/
route_utils.py
581 lines (490 loc) 路 19.9 KB
/
route_utils.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
from __future__ import annotations
import hashlib
import hmac
import json
import shutil
from collections import deque
from dataclasses import dataclass as python_dataclass
from tempfile import NamedTemporaryFile, _TemporaryFileWrapper
from typing import TYPE_CHECKING, AsyncGenerator, BinaryIO, List, Optional, Tuple, Union
import fastapi
import httpx
import multipart
from gradio_client.documentation import document
from multipart.multipart import parse_options_header
from starlette.datastructures import FormData, Headers, UploadFile
from starlette.formparsers import MultiPartException, MultipartPart
from gradio import processing_utils, utils
from gradio.data_classes import PredictBody
from gradio.exceptions import Error
from gradio.helpers import EventData
from gradio.state_holder import SessionState
if TYPE_CHECKING:
from gradio.blocks import Blocks
from gradio.routes import App
class Obj:
"""
Using a class to convert dictionaries into objects. Used by the `Request` class.
Credit: https://www.geeksforgeeks.org/convert-nested-python-dictionary-to-object/
"""
def __init__(self, dict_):
self.__dict__.update(dict_)
for key, value in dict_.items():
if isinstance(value, (dict, list)):
value = Obj(value)
setattr(self, key, value)
def __getitem__(self, item):
return self.__dict__[item]
def __setitem__(self, item, value):
self.__dict__[item] = value
def __iter__(self):
for key, value in self.__dict__.items():
if isinstance(value, Obj):
yield (key, dict(value))
else:
yield (key, value)
def __contains__(self, item) -> bool:
if item in self.__dict__:
return True
for value in self.__dict__.values():
if isinstance(value, Obj) and item in value:
return True
return False
def keys(self):
return self.__dict__.keys()
def values(self):
return self.__dict__.values()
def items(self):
return self.__dict__.items()
def __str__(self) -> str:
return str(self.__dict__)
def __repr__(self) -> str:
return str(self.__dict__)
@document()
class Request:
"""
A Gradio request object that can be used to access the request headers, cookies,
query parameters and other information about the request from within the prediction
function. The class is a thin wrapper around the fastapi.Request class. Attributes
of this class include: `headers`, `client`, `query_params`, and `path_params`. If
auth is enabled, the `username` attribute can be used to get the logged in user.
Example:
import gradio as gr
def echo(text, request: gr.Request):
if request:
print("Request headers dictionary:", request.headers)
print("IP address:", request.client.host)
print("Query parameters:", dict(request.query_params))
return text
io = gr.Interface(echo, "textbox", "textbox").launch()
Demos: request_ip_headers
"""
def __init__(
self,
request: fastapi.Request | None = None,
username: str | None = None,
**kwargs,
):
"""
Can be instantiated with either a fastapi.Request or by manually passing in
attributes (needed for queueing).
Parameters:
request: A fastapi.Request
"""
self.request = request
self.username = username
self.kwargs: dict = kwargs
def dict_to_obj(self, d):
if isinstance(d, dict):
return json.loads(json.dumps(d), object_hook=Obj)
else:
return d
def __getattr__(self, name):
if self.request:
return self.dict_to_obj(getattr(self.request, name))
else:
try:
obj = self.kwargs[name]
except KeyError as ke:
raise AttributeError(
f"'Request' object has no attribute '{name}'"
) from ke
return self.dict_to_obj(obj)
class FnIndexInferError(Exception):
pass
def infer_fn_index(app: App, api_name: str, body: PredictBody) -> int:
if body.fn_index is None:
for i, fn in enumerate(app.get_blocks().dependencies):
if fn["api_name"] == api_name:
return i
raise FnIndexInferError(f"Could not infer fn_index for api_name {api_name}.")
