openai/openai-python
Publicmirrored from https://github.com/openai/openai-pythonAvailable
src/openai/_utils/_utils.py
422lines · modecode
| 1 | from __future__ import annotations |
| 2 | |
| 3 | import os |
| 4 | import re |
| 5 | import inspect |
| 6 | import functools |
| 7 | from typing import ( |
| 8 | Any, |
| 9 | Tuple, |
| 10 | Mapping, |
| 11 | TypeVar, |
| 12 | Callable, |
| 13 | Iterable, |
| 14 | Sequence, |
| 15 | cast, |
| 16 | overload, |
| 17 | ) |
| 18 | from pathlib import Path |
| 19 | from datetime import date, datetime |
| 20 | from typing_extensions import TypeGuard |
| 21 | |
| 22 | import sniffio |
| 23 | |
| 24 | from .._types import NotGiven, FileTypes, NotGivenOr, HeadersLike |
| 25 | from .._compat import parse_date as parse_date, parse_datetime as parse_datetime |
| 26 | |
| 27 | _T = TypeVar("_T") |
| 28 | _TupleT = TypeVar("_TupleT", bound=Tuple[object, ...]) |
| 29 | _MappingT = TypeVar("_MappingT", bound=Mapping[str, object]) |
| 30 | _SequenceT = TypeVar("_SequenceT", bound=Sequence[object]) |
| 31 | CallableT = TypeVar("CallableT", bound=Callable[..., Any]) |
| 32 | |
| 33 | |
| 34 | def flatten(t: Iterable[Iterable[_T]]) -> list[_T]: |
| 35 | return [item for sublist in t for item in sublist] |
| 36 | |
| 37 | |
| 38 | def extract_files( |
| 39 | # TODO: this needs to take Dict but variance issues..... |
| 40 | # create protocol type ? |
| 41 | query: Mapping[str, object], |
| 42 | *, |
| 43 | paths: Sequence[Sequence[str]], |
| 44 | ) -> list[tuple[str, FileTypes]]: |
| 45 | """Recursively extract files from the given dictionary based on specified paths. |
| 46 | |
| 47 | A path may look like this ['foo', 'files', '<array>', 'data']. |
| 48 | |
| 49 | Note: this mutates the given dictionary. |
| 50 | """ |
| 51 | files: list[tuple[str, FileTypes]] = [] |
| 52 | for path in paths: |
| 53 | files.extend(_extract_items(query, path, index=0, flattened_key=None)) |
| 54 | return files |
| 55 | |
| 56 | |
| 57 | def _extract_items( |
| 58 | obj: object, |
| 59 | path: Sequence[str], |
| 60 | *, |
| 61 | index: int, |
| 62 | flattened_key: str | None, |
| 63 | ) -> list[tuple[str, FileTypes]]: |
| 64 | try: |
| 65 | key = path[index] |
| 66 | except IndexError: |
| 67 | if isinstance(obj, NotGiven): |
| 68 | # no value was provided - we can safely ignore |
| 69 | return [] |
| 70 | |
| 71 | # cyclical import |
| 72 | from .._files import assert_is_file_content |
| 73 | |
| 74 | # We have exhausted the path, return the entry we found. |
| 75 | assert flattened_key is not None |
| 76 | |
| 77 | if is_list(obj): |
| 78 | files: list[tuple[str, FileTypes]] = [] |
| 79 | for entry in obj: |
| 80 | assert_is_file_content(entry, key=flattened_key + "[]" if flattened_key else "") |
| 81 | files.append((flattened_key + "[]", cast(FileTypes, entry))) |
| 82 | return files |
| 83 | |
| 84 | assert_is_file_content(obj, key=flattened_key) |
| 85 | return [(flattened_key, cast(FileTypes, obj))] |
| 86 | |
| 87 | index += 1 |
| 88 | if is_dict(obj): |
| 89 | try: |
| 90 | # We are at the last entry in the path so we must remove the field |
| 91 | if (len(path)) == index: |
| 92 | item = obj.pop(key) |
| 93 | else: |
| 94 | item = obj[key] |
| 95 | except KeyError: |
| 96 | # Key was not present in the dictionary, this is not indicative of an error |
| 97 | # as the given path may not point to a required field. We also do not want |
| 98 | # to enforce required fields as the API may differ from the spec in some cases. |
| 99 | return [] |
| 100 | if flattened_key is None: |
| 101 | flattened_key = key |
| 102 | else: |
| 103 | flattened_key += f"[{key}]" |
