1004 Commits (eae90cbc2b21455d1e805f146b369f41a5940763)

Author	SHA1	Message	Date
Jens Ulferts	0d6f420afc	Revert "less magic on default query columns" ... This reverts commit 2293c3735b.	9 years ago
Jens Ulferts	2f764b84b9	Revert "less magic on default query columns" ... This reverts commit 2293c3735b.	9 years ago
Jens Ulferts	2293c3735b	less magic on default query columns ... Use the columns defined to be default for a new query as a default for a new query upon initialization. A caller can chose to overwrite this. This prevents the columns from getting readded when no column is selected. It also allows to improve the performance of the work_packages#index call because an SQL query with less columns is considerably faster.	9 years ago
Jens Ulferts	d060d251bf	eager load in decorator ... Eager loads the represented work packages based on a class variable. This might look like it would break encapsulation at first and it does. When it comes to eager loading one has to break encapsulation at some point. At least this way, we are very near to what we are going to represent and thus know which models we need. Additionally, this enables plugins to register for being eager loaded	9 years ago
Jens Ulferts	2cb05bfc27	remove no longer necessary scope alteration	9 years ago
Jens Ulferts	32b3b9ab9b	replace html path in api representer ... The html path is no longer required as we now have a proper api endpoint a client can call	9 years ago
Jens Ulferts	3de1b3583d	use the api v3 path helper	9 years ago
Jens Ulferts	cadc397239	create all custom field acessors on representer injection	9 years ago
Jens Ulferts	e1b3845582	cache classes with dynamically injected custom attributes	9 years ago
Jens Ulferts	af09bf7017	cache root_path in apiv3 api helper ... A lot of time was spent determining the root path.	9 years ago
Jens Ulferts	7a06b3451b	eager load instances used in the representer ... From an architecture standpoint, this is bad design as it breaks encapsulation. I have to know what the objects I call do in order to have eager loading. From a performance standpoint, this is necessary.	9 years ago
Jens Ulferts	0267938c36	introduce spent_time model	9 years ago
Oliver Günther	f03728c156	Retain partial message for custom field to customizable	9 years ago
Jens Ulferts	fc49e2a7b7	use linked parent resource instead of parentId ... The parentId field should be removed. But we shouldn't do it in a minor version.	9 years ago
Jens Ulferts	9668e7e99e	fix addChild path in work package resource	9 years ago
Jens Ulferts	966d749e45	add parentId field to schema ... This is somewhat debateable. Having a parentId field is wrong in itself. It should be a linked resource. But as the field is also present in the resource, it also needs to be part of the schema for consistency sake.	9 years ago
Markus Kahl	7f88ba5b22	removed attachments#show; only #download remains	9 years ago
Oliver Günther	6259ab433d	Remove embedded activities	9 years ago
Jens Ulferts	932860abd2	only make second request_body merge run on type change	9 years ago
Jens Ulferts	e2f8db9497	eager load colors together with the type	9 years ago
Jens Ulferts	cbbd7ea6e5	apply eager loading to work_packages#activities	9 years ago
Jens Ulferts	f59bbe4008	only need the user_id not the user.id	9 years ago
Jens Ulferts	8065c66256	eager loading associated objects, the dirty way	9 years ago
Jens Ulferts	8284933a5a	mark spentTime and percentDone as optional	9 years ago
Oliver Günther	69fae7bc54	Remove other remnants	9 years ago
Oliver Günther	e996167b55	Remove work_package route, rename appropriately	9 years ago
Oliver Günther	876203c096	Add writable to other custom field injectors ... This was missed in https://github.com/opf/openproject/pull/3728 and properly disguised by a faulty spec fix from my side.	9 years ago
Jens Ulferts	27e2cce32b	add priority to list of attributes inherited from children	9 years ago
Jan Sandbrink	85f60d8bbc	Also allow to add attachments to a WP with permission to add WPs ... This is neccessary to allow adding attachments, after creating a new work package. The undesired part is, that it also allows you to add attachments long after creating the WP.	9 years ago
Oliver Günther	acb4431d8a	Implement UserPreferences endpoint ... Provides the implementation to the user preferences endpoint.	9 years ago
Jens Ulferts	757f6a53d7	eager load preference to avoid N+1 queries	9 years ago
Jan Sandbrink	58aa36ae2a	Check for nil representers explicitly ... Passing nil as model into a representer happens usually only due to an error. If a subclass needs to be able to pass nil as represented, it can specify that by overrding. This however, should be an exception, not the rule.	9 years ago
Oliver Günther	de96916dac	Minor corrections ... This commit addresses the following minor corrections from review: 1. Remove constant URI: It breaks karma, and I can't really tell why 2. Avoid using Object.values Appeared to be supported by chrome, but causes the protractor failure. This also switches the ordering to binary (unshift, push) as that's all we need for now. 3. Remove location.hash when jumping to element `$location.hash` may take precedence over `scrollIntoView` ocassionaly, thus we forcefully remove the location hash when updating the work package on a particular field 4. Avoid showing comment success when empty When opening the comment field, and then submitting on another field with comment being empty and open, the comment was not submitted, but a successful comment notification was still sent. 5. Fix issues for IE 10+ * Force min-height to be set on the element, as IE tends to return 0 as height when empty 6. Misc. * Avoid undefined in comment text * Avoid resizing [msd-elastic] textareas * Rename `add_journal` to `create_journal` Correct wrong bracket for reject function	9 years ago