else:
return body.fn_index
def compile_gr_request(
app: App,
body: PredictBody,
fn_index_inferred: int,
username: Optional[str],
request: Optional[fastapi.Request],
):
# If this fn_index cancels jobs, then the only input we need is the
# current session hash
if app.get_blocks().dependencies[fn_index_inferred]["cancels"]:
body.data = [body.session_hash]
if body.request:
if body.batched:
gr_request = [Request(username=username, request=request)]
else:
gr_request = Request(username=username, request=body.request)
else:
if request is None:
raise ValueError("request must be provided if body.request is None")
gr_request = Request(username=username, request=request)
return gr_request
def restore_session_state(app: App, body: PredictBody):
event_id = body.event_id
session_hash = getattr(body, "session_hash", None)
if session_hash is not None:
session_state = app.state_holder[session_hash]
# The should_reset set keeps track of the fn_indices
# that have been cancelled. When a job is cancelled,
# the /reset route will mark the jobs as having been reset.
# That way if the cancel job finishes BEFORE the job being cancelled
# the job being cancelled will not overwrite the state of the iterator.
if event_id is None:
iterator = None
elif event_id in app.iterators_to_reset:
iterator = None
app.iterators_to_reset.remove(event_id)
else:
iterator = app.iterators.get(event_id)
else:
session_state = SessionState(app.get_blocks())
iterator = None
return session_state, iterator
def prepare_event_data(
blocks: Blocks,
body: PredictBody,
) -> EventData:
target = body.trigger_id
event_data = EventData(
blocks.blocks.get(target) if target else None,
body.event_data,
)
return event_data
async def call_process_api(
app: App,
body: PredictBody,
gr_request: Union[Request, list[Request]],
fn_index_inferred: int,
):
session_state, iterator = restore_session_state(app=app, body=body)
dependency = app.get_blocks().dependencies[fn_index_inferred]
event_data = prepare_event_data(app.get_blocks(), body)
event_id = body.event_id
session_hash = getattr(body, "session_hash", None)
inputs = body.data
batch_in_single_out = not body.batched and dependency["batch"]
if batch_in_single_out:
inputs = [inputs]
try:
with utils.MatplotlibBackendMananger():
output = await app.get_blocks().process_api(
fn_index=fn_index_inferred,
inputs=inputs,
request=gr_request,
state=session_state,
iterator=iterator,
session_hash=session_hash,
event_id=event_id,
event_data=event_data,
in_event_listener=True,
)
iterator = output.pop("iterator", None)
if event_id is not None:
app.iterators[event_id] = iterator # type: ignore
if isinstance(output, Error):
raise output
except BaseException:
iterator = app.iterators.get(event_id) if event_id is not None else None
if iterator is not None: # close off any streams that are still open
run_id = id(iterator)
pending_streams: dict[int, list] = (
app.get_blocks().pending_streams[session_hash].get(run_id, {})
)
for stream in pending_streams.values():
stream.append(None)
raise
if batch_in_single_out:
output["data"] = output["data"][0]
return output
def get_root_url(
request: fastapi.Request, route_path: str, root_path: str | None
) -> str:
"""
Gets the root url of the request, stripping off any query parameters, the route_path, and trailing slashes.
Also ensures that the root url is https if the request is https. If root_path is provided, it is appended to the root url.
The final root url will not have a trailing slash.
"""
root_url = str(request.url)
root_url = httpx.URL(root_url)
root_url = root_url.copy_with(query=None)
root_url = str(root_url).rstrip("/")
if request.headers.get("x-forwarded-proto") == "https":
root_url = root_url.replace("http://", "https://")
route_path = route_path.rstrip("/")
if len(route_path) > 0:
root_url = root_url[: -len(route_path)]
return (root_url.rstrip("/") + (root_path or "")).rstrip("/")
def _user_safe_decode(src: bytes, codec: str) -> str:
try:
return src.decode(codec)
except (UnicodeDecodeError, LookupError):
return src.decode("latin-1")
class GradioUploadFile(UploadFile):
"""UploadFile with a sha attribute."""
