]>
Commit | Line | Data |
---|---|---|
69bec673 | 1 | import collections.abc |
2 | import contextlib | |
3 | import inspect | |
4 | import itertools | |
5 | import re | |
ffbd4f2a | 6 | import xml.etree.ElementTree |
69bec673 | 7 | |
8 | from ._utils import ( | |
9 | IDENTITY, | |
10 | NO_DEFAULT, | |
11 | LazyList, | |
0b6f829b | 12 | deprecation_warning, |
69bec673 | 13 | is_iterable_like, |
14 | try_call, | |
15 | variadic, | |
16 | ) | |
17 | ||
18 | ||
19 | def traverse_obj( | |
20 | obj, *paths, default=NO_DEFAULT, expected_type=None, get_all=True, | |
0b6f829b | 21 | casesense=True, is_user_input=NO_DEFAULT, traverse_string=False): |
69bec673 | 22 | """ |
23 | Safely traverse nested `dict`s and `Iterable`s | |
24 | ||
25 | >>> obj = [{}, {"key": "value"}] | |
26 | >>> traverse_obj(obj, (1, "key")) | |
f9fb3ce8 | 27 | 'value' |
69bec673 | 28 | |
29 | Each of the provided `paths` is tested and the first producing a valid result will be returned. | |
30 | The next path will also be tested if the path branched but no results could be found. | |
31 | Supported values for traversal are `Mapping`, `Iterable` and `re.Match`. | |
32 | Unhelpful values (`{}`, `None`) are treated as the absence of a value and discarded. | |
33 | ||
34 | The paths will be wrapped in `variadic`, so that `'key'` is conveniently the same as `('key', )`. | |
35 | ||
36 | The keys in the path can be one of: | |
37 | - `None`: Return the current object. | |
38 | - `set`: Requires the only item in the set to be a type or function, | |
39 | like `{type}`/`{func}`. If a `type`, returns only values | |
40 | of this type. If a function, returns `func(obj)`. | |
41 | - `str`/`int`: Return `obj[key]`. For `re.Match`, return `obj.group(key)`. | |
42 | - `slice`: Branch out and return all values in `obj[key]`. | |
43 | - `Ellipsis`: Branch out and return a list of all values. | |
44 | - `tuple`/`list`: Branch out and return a list of all matching values. | |
45 | Read as: `[traverse_obj(obj, branch) for branch in branches]`. | |
46 | - `function`: Branch out and return values filtered by the function. | |
47 | Read as: `[value for key, value in obj if function(key, value)]`. | |
48 | For `Iterable`s, `key` is the index of the value. | |
49 | For `re.Match`es, `key` is the group number (0 = full match) | |
50 | as well as additionally any group names, if given. | |
51 | - `dict` Transform the current object and return a matching dict. | |
52 | Read as: `{key: traverse_obj(obj, path) for key, path in dct.items()}`. | |
53 | ||
54 | `tuple`, `list`, and `dict` all support nested paths and branches. | |
55 | ||
56 | @params paths Paths which to traverse by. | |
57 | @param default Value to return if the paths do not match. | |
58 | If the last key in the path is a `dict`, it will apply to each value inside | |
59 | the dict instead, depth first. Try to avoid if using nested `dict` keys. | |
60 | @param expected_type If a `type`, only accept final values of this type. | |
61 | If any other callable, try to call the function on each result. | |
62 | If the last key in the path is a `dict`, it will apply to each value inside | |
63 | the dict instead, recursively. This does respect branching paths. | |
64 | @param get_all If `False`, return the first matching result, otherwise all matching ones. | |
65 | @param casesense If `False`, consider string dictionary keys as case insensitive. | |
66 | ||
0b6f829b | 67 | `traverse_string` is only meant to be used by YoutubeDL.prepare_outtmpl and is not part of the API |
69bec673 | 68 | |
69bec673 | 69 | @param traverse_string Whether to traverse into objects as strings. |
70 | If `True`, any non-compatible object will first be | |
71 | converted into a string and then traversed into. | |
72 | The return value of that path will be a string instead, | |
73 | not respecting any further branching. | |
74 | ||
75 | ||
76 | @returns The result of the object traversal. | |
77 | If successful, `get_all=True`, and the path branches at least once, | |
