-
Notifications
You must be signed in to change notification settings - Fork 239
/
base.py
501 lines (399 loc) · 15.1 KB
/
base.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
# Copyright 2018 Iguazio
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import sys
from base64 import b64encode
from os import getenv, path, remove
from tempfile import mktemp
import fsspec
import orjson
import pandas as pd
import pyarrow.parquet as pq
import requests
import urllib3
import mlrun.errors
from mlrun.utils import is_ipython, logger
verify_ssl = False
if not verify_ssl:
urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning)
class FileStats:
def __init__(self, size, modified, content_type=None):
self.size = size
self.modified = modified
self.content_type = content_type
def __repr__(self):
return f"FileStats(size={self.size}, modified={self.modified}, type={self.content_type})"
class DataStore:
def __init__(self, parent, name, kind, endpoint=""):
self._parent = parent
self.kind = kind
self.name = name
self.endpoint = endpoint
self.subpath = ""
self.secret_pfx = ""
self.options = {}
self.from_spec = False
self._filesystem = None
@property
def is_structured(self):
return False
@property
def is_unstructured(self):
return True
@staticmethod
def uri_to_kfp(endpoint, subpath):
raise ValueError("data store doesnt support KFP URLs")
@staticmethod
def uri_to_ipython(endpoint, subpath):
return ""
def get_filesystem(self, silent=True):
"""return fsspec file system object, if supported"""
return None
def supports_isdir(self):
"""Whether the data store supports isdir"""
return True
def _get_secret_or_env(self, key, default=None):
return self._secret(key) or getenv(key, default)
def get_storage_options(self):
"""get fsspec storage options"""
return None
def open(self, filepath, mode):
fs = self.get_filesystem(False)
return fs.open(filepath, mode)
def _join(self, key):
if self.subpath:
return f"{self.subpath}/{key}"
return key
def _secret(self, key):
return self._parent.secret(self.secret_pfx + key)
@property
def url(self):
return f"{self.kind}://{self.endpoint}"
def get(self, key, size=None, offset=0):
pass
def query(self, key, query="", **kwargs):
raise ValueError("data store doesnt support structured queries")
def put(self, key, data, append=False):
pass
def stat(self, key):
pass
def listdir(self, key):
raise ValueError("data store doesnt support listdir")
def download(self, key, target_path):
data = self.get(key)
mode = "wb"
if isinstance(data, str):
mode = "w"
with open(target_path, mode) as fp:
fp.write(data)
fp.close()
def upload(self, key, src_path):
pass
def as_df(
self,
url,
subpath,
columns=None,
df_module=None,
format="",
start_time=None,
end_time=None,
time_column=None,
**kwargs,
):
df_module = df_module or pd
if url.endswith(".csv") or format == "csv":
if columns:
kwargs["usecols"] = columns
reader = df_module.read_csv
elif url.endswith(".parquet") or url.endswith(".pq") or format == "parquet":
if columns:
kwargs["columns"] = columns
def reader(*args, **kwargs):
if start_time or end_time:
if sys.version_info < (3, 7):
raise ValueError(
f"feature not supported for python version {sys.version_info}"
)
from storey.utils import find_filters
dataset = pq.ParquetDataset(url, filesystem=fs)
if dataset.partitions:
partitions = dataset.partitions.partition_names
time_attributes = [
"year",
"month",
"day",
"hour",
"minute",
"second",
]
partitions_time_attributes = [
j for j in time_attributes if j in partitions
]
else:
partitions_time_attributes = []
filters = []
find_filters(
partitions_time_attributes,
start_time,
end_time,
filters,
time_column,
)
kwargs["filters"] = filters
return df_module.read_parquet(*args, **kwargs)
elif url.endswith(".json") or format == "json":
reader = df_module.read_json
else:
raise Exception(f"file type unhandled {url}")
fs = self.get_filesystem()
if fs:
if self.supports_isdir() and fs.isdir(url):
storage_options = self.get_storage_options()
if storage_options:
kwargs["storage_options"] = storage_options
return reader(url, **kwargs)
else:
