* Increase organization max seat size from 30k to 2b (#1274)
* Increase organization max seat size from 30k to 2b
* PR review. Do not modify unless state matches expected
* Organization sync simultaneous event reporting (#1275)
* Split up azure messages according to max size
* Allow simultaneous login of organization user events
* Early resolve small event lists
* Clarify logic
Co-authored-by: Chad Scharf <3904944+cscharf@users.noreply.github.com>
* Improve readability
This comes at the cost of multiple serializations, but the
improvement in wire-time should more than make up for this
on message where serialization time matters
Co-authored-by: Chad Scharf <3904944+cscharf@users.noreply.github.com>
* Queue emails (#1286)
* Extract common Azure queue methods
* Do not use internal entity framework namespace
* Prefer IEnumerable to IList unless needed
All of these implementations were just using `Count == 1`,
which is easily replicated. This will be used when abstracting Azure queues
* Add model for azure queue message
* Abstract Azure queue for reuse
* Creat service to enqueue mail messages for later processing
Azure queue mail service uses Azure queues.
Blocking just blocks until all the work is done -- This is
how emailing works today
* Provide mail queue service to DI
* Queue organization invite emails for later processing
All emails can later be added to this queue
* Create Admin hosted service to process enqueued mail messages
* Prefer constructors to static generators
* Mass delete organization users (#1287)
* Add delete many to Organization Users
* Correct formatting
* Remove erroneous migration
* Clarify parameter name
* Formatting fixes
* Simplify bump account revision sproc
* Formatting fixes
* Match file names to objects
* Indicate if large import is expected
* Early pull all existing users we were planning on inviting (#1290)
* Early pull all existing users we were planning on inviting
* Improve sproc name
* Batch upsert org users (#1289)
* Add UpsertMany sprocs to OrganizationUser
* Add method to create TVPs from any object.
Uses DbOrder attribute to generate.
Sproc will fail unless TVP column order matches that of the db type
* Combine migrations
* Correct formatting
* Include sql objects in sql project
* Keep consisten parameter names
* Batch deletes for performance
* Correct formatting
* consolidate migrations
* Use batch methods in OrganizationImport
* Declare @BatchSize
* Transaction names limited to 32 chars
Drop sproc before creating it if it exists
* Update import tests
* Allow for more users in org upgrades
* Fix formatting
* Improve class hierarchy structure
* Use name tuple types
* Fix formatting
* Front load all reflection
* Format constructor
* Simplify ToTvp as class-specific extension
Co-authored-by: Chad Scharf <3904944+cscharf@users.noreply.github.com>
* Get limited life attachment download URL
This change limits url download to a 1min lifetime.
This requires moving to a new container to allow for non-public blob
access.
Clients will have to call GetAttachmentData api function to receive the download
URL. For backwards compatibility, attachment URLs are still present, but will not
work for attachments stored in non-public access blobs.
* Make GlobalSettings interface for testing
* Test LocalAttachmentStorageService equivalence
* Remove comment
* Add missing globalSettings using
* Simplify default attachment container
* Default to attachments containe for existing methods
A new upload method will be made for uploading to attachments-v2.
For compatibility for clients which don't use these new methods, we need
to still use the old container. The new container will be used only for
new uploads
* Remove Default MetaData fixture.
* Keep attachments container blob-level security for all instances
* Close unclosed FileStream
* Favor default value for noop services
Following the paradigms illustrated in "Working Effectively with Legacy
Code", this commit introduces at least one test for each service class
implementation. This test is a simple construction test -- we just
create each service and assert that it exists. Each test suite includes
a comment instructing the developer who comes next to remove the
constructor test. We don't want to keep these tests as the codebase
matures, as they aren't useful in the longterm. They only prove that we
have that class under test.
Where test suites failed to construct their associated classes, we skip
the test but leave behind the implementation. This is by design, so that
as the constructors for those classes change, we are forced to keep the
test suite current by leaning on the compiler.