78 | then a list of results is returned instead. | |
79 | If no `default` is given and the last path branches, a `list` of results | |
80 | is always returned. If a path ends on a `dict` that result will always be a `dict`. | |
81 | """ | |
0b6f829b SS |
82 | if is_user_input is not NO_DEFAULT: |
83 | deprecation_warning('The is_user_input parameter is deprecated and no longer works') | |
84 | ||
69bec673 | 85 | casefold = lambda k: k.casefold() if isinstance(k, str) else k |
86 | ||
87 | if isinstance(expected_type, type): | |
88 | type_test = lambda val: val if isinstance(val, expected_type) else None | |
89 | else: | |
90 | type_test = lambda val: try_call(expected_type or IDENTITY, args=(val,)) | |
91 | ||
92 | def apply_key(key, obj, is_last): | |
93 | branching = False | |
94 | result = None | |
95 | ||
96 | if obj is None and traverse_string: | |
97 | if key is ... or callable(key) or isinstance(key, slice): | |
98 | branching = True | |
99 | result = () | |
100 | ||
101 | elif key is None: | |
102 | result = obj | |
103 | ||
104 | elif isinstance(key, set): | |
105 | assert len(key) == 1, 'Set should only be used to wrap a single item' | |
106 | item = next(iter(key)) | |
107 | if isinstance(item, type): | |
108 | if isinstance(obj, item): | |
109 | result = obj | |
110 | else: | |
111 | result = try_call(item, args=(obj,)) | |
112 | ||
113 | elif isinstance(key, (list, tuple)): | |
114 | branching = True | |
115 | result = itertools.chain.from_iterable( | |
116 | apply_path(obj, branch, is_last)[0] for branch in key) | |
117 | ||
118 | elif key is ...: | |
119 | branching = True | |
120 | if isinstance(obj, collections.abc.Mapping): | |
121 | result = obj.values() | |
ffbd4f2a | 122 | elif is_iterable_like(obj) or isinstance(obj, xml.etree.ElementTree.Element): |
69bec673 | 123 | result = obj |
124 | elif isinstance(obj, re.Match): | |
125 | result = obj.groups() | |
126 | elif traverse_string: | |
127 | branching = False | |
128 | result = str(obj) | |
129 | else: | |
130 | result = () | |
131 | ||
132 | elif callable(key): | |
133 | branching = True | |
134 | if isinstance(obj, collections.abc.Mapping): | |
135 | iter_obj = obj.items() | |
ffbd4f2a | 136 | elif is_iterable_like(obj) or isinstance(obj, xml.etree.ElementTree.Element): |
69bec673 | 137 | iter_obj = enumerate(obj) |
138 | elif isinstance(obj, re.Match): | |
139 | iter_obj = itertools.chain( | |
140 | enumerate((obj.group(), *obj.groups())), | |
141 | obj.groupdict().items()) | |
142 | elif traverse_string: | |
143 | branching = False | |
144 | iter_obj = enumerate(str(obj)) | |
145 | else: | |
146 | iter_obj = () | |
147 | ||
148 | result = (v for k, v in iter_obj if try_call(key, args=(k, v))) | |
149 | if not branching: # string traversal | |
150 | result = ''.join(result) | |
151 | ||
152 | elif isinstance(key, dict): | |
153 | iter_obj = ((k, _traverse_obj(obj, v, False, is_last)) for k, v in key.items()) | |
154 | result = { | |
155 | k: v if v is not None else default for k, v in iter_obj | |
156 | if v is not None or default is not NO_DEFAULT | |
157 | } or None | |
158 | ||
159 | elif isinstance(obj, collections.abc.Mapping): | |
160 | result = (try_call(obj.get, args=(key,)) if casesense or try_call(obj.__contains__, args=(key,)) else | |
161 | next((v for k, v in obj.items() if casefold(k) == key), None)) | |
162 | ||
163 | elif isinstance(obj, re.Match): | |
164 | if isinstance(key, int) or casesense: | |
165 | with contextlib.suppress(IndexError): | |
166 | result = obj.group(key) | |
167 | ||
168 | elif isinstance(key, str): | |
169 | result = next((v for k, v in obj.groupdict().items() if casefold(k) == key), None) | |
170 | ||
171 | elif isinstance(key, (int, slice)): | |
ffbd4f2a | 172 | if is_iterable_like(obj, (collections.abc.Sequence, xml.etree.ElementTree.Element)): |
69bec673 | 173 | branching = isinstance(key, slice) |
174 | with contextlib.suppress(IndexError): | |
175 | result = obj[key] | |
