- Helper functions for splitting string nested types ------- |
6 | static bool IsNull(const char *buf, idx_t start_pos, Vector &child, idx_t row_idx) { |
7 | if (buf[start_pos] == 'N' && buf[start_pos + 1] == 'U' && buf[start_pos + 2] == 'L' && buf[start_pos + 3] == 'L') { |
8 | FlatVector::SetNull(vector&: child, idx: row_idx, is_null: true); |
9 | return true; |
10 | } |
11 | return false; |
12 | } |
13 | |
14 | inline static void SkipWhitespace(const char *buf, idx_t &pos, idx_t len) { |
15 | while (pos < len && StringUtil::CharacterIsSpace(c: buf[pos])) { |
16 | pos++; |
17 | } |
18 | } |
19 | |
20 | static bool SkipToCloseQuotes(idx_t &pos, const char *buf, idx_t &len) { |
21 | char quote = buf[pos]; |
22 | pos++; |
23 | |
24 | while (pos < len) { |
25 | if (buf[pos] == quote) { |
26 | return true; |
27 | } |
28 | pos++; |
29 | } |
30 | return false; |
31 | } |
32 | |
33 | static bool SkipToClose(idx_t &idx, const char *buf, idx_t &len, idx_t &lvl, char close_bracket) { |
34 | idx++; |
35 | |
36 | while (idx < len) { |
37 | if (buf[idx] == '"' || buf[idx] == '\'') { |
38 | if (!SkipToCloseQuotes(pos&: idx, buf, len)) { |
39 | return false; |
40 | } |
41 | } else if (buf[idx] == '{') { |
42 | if (!SkipToClose(idx, buf, len, lvl, close_bracket: '}')) { |
43 | return false; |
44 | } |
45 | } else if (buf[idx] == '[') { |
46 | if (!SkipToClose(idx, buf, len, lvl, close_bracket: ']')) { |
47 | return false; |
48 | } |
49 | lvl++; |
50 | } else if (buf[idx] == close_bracket) { |
51 | if (close_bracket == ']') { |
52 | lvl--; |
53 | } |
54 | return true; |
55 | } |
56 | idx++; |
57 | } |
58 | return false; |
59 | } |
60 | |
61 | static idx_t StringTrim(const char *buf, idx_t &start_pos, idx_t pos) { |
62 | idx_t trailing_whitespace = 0; |
63 | while (StringUtil::CharacterIsSpace(c: buf[pos - trailing_whitespace - 1])) { |
64 | trailing_whitespace++; |
65 | } |
66 | if ((buf[start_pos] == '"' && buf[pos - trailing_whitespace - 1] == '"') || |
67 | (buf[start_pos] == '\'' && buf[pos - trailing_whitespace - 1] == '\'')) { |
68 | start_pos++; |
69 | trailing_whitespace++; |
70 | } |
71 | return (pos - trailing_whitespace); |
72 | } |
73 | |
74 | struct CountPartOperation { |
75 | idx_t count = 0; |
76 | |
77 | bool HandleKey(const char *buf, idx_t start_pos, idx_t pos) { |
78 | count++; |
79 | return true; |
80 | } |
81 | void HandleValue(const char *buf, idx_t start_pos, idx_t pos) { |
82 | count++; |
83 | } |
84 | }; |
85 | |
86 | // ------- LIST SPLIT ------- |
87 | struct SplitStringListOperation { |
88 | SplitStringListOperation(string_t *child_data, idx_t &child_start, Vector &child) |
89 | : child_data(child_data), child_start(child_start), child(child) { |
90 | } |
91 | |
92 | string_t *child_data; |
93 | idx_t &child_start; |
94 | Vector &child; |
95 | |
96 | void HandleValue(const char *buf, idx_t start_pos, idx_t pos) { |
97 | if ((pos - start_pos) == 4 && IsNull(buf, start_pos, child, row_idx: child_start)) { |
98 | child_start++; |
99 | return; |
100 | } |
101 | child_data[child_start] = StringVector::AddString(vector&: child, data: buf + start_pos, len: pos - start_pos); |
102 | child_start++; |
103 | } |
104 | }; |
105 | |
106 | template <class OP> |
107 | static bool SplitStringListInternal(const string_t &input, OP &state) { |
108 | const char *buf = input.GetData(); |
109 | idx_t len = input.GetSize(); |
110 | idx_t lvl = 1; |
111 | idx_t pos = 0; |
112 | |
113 | SkipWhitespace(buf, pos, len); |
114 | if (pos == len || buf[pos] != '[') { |
115 | return false; |
116 | } |
117 | |
118 | SkipWhitespace(buf, pos&: ++pos, len); |
119 | idx_t start_pos = pos; |
120 | while (pos < len) { |
121 | if (buf[pos] == '[') { |
122 | if (!SkipToClose(idx&: pos, buf, len, lvl&: ++lvl, close_bracket: ']')) { |
123 | return false; |
124 | } |
125 | } else if ((buf[pos] == '"' || buf[pos] == '\'') && pos == start_pos) { |
126 | SkipToCloseQuotes(pos, buf, len); |
127 | } else if (buf[pos] == '{') { |
128 | idx_t struct_lvl = 0; |
129 | SkipToClose(idx&: pos, buf, len, lvl&: struct_lvl, close_bracket: '}'); |