def __init__(
self,
file: BinaryIO,
*,
size: int | None = None,
filename: str | None = None,
headers: Headers | None = None,
) -> None:
super().__init__(file, size=size, filename=filename, headers=headers)
self.sha = hashlib.sha1()
@python_dataclass(frozen=True)
class FileUploadProgressUnit:
filename: str
chunk_size: int
@python_dataclass
class FileUploadProgressTracker:
deque: deque[FileUploadProgressUnit]
is_done: bool
class FileUploadProgressNotTrackedError(Exception):
pass
class FileUploadProgressNotQueuedError(Exception):
pass
class FileUploadProgress:
def __init__(self) -> None:
self._statuses: dict[str, FileUploadProgressTracker] = {}
def track(self, upload_id: str):
if upload_id not in self._statuses:
self._statuses[upload_id] = FileUploadProgressTracker(deque(), False)
def append(self, upload_id: str, filename: str, message_bytes: bytes):
if upload_id not in self._statuses:
self.track(upload_id)
queue = self._statuses[upload_id].deque
if len(queue) == 0:
queue.append(FileUploadProgressUnit(filename, len(message_bytes)))
else:
last_unit = queue.popleft()
if last_unit.filename != filename:
queue.append(FileUploadProgressUnit(filename, len(message_bytes)))
else:
queue.append(
FileUploadProgressUnit(
filename,
last_unit.chunk_size + len(message_bytes),
)
)
def set_done(self, upload_id: str):
if upload_id not in self._statuses:
self.track(upload_id)
self._statuses[upload_id].is_done = True
def is_done(self, upload_id: str):
if upload_id not in self._statuses:
raise FileUploadProgressNotTrackedError()
return self._statuses[upload_id].is_done
def stop_tracking(self, upload_id: str):
if upload_id in self._statuses:
del self._statuses[upload_id]
def pop(self, upload_id: str) -> FileUploadProgressUnit:
if upload_id not in self._statuses:
raise FileUploadProgressNotTrackedError()
try:
return self._statuses[upload_id].deque.pop()
except IndexError as e:
raise FileUploadProgressNotQueuedError() from e
class GradioMultiPartParser:
"""Vendored from starlette.MultipartParser.
Thanks starlette!
Made the following modifications
- Use GradioUploadFile instead of UploadFile
- Use NamedTemporaryFile instead of SpooledTemporaryFile
- Compute hash of data as the request is streamed
"""
max_file_size = 1024 * 1024
def __init__(
self,
headers: Headers,
stream: AsyncGenerator[bytes, None],
*,
max_files: Union[int, float] = 1000,
max_fields: Union[int, float] = 1000,
upload_id: str | None = None,
upload_progress: FileUploadProgress | None = None,
) -> None:
assert (
multipart is not None
), "The `python-multipart` library must be installed to use form parsing."
self.headers = headers
self.stream = stream
self.max_files = max_files
self.max_fields = max_fields
self.items: List[Tuple[str, Union[str, UploadFile]]] = []
self.upload_id = upload_id
self.upload_progress = upload_progress
self._current_files = 0
self._current_fields = 0
self._current_partial_header_name: bytes = b""
self._current_partial_header_value: bytes = b""
self._current_part = MultipartPart()
self._charset = ""
self._file_parts_to_write: List[Tuple[MultipartPart, bytes]] = []
self._file_parts_to_finish: List[MultipartPart] = []
self._files_to_close_on_error: List[_TemporaryFileWrapper] = []
def on_part_begin(self) -> None:
self._current_part = MultipartPart()
def on_part_data(self, data: bytes, start: int, end: int) -> None:
message_bytes = data[start:end]
if self.upload_progress is not None:
self.upload_progress.append(
self.upload_id, # type: ignore
self._current_part.file.filename, # type: ignore
message_bytes,
)
if self._current_part.file is None:
self._current_part.data += message_bytes
else:
self._file_parts_to_write.append((self._current_part, message_bytes))
def on_part_end(self) -> None:
if self._current_part.file is None:
self.items.append(
(
self._current_part.field_name,
_user_safe_decode(self._current_part.data, self._charset),
)
)
else:
self._file_parts_to_finish.append(self._current_part)