176 | elif traverse_string: | |
177 | with contextlib.suppress(IndexError): | |
178 | result = str(obj)[key] | |
179 | ||
ffbd4f2a SS |
180 | elif isinstance(obj, xml.etree.ElementTree.Element) and isinstance(key, str): |
181 | xpath, _, special = key.rpartition('/') | |
182 | if not special.startswith('@') and special != 'text()': | |
183 | xpath = key | |
184 | special = None | |
185 | ||
186 | # Allow abbreviations of relative paths, absolute paths error | |
187 | if xpath.startswith('/'): | |
188 | xpath = f'.{xpath}' | |
189 | elif xpath and not xpath.startswith('./'): | |
190 | xpath = f'./{xpath}' | |
191 | ||
192 | def apply_specials(element): | |
193 | if special is None: | |
194 | return element | |
195 | if special == '@': | |
196 | return element.attrib | |
197 | if special.startswith('@'): | |
198 | return try_call(element.attrib.get, args=(special[1:],)) | |
199 | if special == 'text()': | |
200 | return element.text | |
201 | assert False, f'apply_specials is missing case for {special!r}' | |
202 | ||
203 | if xpath: | |
204 | result = list(map(apply_specials, obj.iterfind(xpath))) | |
205 | else: | |
206 | result = apply_specials(obj) | |
207 | ||
69bec673 | 208 | return branching, result if branching else (result,) |
209 | ||
210 | def lazy_last(iterable): | |
211 | iterator = iter(iterable) | |
212 | prev = next(iterator, NO_DEFAULT) | |
213 | if prev is NO_DEFAULT: | |
214 | return | |
215 | ||
216 | for item in iterator: | |
217 | yield False, prev | |
218 | prev = item | |
219 | ||
220 | yield True, prev | |
221 | ||
222 | def apply_path(start_obj, path, test_type): | |
223 | objs = (start_obj,) | |
224 | has_branched = False | |
225 | ||
226 | key = None | |
227 | for last, key in lazy_last(variadic(path, (str, bytes, dict, set))): | |
69bec673 | 228 | if not casesense and isinstance(key, str): |
229 | key = key.casefold() | |
230 | ||
3699eeb6 SS |
231 | if key in (any, all): |
232 | has_branched = False | |
233 | filtered_objs = (obj for obj in objs if obj not in (None, {})) | |
234 | if key is any: | |
235 | objs = (next(filtered_objs, None),) | |
236 | else: | |
237 | objs = (list(filtered_objs),) | |
238 | continue | |
239 | ||
69bec673 | 240 | if __debug__ and callable(key): |
241 | # Verify function signature | |
242 | inspect.signature(key).bind(None, None) | |
243 | ||
244 | new_objs = [] | |
245 | for obj in objs: | |
246 | branching, results = apply_key(key, obj, last) | |
247 | has_branched |= branching | |
248 | new_objs.append(results) | |
249 | ||
250 | objs = itertools.chain.from_iterable(new_objs) | |
251 | ||
252 | if test_type and not isinstance(key, (dict, list, tuple)): | |
253 | objs = map(type_test, objs) | |
254 | ||
255 | return objs, has_branched, isinstance(key, dict) | |
256 | ||
257 | def _traverse_obj(obj, path, allow_empty, test_type): | |
258 | results, has_branched, is_dict = apply_path(obj, path, test_type) | |
259 | results = LazyList(item for item in results if item not in (None, {})) | |
260 | if get_all and has_branched: | |
261 | if results: | |
262 | return results.exhaust() | |
263 | if allow_empty: | |
264 | return [] if default is NO_DEFAULT else default | |
265 | return None | |
266 | ||
267 | return results[0] if results else {} if allow_empty and is_dict else None | |
268 | ||
269 | for index, path in enumerate(paths, 1): | |
270 | result = _traverse_obj(obj, path, index == len(paths), True) | |
271 | if result is not None: | |
272 | return result | |
273 | ||
274 | return None if default is NO_DEFAULT else default | |
275 | ||
276 | ||
277 | def get_first(obj, *paths, **kwargs): | |
278 | return traverse_obj(obj, *((..., *variadic(keys)) for keys in paths), **kwargs, get_all=False) | |
279 | ||
280 | ||
281 | def dict_get(d, key_or_keys, default=None, skip_false_values=True): | |
282 | for val in map(d.get, variadic(key_or_keys)): | |
283 | if val is not None and (val or not skip_false_values): | |
284 | return val | |
285 | return default |