| 104 | return _extract_items( |
| 105 | item, |
| 106 | path, |
| 107 | index=index, |
| 108 | flattened_key=flattened_key, |
| 109 | ) |
| 110 | elif is_list(obj): |
| 111 | if key != "<array>": |
| 112 | return [] |
| 113 | |
| 114 | return flatten( |
| 115 | [ |
| 116 | _extract_items( |
| 117 | item, |
| 118 | path, |
| 119 | index=index, |
| 120 | flattened_key=flattened_key + "[]" if flattened_key is not None else "[]", |
| 121 | ) |
| 122 | for item in obj |
| 123 | ] |
| 124 | ) |
| 125 | |
| 126 | # Something unexpected was passed, just ignore it. |
| 127 | return [] |
| 128 | |
| 129 | |
| 130 | def is_given(obj: NotGivenOr[_T]) -> TypeGuard[_T]: |
| 131 | return not isinstance(obj, NotGiven) |
| 132 | |
| 133 | |
| 134 | # Type safe methods for narrowing types with TypeVars. |
| 135 | # The default narrowing for isinstance(obj, dict) is dict[unknown, unknown], |
| 136 | # however this cause Pyright to rightfully report errors. As we know we don't |
| 137 | # care about the contained types we can safely use `object` in it's place. |
| 138 | # |
| 139 | # There are two separate functions defined, `is_*` and `is_*_t` for different use cases. |
| 140 | # `is_*` is for when you're dealing with an unknown input |
| 141 | # `is_*_t` is for when you're narrowing a known union type to a specific subset |
| 142 | |
| 143 | |
| 144 | def is_tuple(obj: object) -> TypeGuard[tuple[object, ...]]: |
| 145 | return isinstance(obj, tuple) |
| 146 | |
| 147 | |
| 148 | def is_tuple_t(obj: _TupleT | object) -> TypeGuard[_TupleT]: |
| 149 | return isinstance(obj, tuple) |
| 150 | |
| 151 | |
| 152 | def is_sequence(obj: object) -> TypeGuard[Sequence[object]]: |
| 153 | return isinstance(obj, Sequence) |
| 154 | |
| 155 | |
| 156 | def is_sequence_t(obj: _SequenceT | object) -> TypeGuard[_SequenceT]: |
| 157 | return isinstance(obj, Sequence) |
| 158 | |
| 159 | |
| 160 | def is_mapping(obj: object) -> TypeGuard[Mapping[str, object]]: |
| 161 | return isinstance(obj, Mapping) |
| 162 | |
| 163 | |
| 164 | def is_mapping_t(obj: _MappingT | object) -> TypeGuard[_MappingT]: |
| 165 | return isinstance(obj, Mapping) |
| 166 | |
| 167 | |
| 168 | def is_dict(obj: object) -> TypeGuard[dict[object, object]]: |
| 169 | return isinstance(obj, dict) |
| 170 | |
| 171 | |
| 172 | def is_list(obj: object) -> TypeGuard[list[object]]: |
| 173 | return isinstance(obj, list) |
| 174 | |
| 175 | |
| 176 | def is_iterable(obj: object) -> TypeGuard[Iterable[object]]: |
| 177 | return isinstance(obj, Iterable) |
| 178 | |
| 179 | |
| 180 | def deepcopy_minimal(item: _T) -> _T: |
| 181 | """Minimal reimplementation of copy.deepcopy() that will only copy certain object types: |
| 182 | |
| 183 | - mappings, e.g. `dict` |
| 184 | - list |
| 185 | |
| 186 | This is done for performance reasons. |
| 187 | """ |
| 188 | if is_mapping(item): |
| 189 | return cast(_T, {k: deepcopy_minimal(v) for k, v in item.items()}) |
| 190 | if is_list(item): |
| 191 | return cast(_T, [deepcopy_minimal(entry) for entry in item]) |
| 192 | return item |
| 193 | |
| 194 | |
| 195 | # copied from https://github.com/Rapptz/RoboDanny |
| 196 | def human_join(seq: Sequence[str], *, delim: str = ", ", final: str = "or") -> str: |
| 197 | size = len(seq) |
| 198 | if size == 0: |
| 199 | return "" |
| 200 | |
| 201 | if size == 1: |
| 202 | return seq[0] |
| 203 | |
| 204 | if size == 2: |
| 205 | return f"{seq[0]} {final} {seq[1]}" |
| 206 | |
| 207 | return delim.join(seq[:-1]) + f" {final} {seq[-1]}" |
| 208 | |
| 209 | |
| 210 | def quote(string: str) -> str: |