130 | } else if (buf[pos] == ',' || buf[pos] == ']') { |
131 | idx_t trailing_whitespace = 0; |
132 | while (StringUtil::CharacterIsSpace(c: buf[pos - trailing_whitespace - 1])) { |
133 | trailing_whitespace++; |
134 | } |
135 | if (!(buf[pos] == ']' && start_pos == pos)) { |
136 | state.HandleValue(buf, start_pos, pos - trailing_whitespace); |
137 | } // else the list is empty |
138 | if (buf[pos] == ']') { |
139 | lvl--; |
140 | break; |
141 | } |
142 | SkipWhitespace(buf, pos&: ++pos, len); |
143 | start_pos = pos; |
144 | continue; |
145 | } |
146 | pos++; |
147 | } |
148 | SkipWhitespace(buf, pos&: ++pos, len); |
149 | return (pos == len && lvl == 0); |
150 | } |
151 | |
152 | bool VectorStringToList::SplitStringList(const string_t &input, string_t *child_data, idx_t &child_start, |
153 | Vector &child) { |
154 | SplitStringListOperation state(child_data, child_start, child); |
155 | return SplitStringListInternal<SplitStringListOperation>(input, state); |
156 | } |
157 | |
158 | idx_t VectorStringToList::CountPartsList(const string_t &input) { |
159 | CountPartOperation state; |
160 | SplitStringListInternal<CountPartOperation>(input, state); |
161 | return state.count; |
162 | } |
163 | |
164 | // ------- MAP SPLIT ------- |
165 | struct SplitStringMapOperation { |
166 | SplitStringMapOperation(string_t *child_key_data, string_t *child_val_data, idx_t &child_start, Vector &varchar_key, |
167 | Vector &varchar_val) |
168 | : child_key_data(child_key_data), child_val_data(child_val_data), child_start(child_start), |
169 | varchar_key(varchar_key), varchar_val(varchar_val) { |
170 | } |
171 | |
172 | string_t *child_key_data; |
173 | string_t *child_val_data; |
174 | idx_t &child_start; |
175 | Vector &varchar_key; |
176 | Vector &varchar_val; |
177 | |
178 | bool HandleKey(const char *buf, idx_t start_pos, idx_t pos) { |
179 | if ((pos - start_pos) == 4 && IsNull(buf, start_pos, child&: varchar_key, row_idx: child_start)) { |
180 | FlatVector::SetNull(vector&: varchar_val, idx: child_start, is_null: true); |
181 | child_start++; |
182 | return false; |
183 | } |
184 | child_key_data[child_start] = StringVector::AddString(vector&: varchar_key, data: buf + start_pos, len: pos - start_pos); |
185 | return true; |
186 | } |
187 | |
188 | void HandleValue(const char *buf, idx_t start_pos, idx_t pos) { |
189 | if ((pos - start_pos) == 4 && IsNull(buf, start_pos, child&: varchar_val, row_idx: child_start)) { |
190 | child_start++; |
191 | return; |
192 | } |
193 | child_val_data[child_start] = StringVector::AddString(vector&: varchar_val, data: buf + start_pos, len: pos - start_pos); |
194 | child_start++; |
195 | } |
196 | }; |
197 | |
198 | template <class OP> |
199 | static bool FindKeyOrValueMap(const char *buf, idx_t len, idx_t &pos, OP &state, bool key) { |
200 | auto start_pos = pos; |
201 | idx_t lvl = 0; |
202 | while (pos < len) { |
203 | if (buf[pos] == '"' || buf[pos] == '\'') { |
204 | SkipToCloseQuotes(pos, buf, len); |
205 | } else if (buf[pos] == '{') { |
206 | SkipToClose(idx&: pos, buf, len, lvl, close_bracket: '}'); |
207 | } else if (buf[pos] == '[') { |
208 | SkipToClose(idx&: pos, buf, len, lvl, close_bracket: ']'); |
209 | } else if (key && buf[pos] == '=') { |
210 | idx_t end_pos = StringTrim(buf, start_pos, pos); |
211 | return state.HandleKey(buf, start_pos, end_pos); // put string in KEY_child_vector |
212 | } else if (!key && (buf[pos] == ',' || buf[pos] == '}')) { |
213 | idx_t end_pos = StringTrim(buf, start_pos, pos); |
214 | state.HandleValue(buf, start_pos, end_pos); // put string in VALUE_child_vector |
215 | return true; |
216 | } |
217 | pos++; |
218 | } |
219 | return false; |
220 | } |
221 | |
222 | template <class OP> |
223 | static bool SplitStringMapInternal(const string_t &input, OP &state) { |
224 | const char *buf = input.GetData(); |
225 | idx_t len = input.GetSize(); |
226 | idx_t pos = 0; |
227 | |
228 | SkipWhitespace(buf, pos, len); |
229 | if (pos == len || buf[pos] != '{') { |
230 | return false; |
