This repository has been archived on 2020-03-24. You can view files and clone it, but cannot push or open issues or pull requests.
quay/test/test_cloud_storage.py

234 lines
8.7 KiB
Python

import unittest
import moto
import boto
import os
from storage import S3Storage, StorageContext
from storage.cloud import _CloudStorage, _PartUploadMetadata
from storage.cloud import _CHUNKS_KEY
from StringIO import StringIO
_TEST_CONTENT = os.urandom(1024)
_TEST_BUCKET = 'some_bucket'
_TEST_USER = 'someuser'
_TEST_PASSWORD = 'somepassword'
_TEST_PATH = 'some/cool/path'
_TEST_CONTEXT = StorageContext('nyc', None, None, None, None)
class TestCloudStorage(unittest.TestCase):
def setUp(self):
self.mock = moto.mock_s3()
self.mock.start()
# Create a test bucket and put some test content.
boto.connect_s3().create_bucket(_TEST_BUCKET)
self.engine = S3Storage(_TEST_CONTEXT, 'some/path', _TEST_BUCKET, _TEST_USER, _TEST_PASSWORD)
self.engine.put_content(_TEST_PATH, _TEST_CONTENT)
def tearDown(self):
self.mock.stop()
def test_basicop(self):
# Ensure the content exists.
self.assertTrue(self.engine.exists(_TEST_PATH))
# Verify it can be retrieved.
self.assertEquals(_TEST_CONTENT, self.engine.get_content(_TEST_PATH))
# Retrieve a checksum for the content.
self.engine.get_checksum(_TEST_PATH)
# Remove the file.
self.engine.remove(_TEST_PATH)
# Ensure it no longer exists.
with self.assertRaises(IOError):
self.engine.get_content(_TEST_PATH)
with self.assertRaises(IOError):
self.engine.get_checksum(_TEST_PATH)
self.assertFalse(self.engine.exists(_TEST_PATH))
def test_copy_samecreds(self):
# Copy the content to another engine.
another_engine = S3Storage(_TEST_CONTEXT, 'another/path', _TEST_BUCKET, _TEST_USER,
_TEST_PASSWORD)
self.engine.copy_to(another_engine, _TEST_PATH)
# Verify it can be retrieved.
self.assertEquals(_TEST_CONTENT, another_engine.get_content(_TEST_PATH))
def test_copy_differentcreds(self):
# Copy the content to another engine.
another_engine = S3Storage(_TEST_CONTEXT, 'another/path', 'another_bucket', 'blech',
'password')
boto.connect_s3().create_bucket('another_bucket')
self.engine.copy_to(another_engine, _TEST_PATH)
# Verify it can be retrieved.
self.assertEquals(_TEST_CONTENT, another_engine.get_content(_TEST_PATH))
def test_stream_read(self):
# Read the streaming content.
data = ''.join(self.engine.stream_read(_TEST_PATH))
self.assertEquals(_TEST_CONTENT, data)
def test_stream_read_file(self):
with self.engine.stream_read_file(_TEST_PATH) as f:
self.assertEquals(_TEST_CONTENT, f.read())
def test_stream_write(self):
new_data = os.urandom(4096)
self.engine.stream_write(_TEST_PATH, StringIO(new_data), content_type='Cool/Type')
self.assertEquals(new_data, self.engine.get_content(_TEST_PATH))
def test_chunked_upload_no_chunks(self):
self._chunk_upload_test(0)
def test_chunked_upload_single_chunk(self):
self._chunk_upload_test(1)
def test_chunked_upload_multiple_chunks(self):
self._chunk_upload_test(50)
def test_chunked_upload_single_chunk_client_side(self):
self._chunk_upload_test(1, force_client_side=True)
def test_chunked_upload_multiple_chunks_client_side(self):
self._chunk_upload_test(50, force_client_side=True)
def _chunk_upload_test(self, chunk_count, force_client_side=False):
upload_id, metadata = self.engine.initiate_chunked_upload()
final_data = ''
for index in range(0, chunk_count):
chunk_data = os.urandom(1024)
final_data = final_data + chunk_data
bytes_written, new_metadata, error = self.engine.stream_upload_chunk(upload_id, 0,
len(chunk_data),
StringIO(chunk_data),
metadata)
metadata = new_metadata
self.assertEquals(len(chunk_data), bytes_written)
self.assertIsNone(error)
self.assertEquals(index + 1, len(metadata[_CHUNKS_KEY]))