| 211 | """Add single quotation marks around the given string. Does *not* do any escaping.""" |
| 212 | return f"'{string}'" |
| 213 | |
| 214 | |
| 215 | def required_args(*variants: Sequence[str]) -> Callable[[CallableT], CallableT]: |
| 216 | """Decorator to enforce a given set of arguments or variants of arguments are passed to the decorated function. |
| 217 | |
| 218 | Useful for enforcing runtime validation of overloaded functions. |
| 219 | |
| 220 | Example usage: |
| 221 | ```py |
| 222 | @overload |
| 223 | def foo(*, a: str) -> str: ... |
| 224 | |
| 225 | |
| 226 | @overload |
| 227 | def foo(*, b: bool) -> str: ... |
| 228 | |
| 229 | |
| 230 | # This enforces the same constraints that a static type checker would |
| 231 | # i.e. that either a or b must be passed to the function |
| 232 | @required_args(["a"], ["b"]) |
| 233 | def foo(*, a: str | None = None, b: bool | None = None) -> str: ... |
| 234 | ``` |
| 235 | """ |
| 236 | |
| 237 | def inner(func: CallableT) -> CallableT: |
| 238 | params = inspect.signature(func).parameters |
| 239 | positional = [ |
| 240 | name |
| 241 | for name, param in params.items() |
| 242 | if param.kind |
| 243 | in { |
| 244 | param.POSITIONAL_ONLY, |
| 245 | param.POSITIONAL_OR_KEYWORD, |
| 246 | } |
| 247 | ] |
| 248 | |
| 249 | @functools.wraps(func) |
| 250 | def wrapper(*args: object, **kwargs: object) -> object: |
| 251 | given_params: set[str] = set() |
| 252 | for i, _ in enumerate(args): |
| 253 | try: |
| 254 | given_params.add(positional[i]) |
| 255 | except IndexError: |
| 256 | raise TypeError( |
| 257 | f"{func.__name__}() takes {len(positional)} argument(s) but {len(args)} were given" |
| 258 | ) from None |
| 259 | |
| 260 | for key in kwargs.keys(): |
| 261 | given_params.add(key) |
| 262 | |
| 263 | for variant in variants: |
| 264 | matches = all((param in given_params for param in variant)) |
| 265 | if matches: |
| 266 | break |
| 267 | else: # no break |
| 268 | if len(variants) > 1: |
| 269 | variations = human_join( |
| 270 | ["(" + human_join([quote(arg) for arg in variant], final="and") + ")" for variant in variants] |
| 271 | ) |
| 272 | msg = f"Missing required arguments; Expected either {variations} arguments to be given" |
| 273 | else: |
| 274 | assert len(variants) > 0 |
| 275 | |
| 276 | # TODO: this error message is not deterministic |
| 277 | missing = list(set(variants[0]) - given_params) |
| 278 | if len(missing) > 1: |
| 279 | msg = f"Missing required arguments: {human_join([quote(arg) for arg in missing])}" |
| 280 | else: |
| 281 | msg = f"Missing required argument: {quote(missing[0])}" |
| 282 | raise TypeError(msg) |
| 283 | return func(*args, **kwargs) |
| 284 | |
| 285 | return wrapper # type: ignore |
| 286 | |
| 287 | return inner |
| 288 | |
| 289 | |
| 290 | _K = TypeVar("_K") |
| 291 | _V = TypeVar("_V") |
| 292 | |
| 293 | |
| 294 | @overload |
| 295 | def strip_not_given(obj: None) -> None: ... |
| 296 | |
| 297 | |
| 298 | @overload |
| 299 | def strip_not_given(obj: Mapping[_K, _V | NotGiven]) -> dict[_K, _V]: ... |
| 300 | |
| 301 | |
| 302 | @overload |
| 303 | def strip_not_given(obj: object) -> object: ... |
| 304 | |
| 305 | |
| 306 | def strip_not_given(obj: object | None) -> object: |
| 307 | """Remove all top-level keys where their values are instances of `NotGiven`""" |
| 308 | if obj is None: |
| 309 | return None |
| 310 | |
| 311 | if not is_mapping(obj): |
| 312 | return obj |
| 313 | |
| 314 | return {key: value for key, value in obj.items() if not isinstance(value, NotGiven)} |
| 315 | |
| 316 | |
| 317 | def coerce_integer(val: str) -> int: |
| 318 | return int(val, base=10) |