# The file can be added to the items right now even though it's not
# finished yet, because it will be finished in the `parse()` method, before
# self.items is used in the return value.
self.items.append((self._current_part.field_name, self._current_part.file))
def on_header_field(self, data: bytes, start: int, end: int) -> None:
self._current_partial_header_name += data[start:end]
def on_header_value(self, data: bytes, start: int, end: int) -> None:
self._current_partial_header_value += data[start:end]
def on_header_end(self) -> None:
field = self._current_partial_header_name.lower()
if field == b"content-disposition":
self._current_part.content_disposition = self._current_partial_header_value
self._current_part.item_headers.append(
(field, self._current_partial_header_value)
)
self._current_partial_header_name = b""
self._current_partial_header_value = b""
def on_headers_finished(self) -> None:
disposition, options = parse_options_header(
self._current_part.content_disposition
)
try:
self._current_part.field_name = _user_safe_decode(
options[b"name"], self._charset
)
except KeyError as e:
raise MultiPartException(
'The Content-Disposition header field "name" must be ' "provided."
) from e
if b"filename" in options:
self._current_files += 1
if self._current_files > self.max_files:
raise MultiPartException(
f"Too many files. Maximum number of files is {self.max_files}."
)
filename = _user_safe_decode(options[b"filename"], self._charset)
tempfile = NamedTemporaryFile(delete=False)
self._files_to_close_on_error.append(tempfile)
self._current_part.file = GradioUploadFile(
file=tempfile, # type: ignore[arg-type]
size=0,
filename=filename,
headers=Headers(raw=self._current_part.item_headers),
)
else:
self._current_fields += 1
if self._current_fields > self.max_fields:
raise MultiPartException(
f"Too many fields. Maximum number of fields is {self.max_fields}."
)
self._current_part.file = None
def on_end(self) -> None:
pass
async def parse(self) -> FormData:
# Parse the Content-Type header to get the multipart boundary.
_, params = parse_options_header(self.headers["Content-Type"])
charset = params.get(b"charset", "utf-8")
if isinstance(charset, bytes):
charset = charset.decode("latin-1")
self._charset = charset
try:
boundary = params[b"boundary"]
except KeyError as e:
raise MultiPartException("Missing boundary in multipart.") from e
# Callbacks dictionary.
callbacks = {
"on_part_begin": self.on_part_begin,
"on_part_data": self.on_part_data,
"on_part_end": self.on_part_end,
"on_header_field": self.on_header_field,
"on_header_value": self.on_header_value,
"on_header_end": self.on_header_end,
"on_headers_finished": self.on_headers_finished,
"on_end": self.on_end,
}
# Create the parser.
parser = multipart.MultipartParser(boundary, callbacks)
try:
# Feed the parser with data from the request.
async for chunk in self.stream:
parser.write(chunk)
# Write file data, it needs to use await with the UploadFile methods
# that call the corresponding file methods *in a threadpool*,
# otherwise, if they were called directly in the callback methods above
# (regular, non-async functions), that would block the event loop in
# the main thread.
for part, data in self._file_parts_to_write:
assert part.file # for type checkers
await part.file.write(data)
part.file.sha.update(data) # type: ignore
for part in self._file_parts_to_finish:
assert part.file # for type checkers
await part.file.seek(0)
self._file_parts_to_write.clear()
self._file_parts_to_finish.clear()
except MultiPartException as exc:
# Close all the files if there was an error.
for file in self._files_to_close_on_error:
file.close()
raise exc
parser.finalize()
if self.upload_progress is not None:
self.upload_progress.set_done(self.upload_id) # type: ignore
return FormData(self.items)
def move_uploaded_files_to_cache(files: list[str], destinations: list[str]) -> None:
for file, dest in zip(files, destinations):
shutil.move(file, dest)
def update_root_in_config(config: dict, root: str) -> dict:
"""
Updates the root "key" in the config dictionary to the new root url. If the
root url has changed, all of the urls in the config that correspond to component
file urls are updated to use the new root url.
"""
previous_root = config.get("root")
if previous_root is None or previous_root != root:
config["root"] = root
config = processing_utils.add_root_url(config, root, previous_root)
return config
def compare_passwords_securely(input_password: str, correct_password: str) -> bool:
return hmac.compare_digest(input_password.encode(), correct_password.encode())