Oliver Günther	3ca36e1e1d	Parallel sending of comment and fields in WP split view ... Parallel submitting of comments alongside WP edits wasn't possible natively due to the increased lock_version in work packages, caused by `touch`ing and explicitly saving the work package in the process of adding a journal. This commit works around this issue by updating the timestamp columns directly, bypassing optimistic locking in the work package table. This allows comments to be submitted transparently with other fields to the user, while actually, we send it in two subsequent steps.	9 years ago
Oliver Günther	0883cf9834	Replace exclusive-edit with transcluded comment directive ... This commit removes the somewhat superfluous exlusiveEditController in favor of event-based passing of quoted text to the correct comment field. The comment directive now decides for itself whether it wants to be placed above or below the activities and uses transclusion for that purpose. Editing comments is now no longer exlusive, but that shouldn't be necessary anyway.	9 years ago
Oliver Günther	3bfad28db0	Correct comment behavior with optional notifications ... This allows submitting comments with and without notification for the created journal. Using hyperagent resource on activities, the resource `addComment` is replaced when executing `fetch` on the link. This breaks subsequent comment requests. This is fixed by requesting manually on the resource link using `$http`. This commit also provides the functionality required to comment with or without notification propagation, and changes the backend to comply to the APIv3 specification, which defines comments to be sent as a JSON `{ comment: { raw: 'my comment' } }`. This commit also provides: * Displaying the comment input on overview depending on sort setting * Disable inplace controls when comment is empty * correctly hiding the comment based on link availability	9 years ago
Alex Coles	2f659e76a9	Fix deprecated Model.exists? arg in UpdateContract ... Signed-off-by: Alex Coles <alex@alexbcoles.com>	9 years ago
Jan Sandbrink	0702683116	Fix bad method call for version custom fields ... exploring the down sides of isolated unit testing :(	9 years ago
Jan Sandbrink	338f9d40d0	Fix (untested) relation representer ... This one still is not very APIv3, but there ain't no time to improve it... The angular frontend also seems to rely on the quirks with the relation type changing depending which work package embeds the relation.	9 years ago
Jan Sandbrink	88ed4d2df9	Promote ::create to a common interface method ... This one is already used by the representers that require a custom field injector. However, by having a separate method this also means collections (or the new DecoratorFactory) need to take care and call the ::create method instead of ::new. I therefore made create the default interface to create representers, which is now used by Factories and Collections. Note: I did not yet change all representer instantiations to ::create. I hope this will happen gradually.	9 years ago
Jan Sandbrink	6e9ed008fb	Enforce passing of current_user into representers ... Thereby permission checks will no longer silently fail if no user was passed. This uncovered a lot of places where we actually expected a user to be passed, when it wasn't. This was especially true for collections. The basic idea is, that virtually all representers could change their behaviour based on the current user. Representers where this is not the case (e.g. Digest), can still explicitly opt-out.	9 years ago
Jan Sandbrink	f121e610db	Pass current_user into all AttachmentRepresenters	9 years ago
Jan Sandbrink	ab9d502009	Allow to delete work packages via APIv3	9 years ago
Jan Sandbrink	3f269e43ae	Add link to delete an attachment to its representer	9 years ago
Jan Sandbrink	e7a6ae629e	Add specs for update links with missing permissions ... - also (for now) remove title text which was not localized - consistently use link helpers for the corresponding specs	9 years ago
Jan Sandbrink	aad39544f5	Hide addAttachment link when user is not allowed	9 years ago
Jan Sandbrink	5de0f15898	Correctly link activities ... - they might not be embedded - create an endpoint for activities - document it properly - adapt angular client	9 years ago
Jan Sandbrink	14a55b82d1	Do not embed linked resources in WP collections	9 years ago
Jan Sandbrink	5d473e9a9c	fix HTTP response of DELETE attachment ... this was wrongfully HTTP 202 (Accepted), which would only be valid if we did the deletion in a delayed/deferred way. For now using HTTP 204 (No Content), since we have no response body.	9 years ago
Jens Ulferts	db3681661d	fixes attachment download path in api ... Kudos to @oliverguenter for the fix	9 years ago