# If not dir, use fs.open() to avoid regression when pandas < 1.2 and does not
# support the storage_options parameter.
return reader(fs.open(url), **kwargs)
tmp = mktemp()
self.download(self._join(subpath), tmp)
df = reader(tmp, **kwargs)
remove(tmp)
return df
def to_dict(self):
return {
"name": self.name,
"url": f"{self.kind}://{self.endpoint}/{self.subpath}",
"secret_pfx": self.secret_pfx,
"options": self.options,
}
def rm(self, path, recursive=False, maxdepth=None):
self.get_filesystem().rm(path=path, recursive=recursive, maxdepth=maxdepth)
class DataItem:
"""Data input/output class abstracting access to various local/remote data sources
DataItem objects are passed into functions and can be used inside the function, when a function run completes
users can access the run data via the run.artifact(key) which returns a DataItem object.
users can also convert a data url (e.g. s3://bucket/key.csv) to a DataItem using `mlrun.get_dataitem(url)`.
Example::
# using data item inside a function
def my_func(context, data: DataItem):
df = data.as_df()
# reading run results using DataItem (run.artifact())
train_run = train_iris_func.run(inputs={'dataset': dataset},
params={'label_column': 'label'})
train_run.artifact('confusion-matrix').show()
test_set = train_run.artifact('test_set').as_df()
# create and use DataItem from uri
data = mlrun.get_dataitem('http://xyz/data.json').get()
"""
def __init__(
self,
key: str,
store: DataStore,
subpath: str,
url: str = "",
meta=None,
artifact_url=None,
):
self._store = store
self._key = key
self._url = url
self._path = subpath
self._meta = meta
self._artifact_url = artifact_url
self._local_path = ""
@property
def key(self):
"""DataItem key"""
return self._key
@property
def suffix(self):
"""DataItem suffix (file extension) e.g. '.png'"""
_, file_ext = path.splitext(self._path)
return file_ext
@property
def store(self):
"""DataItem store object"""
return self._store
@property
def kind(self):
"""DataItem store kind (file, s3, v3io, ..)"""
return self._store.kind
@property
def meta(self):
"""Artifact Metadata, when the DataItem is read from the artifacts store"""
return self._meta
@property
def artifact_url(self):
"""DataItem artifact url (when its an artifact) or url for simple dataitems"""
return self._artifact_url or self._url
@property
def url(self):
"""DataItem url e.g. /dir/path, s3://bucket/path"""
return self._url
def get(self, size=None, offset=0, encoding=None):
"""read all or a byte range and return the content
:param size: number of bytes to get
:param offset: fetch from offset (in bytes)
:param encoding: encoding (e.g. "utf-8") for converting bytes to str
"""
body = self._store.get(self._path, size=size, offset=offset)
if encoding and isinstance(body, bytes):
body = body.decode(encoding)
return body
def download(self, target_path):
"""download to the target dir/path
:param target_path: local target path for the downloaded item
"""
self._store.download(self._path, target_path)
def put(self, data, append=False):
"""write/upload the data, append is only supported by some datastores
:param data: data (bytes/str) to write
:param append: append data to the end of the object, NOT SUPPORTED BY SOME OBJECT STORES!
"""
self._store.put(self._path, data, append=append)
def upload(self, src_path):
"""upload the source file (src_path)
:param src_path: source file path to read from and upload
"""
self._store.upload(self._path, src_path)
def stat(self):
"""return FileStats class (size, modified, content_type)"""
return self._store.stat(self._path)
def open(self, mode):
"""return fsspec file handler, if supported"""
return self._store.open(self._url, mode)
def ls(self):
"""return a list of child file names"""
return self._store.listdir(self._path)
def listdir(self):
"""return a list of child file names"""
return self._store.listdir(self._path)
def local(self):
"""get the local path of the file, download to tmp first if its a remote object"""
if self.kind == "file":
return self._path
if self._local_path:
return self._local_path
dot = self._path.rfind(".")
self._local_path = mktemp() if dot == -1 else mktemp(self._path[dot:])
logger.info(f"downloading {self.url} to local tmp")
self.download(self._local_path)
return self._local_path
def as_df(
self, columns=None, df_module=None, format="", **kwargs,
):
"""return a dataframe object (generated from the dataitem).