# Complete the chunked upload.
self.engine.complete_chunked_upload(upload_id, 'some/chunked/path', metadata,
force_client_side=force_client_side)
# Ensure the file contents are valid.
self.assertEquals(final_data, self.engine.get_content('some/chunked/path'))
def test_cancel_chunked_upload_single_chunk(self):
self._cancel_chunked_upload_test(1)
def test_cancel_chunked_upload_multiple_chunks(self):
self._cancel_chunked_upload_test(50)
def _cancel_chunked_upload_test(self, chunk_count):
upload_id, metadata = self.engine.initiate_chunked_upload()
for _ in range(0, chunk_count):
chunk_data = os.urandom(1024)
_, new_metadata, _ = self.engine.stream_upload_chunk(upload_id, 0,
len(chunk_data),
StringIO(chunk_data),
metadata)
metadata = new_metadata
# Cancel the upload.
self.engine.cancel_chunked_upload(upload_id, metadata)
# Ensure all chunks were deleted.
for chunk in metadata[_CHUNKS_KEY]:
self.assertFalse(self.engine.exists(chunk.path))
def test_large_chunks_upload(self):
# Make the max chunk size much smaller for testing.
self.engine.maximum_chunk_size = self.engine.minimum_chunk_size * 2
upload_id, metadata = self.engine.initiate_chunked_upload()
# Write a "super large" chunk, to ensure that it is broken into smaller chunks.
chunk_data = os.urandom(int(self.engine.maximum_chunk_size * 2.5))
bytes_written, new_metadata, _ = self.engine.stream_upload_chunk(upload_id, 0,
-1,
StringIO(chunk_data),
metadata)
self.assertEquals(bytes_written, len(chunk_data))
# Complete the chunked upload.
self.engine.complete_chunked_upload(upload_id, 'some/chunked/path', new_metadata)
# Ensure the file contents are valid.
self.assertEquals(len(self.engine.get_content('some/chunked/path')), len(chunk_data))
self.assertEquals(chunk_data, self.engine.get_content('some/chunked/path'))
def test_large_chunks_with_ragged_edge(self):
# Make the max chunk size much smaller for testing and force it to have a ragged edge.
self.engine.maximum_chunk_size = self.engine.minimum_chunk_size * 2 + 10
upload_id, metadata = self.engine.initiate_chunked_upload()
# Write a few "super large" chunks, to ensure that it is broken into smaller chunks.
all_data = ''
for _ in range(0, 2):
chunk_data = os.urandom(int(self.engine.maximum_chunk_size) + 20)
bytes_written, new_metadata, _ = self.engine.stream_upload_chunk(upload_id, 0,
-1,
StringIO(chunk_data),
metadata)
self.assertEquals(bytes_written, len(chunk_data))
all_data = all_data + chunk_data
metadata = new_metadata
# Complete the chunked upload.
self.engine.complete_chunked_upload(upload_id, 'some/chunked/path', new_metadata)
# Ensure the file contents are valid.
self.assertEquals(len(self.engine.get_content('some/chunked/path')), len(all_data))
self.assertEquals(all_data, self.engine.get_content('some/chunked/path'))
def assertRechunked(self, chunk, max_size, *args):
rechunked = list(_CloudStorage._rechunk(chunk, max_size))
self.assertEquals(len(rechunked), len(args), rechunked)
for index, chunk in enumerate(rechunked):
self.assertEquals(args[index], chunk)
def test_rechunking(self):
chunk = _PartUploadMetadata('foo', 0, 100)
self.assertRechunked(chunk, 50,
_PartUploadMetadata('foo', 0, 50),
_PartUploadMetadata('foo', 50, 50))
self.assertRechunked(chunk, 40,
_PartUploadMetadata('foo', 0, 25),
_PartUploadMetadata('foo', 25, 25),
_PartUploadMetadata('foo', 50, 25),
_PartUploadMetadata('foo', 75, 25))
self.assertRechunked(chunk, 51,
_PartUploadMetadata('foo', 0, 50),
_PartUploadMetadata('foo', 50, 50))
self.assertRechunked(chunk, 49,
_PartUploadMetadata('foo', 0, 25),
_PartUploadMetadata('foo', 25, 25),
_PartUploadMetadata('foo', 50, 25),
_PartUploadMetadata('foo', 75, 25))
self.assertRechunked(chunk, 99,
_PartUploadMetadata('foo', 0, 50),
_PartUploadMetadata('foo', 50, 50))
self.assertRechunked(chunk, 100,
_PartUploadMetadata('foo', 0, 100))
if __name__ == '__main__':
unittest.main()