| 319 | |
| 320 | |
| 321 | def coerce_float(val: str) -> float: |
| 322 | return float(val) |
| 323 | |
| 324 | |
| 325 | def coerce_boolean(val: str) -> bool: |
| 326 | return val == "true" or val == "1" or val == "on" |
| 327 | |
| 328 | |
| 329 | def maybe_coerce_integer(val: str | None) -> int | None: |
| 330 | if val is None: |
| 331 | return None |
| 332 | return coerce_integer(val) |
| 333 | |
| 334 | |
| 335 | def maybe_coerce_float(val: str | None) -> float | None: |
| 336 | if val is None: |
| 337 | return None |
| 338 | return coerce_float(val) |
| 339 | |
| 340 | |
| 341 | def maybe_coerce_boolean(val: str | None) -> bool | None: |
| 342 | if val is None: |
| 343 | return None |
| 344 | return coerce_boolean(val) |
| 345 | |
| 346 | |
| 347 | def removeprefix(string: str, prefix: str) -> str: |
| 348 | """Remove a prefix from a string. |
| 349 | |
| 350 | Backport of `str.removeprefix` for Python < 3.9 |
| 351 | """ |
| 352 | if string.startswith(prefix): |
| 353 | return string[len(prefix) :] |
| 354 | return string |
| 355 | |
| 356 | |
| 357 | def removesuffix(string: str, suffix: str) -> str: |
| 358 | """Remove a suffix from a string. |
| 359 | |
| 360 | Backport of `str.removesuffix` for Python < 3.9 |
| 361 | """ |
| 362 | if string.endswith(suffix): |
| 363 | return string[: -len(suffix)] |
| 364 | return string |
| 365 | |
| 366 | |
| 367 | def file_from_path(path: str) -> FileTypes: |
| 368 | contents = Path(path).read_bytes() |
| 369 | file_name = os.path.basename(path) |
| 370 | return (file_name, contents) |
| 371 | |
| 372 | |
| 373 | def get_required_header(headers: HeadersLike, header: str) -> str: |
| 374 | lower_header = header.lower() |
| 375 | if is_mapping_t(headers): |
| 376 | # mypy doesn't understand the type narrowing here |
| 377 | for k, v in headers.items(): # type: ignore |
| 378 | if k.lower() == lower_header and isinstance(v, str): |
| 379 | return v |
| 380 | |
| 381 | # to deal with the case where the header looks like Stainless-Event-Id |
| 382 | intercaps_header = re.sub(r"([^\w])(\w)", lambda pat: pat.group(1) + pat.group(2).upper(), header.capitalize()) |
| 383 | |
| 384 | for normalized_header in [header, lower_header, header.upper(), intercaps_header]: |
| 385 | value = headers.get(normalized_header) |
| 386 | if value: |
| 387 | return value |
| 388 | |
| 389 | raise ValueError(f"Could not find {header} header") |
| 390 | |
| 391 | |
| 392 | def get_async_library() -> str: |
| 393 | try: |
| 394 | return sniffio.current_async_library() |
| 395 | except Exception: |
| 396 | return "false" |
| 397 | |
| 398 | |
| 399 | def lru_cache(*, maxsize: int | None = 128) -> Callable[[CallableT], CallableT]: |
| 400 | """A version of functools.lru_cache that retains the type signature |
| 401 | for the wrapped function arguments. |
| 402 | """ |
| 403 | wrapper = functools.lru_cache( # noqa: TID251 |
| 404 | maxsize=maxsize, |
| 405 | ) |
| 406 | return cast(Any, wrapper) # type: ignore[no-any-return] |
| 407 | |
| 408 | |
| 409 | def json_safe(data: object) -> object: |
| 410 | """Translates a mapping / sequence recursively in the same fashion |
| 411 | as `pydantic` v2's `model_dump(mode="json")`. |
| 412 | """ |
| 413 | if is_mapping(data): |
| 414 | return {json_safe(key): json_safe(value) for key, value in data.items()} |
| 415 | |
| 416 | if is_iterable(data) and not isinstance(data, (str, bytes, bytearray)): |
| 417 | return [json_safe(item) for item in data] |
| 418 | |
| 419 | if isinstance(data, (datetime, date)): |
| 420 | return data.isoformat() |
| 421 | |
| 422 | return data |
| 423 | |