:param columns: optional, list of columns to select
:param df_module: optional, dataframe class (e.g. pd, dd, cudf, ..)
:param format: file format, if not specified it will be deducted from the suffix
"""
return self._store.as_df(
self._url,
self._path,
columns=columns,
df_module=df_module,
format=format,
**kwargs,
)
def show(self, format=None):
"""show the data object content in Jupyter
:param format: format to use (when there is no/wrong suffix), e.g. 'png'
"""
if not is_ipython:
logger.warning(
"Jupyter/IPython was not detected, .show() will only display inside Jupyter"
)
return
from IPython import display
suffix = self.suffix.lower()
if format:
suffix = "." + format
if suffix in [".jpg", ".png", ".gif"]:
display.display(display.Image(self.get(), format=suffix[1:]))
elif suffix in [".htm", ".html"]:
display.display(display.HTML(self.get(encoding="utf-8")))
elif suffix in [".csv", ".pq", ".parquet"]:
display.display(self.as_df())
elif suffix in [".yaml", ".txt", ".py"]:
display.display(display.Pretty(self.get(encoding="utf-8")))
elif suffix == ".json":
display.display(display.JSON(orjson.loads(self.get())))
elif suffix == ".md":
display.display(display.Markdown(self.get(encoding="utf-8")))
else:
logger.error(f"unsupported show() format {suffix} for {self.url}")
def __str__(self):
return self.url
def __repr__(self):
return f"'{self.url}'"
def get_range(size, offset):
byterange = f"bytes={offset}-"
if size:
byterange += str(offset + size)
return byterange
def basic_auth_header(user, password):
username = user.encode("latin1")
password = password.encode("latin1")
base = b64encode(b":".join((username, password))).strip()
authstr = "Basic " + base.decode("ascii")
return {"Authorization": authstr}
def http_get(url, headers=None, auth=None):
try:
response = requests.get(url, headers=headers, auth=auth, verify=verify_ssl)
except OSError as exc:
raise OSError(f"error: cannot connect to {url}: {exc}")
mlrun.errors.raise_for_status(response)
return response.content
def http_head(url, headers=None, auth=None):
try:
response = requests.head(url, headers=headers, auth=auth, verify=verify_ssl)
except OSError as exc:
raise OSError(f"error: cannot connect to {url}: {exc}")
mlrun.errors.raise_for_status(response)
return response.headers
def http_put(url, data, headers=None, auth=None):
try:
response = requests.put(
url, data=data, headers=headers, auth=auth, verify=verify_ssl
)
except OSError as exc:
raise OSError(f"error: cannot connect to {url}: {exc}")
mlrun.errors.raise_for_status(response)
def http_upload(url, file_path, headers=None, auth=None):
with open(file_path, "rb") as data:
http_put(url, data, headers, auth)
class HttpStore(DataStore):
def __init__(self, parent, schema, name, endpoint=""):
super().__init__(parent, name, schema, endpoint)
self.auth = None
def get_filesystem(self, silent=True):
"""return fsspec file system object, if supported"""
if not self._filesystem:
self._filesystem = fsspec.filesystem("http")
return self._filesystem
def supports_isdir(self):
return False
def upload(self, key, src_path):
raise ValueError("unimplemented")
def put(self, key, data, append=False):
raise ValueError("unimplemented")
def get(self, key, size=None, offset=0):
data = http_get(self.url + self._join(key), None, self.auth)
if offset:
data = data[offset:]
if size:
data = data[:size]
return data