sachet-server/TODO.txt

[ ]: not started
[-]: in progress
[x]: done
[!]: wontfix
[?]: maybe

[x] authentication
	[x] /users/login POST
	[x] /users/logout POST
	[x] /users/extend POST

[x] permissions
	[x] change the db schema
		[x] use a library for serializing classes
		[x] permissions are:
			- create file
			- modify any file
			- delete any file
			- lock/unlock any file
				no modification, no deletion
			- list all files
			- admin
				doesn't imply other privileges directly,
				but means you can set them to yourself
	[x] /users/<user> GET
		this endpoint will now contain permission data too
	[x] /users/<user> PATCH
	[x] /users/<user> PUT
		we'll use these to set permissions

	[x] make the "generic model api"
		endpoints like /users/<id> and /admin/settings
		that directly get/set models will inherit from this class
		any authentication/permission logic belongs to the children class

		[x] make schemas members of their parent models

	[x] settings endpoint
		[x] initialize settings to defaults

		[x] tests
		[x] /admin/settings GET
		[x] /admin/settings PATCH
		[x] /admin/settings PUT
			settings will set the default non-authenticated user's permissions

[x] linter

[x] file management backend
	we want to abstract away the filesystem
	[x] write todos for this entry
		read, write, list
		but how are we going to implement metadata and stuff?
		are we going to use UUIDs?
		hashing?

	[x] create the module sachet.storage
		[x] class Storage
			generic storage
			- get handle to file
			- list files
		[x] class File
			- delete file
			- r/w file
			- read metadata
			- write metadata
			- rename file

		our "real" classes will inherit from the Storage interface
		and they will replace all the methods (and the File class)

		[x] class FileSystem
			just hook up python's os filesystem to FileSystem

			[x] rewrite tests
			[x] class File
				- create file and its associated metadata file
				- delete file
				- r/w file
				- read metadata
				- write metadata
				- rename file

[x] file sharing
	[x] Share model
		[x] docstring

	[x] auth_required: require certain permissions

	metadata
		[x] /files POST
			client will post metadata as json
			returns a 201 and set Location header to the created entry
			depends on create file permission
		[x] /files/<file> PUT
			metadata update
		[x] /files/<file> PATCH
			metadata update
		[x] /files/<file> GET
			metadata get
		[x] /files/<file> DELETE
			deletes file

	files
		[x] /files/<file>/content POST
			uploads file
		[x] /files/<file>/content PUT
			modifies file
		[x] /files/<file>/content GET
			downloads file, depends on read permission

	[x] debugging

	[x] testing

	[x] metadata
		[x] implement a filename for Share
		[x] test filename

	[x] test modification

	[x] implement listing files
		[x] split POST off ModelClass
			one class will be for editing models themselves
			other class, ModelListClass
				POST creates a new Model
				GET lists all models (paginated)
		[x] implement ModelListClass GET (pagination!)

	[x] implement anonymous permissions
		[x] test adversarial conditions

	[x] fix that one bug where sending a non-int page number causes 500
		[x] write a test case to prevent it

	[x] fix modification
		supposedly users should only be able to modify their own pastes

	locking
		[x] add locked attribute to Share
		[x] implement authorization to modify/delete
		[x] implement changing lock/unlock state
			setting a bool flag in /files/<file> PUT/PATCH/POST
			has the be authorized, though

[!] access docstrings without starting up the webserver
	https://stackoverflow.com/questions/18214612/how-to-access-app-config-in-a-blueprint
	instead of importing app, import this

[x] database migrations
[x] database cleanup
	move all the cleanup to a flask cmd
	"delete where date < expiry"

[x] write in README about db cleanup and migrations

[x] implement chunked upload
	[x] Upload model
		does not have a REST API endpoint
		[x] upload_id (primary key a.k.a. dzuuid)
		[x] backref to Share
		[x] list of Chunks
		[x] datetime
	[x] Chunk model
		does not have a REST API endpoint
		[x] id (autoincrement)
		[x] index
		[x] backref to Upload
		[x] datetime

	we will store chunks as individual files: share uuid + suffix
	suffix is
	- upload uuid (prevents race condition for two concurrent uploads)
	- chunk index
	
	[x] write tests for chunked upload
	[?] write more rigorous tests
		poke at the chunks themselves
		what happens when you send chunks in a race condition,
		at the wrong time, with the wrong permissions, etc?

	[x] investigate why share_id is set to none
	[x] clear out chunk files after they're used
	[x] clear out files after they are deleted

[x] implement chunked download
	[x] implement File.size
		[x] write tests
	[x] write test for Range
	[x] implement HTTP 416
		[x] write appropriate tests

[x] write periodic cleanup for crusty stale uploads and chunks
	so right now the issue is that the chunk on_delete event isn't triggering
	and we're leaving random files on the disk

[x] add /users API
	[x] fix url for users (returned when posting)
		[x] create test case
	[x] add /users DELETE
		[x] added test case
		[x] add a note for this in the docs
	[x] password change endpoint
		[x] tests
		[x] docs

[x] investigate "FOREIGN KEY constraint failed"

[x] investigate what happens when you change ownership of a share to a
	non-existent user
	add a note maybe to the docs
	[x] add note about ownership transfers
	[x] add note that setting owner to null allows anon users to own it

[x] make sure that reserved names like `login`, `renew` are actually necessary
	right now we're only implementing them on the CLI interface
	maybe it doesn't matter because we're using POST only for these
	and users don't have POST
	[x] if they're not necessary we'll tear out the safeguards
	[x] also add a warning to related endpoints to be careful

[x] fix bug where users can modify other users' shares' metadata

[x] proper documentation
	[x] use a linter on docstrings
	[x] Authentication
	[x] Paginated APIs
	[x] Permissions
	[x] User API
		[x] url_for docs
	[x] Admin API
		[x] anon perms
	[x] Files API
		[x] metadata API
		[x] list API
			[x] url_for docs
		[x] content API
			[x] chunked upload protocol
			[x] POST
			[x] PUT
			[x] GET
		[x] lock API
			(reference this in permissions)
			(reference this in the files schema)
	[x] cli
	[x] getting started (dev)

[x] implement /whoami endpoint
	[x] tests
	[x] docs

[ ] expose filesize in share info

[ ] investigate cleanup being in the user subcmd
[ ] investigate cleanup cmd triggering foreign key failure

[ ] if you create a new user without a required field it gives 500
[ ] if you create a new user with the same name as an existing one it gives 500
[x] if you try to read a share with an invalid uuid it gives 500

[ ] investigate if you can interfere with an upload by setting the same id

[ ] prod deployment config WSGI
	[ ] write info about this in docs