231 | } |
232 | SkipWhitespace(buf, pos&: ++pos, len); |
233 | if (pos == len) { |
234 | return false; |
235 | } |
236 | if (buf[pos] == '}') { |
237 | SkipWhitespace(buf, pos&: ++pos, len); |
238 | return (pos == len); |
239 | } |
240 | while (pos < len) { |
241 | if (!FindKeyOrValueMap(buf, len, pos, state, true)) { |
242 | return false; |
243 | } |
244 | SkipWhitespace(buf, pos&: ++pos, len); |
245 | if (!FindKeyOrValueMap(buf, len, pos, state, false)) { |
246 | return false; |
247 | } |
248 | SkipWhitespace(buf, pos&: ++pos, len); |
249 | } |
250 | return true; |
251 | } |
252 | |
253 | bool VectorStringToMap::SplitStringMap(const string_t &input, string_t *child_key_data, string_t *child_val_data, |
254 | idx_t &child_start, Vector &varchar_key, Vector &varchar_val) { |
255 | SplitStringMapOperation state(child_key_data, child_val_data, child_start, varchar_key, varchar_val); |
256 | return SplitStringMapInternal<SplitStringMapOperation>(input, state); |
257 | } |
258 | |
259 | idx_t VectorStringToMap::CountPartsMap(const string_t &input) { |
260 | CountPartOperation state; |
261 | SplitStringMapInternal<CountPartOperation>(input, state); |
262 | return state.count; |
263 | } |
264 | |
265 | // ------- STRUCT SPLIT ------- |
266 | static bool FindKeyStruct(const char *buf, idx_t len, idx_t &pos) { |
267 | while (pos < len) { |
268 | if (buf[pos] == ':') { |
269 | return true; |
270 | } |
271 | pos++; |
272 | } |
273 | return false; |
274 | } |
275 | |
276 | static bool FindValueStruct(const char *buf, idx_t len, idx_t &pos, Vector &varchar_child, idx_t &row_idx, |
277 | ValidityMask *child_mask) { |
278 | auto start_pos = pos; |
279 | idx_t lvl = 0; |
280 | while (pos < len) { |
281 | if (buf[pos] == '"' || buf[pos] == '\'') { |
282 | SkipToCloseQuotes(pos, buf, len); |
283 | } else if (buf[pos] == '{') { |
284 | SkipToClose(idx&: pos, buf, len, lvl, close_bracket: '}'); |
285 | } else if (buf[pos] == '[') { |
286 | SkipToClose(idx&: pos, buf, len, lvl, close_bracket: ']'); |
287 | } else if (buf[pos] == ',' || buf[pos] == '}') { |
288 | idx_t end_pos = StringTrim(buf, start_pos, pos); |
289 | if ((end_pos - start_pos) == 4 && IsNull(buf, start_pos, child&: varchar_child, row_idx)) { |
290 | return true; |
291 | } |
292 | FlatVector::GetData<string_t>(vector&: varchar_child)[row_idx] = |
293 | StringVector::AddString(vector&: varchar_child, data: buf + start_pos, len: end_pos - start_pos); |
294 | child_mask->SetValid(row_idx); // any child not set to valid will remain invalid |
295 | return true; |
296 | } |
297 | pos++; |
298 | } |
299 | return false; |
300 | } |
301 | |
302 | bool VectorStringToStruct::SplitStruct(const string_t &input, vector<unique_ptr<Vector>> &varchar_vectors, |
303 | idx_t &row_idx, string_map_t<idx_t> &child_names, |
304 | vector<ValidityMask *> &child_masks) { |
305 | const char *buf = input.GetData(); |
306 | idx_t len = input.GetSize(); |
307 | idx_t pos = 0; |
308 | idx_t child_idx; |
309 | |
310 | SkipWhitespace(buf, pos, len); |
311 | if (pos == len || buf[pos] != '{') { |
312 | return false; |
313 | } |
314 | SkipWhitespace(buf, pos&: ++pos, len); |
315 | if (buf[pos] == '}') { |
316 | pos++; |
317 | } else { |
318 | while (pos < len) { |
319 | auto key_start = pos; |
320 | if (!FindKeyStruct(buf, len, pos)) { |
321 | return false; |
322 | } |
323 | auto key_end = StringTrim(buf, start_pos&: key_start, pos); |
324 | string_t found_key(buf + key_start, key_end - key_start); |
325 | |
326 | auto it = child_names.find(x: found_key); |
327 | if (it == child_names.end()) { |
328 | return false; // false key |
329 | } |
330 | child_idx = it->second; |
331 | SkipWhitespace(buf, pos&: ++pos, len); |
332 | if (!FindValueStruct(buf, len, pos, varchar_child&: *varchar_vectors[child_idx], row_idx, child_mask: child_masks[child_idx])) { |
333 | return false; |
334 | } |
335 | SkipWhitespace(buf, pos&: ++pos, len); |
336 | } |
337 | } |
338 | SkipWhitespace(buf, pos, len); |
339 | return (pos == len); |
340 | } |
341 | |
342 | } // namespace duckdb |
343 | |